What is Secondary 3 English Vocabulary | Artificial Intelligence, Technology and Digital Literacy for Grade 9

Secondary 3 English vocabulary for artificial intelligence, technology and digital literacy gives Grade 9 students the language to discuss AI systems, generative AI, algorithms, data, bias, privacy, verification, plagiarism, prompts, models and responsible use with more precision than broad terms such as “smart,” “robot” or “computer.” Students searching for Secondary 3 vocabulary, Grade 9 AI vocabulary, artificial intelligence vocabulary, generative AI vocabulary, technology vocabulary, digital literacy vocabulary, machine learning terms and 9th grade AI literacy need a current conceptual system rather than a fashionable word list.

Current student AI-literacy frameworks separate several important dimensions. UNESCO’s AI Competency Framework for Students organises learning around a human-centred mindset, ethics of AI, AI techniques and applications, and AI system design, with progression from understanding to applying and creating. NIST’s current terminology distinguishes generative artificial intelligence as systems that generate synthetic content and uses confabulation for outputs that confidently present erroneous or false content, often colloquially called hallucinations. These terms matter because students need to discuss both capability and limitation.

This article is the technology-and-AI child of eduKateSG’s Secondary 3 English Vocabulary apex. It does not turn Secondary 3 English into a computer-science course. This page owns AI and digital-literacy vocabulary for reading, writing, discussion, source evaluation and responsible school use: enough technical precision to reason clearly about contemporary technology while keeping language accessible to Grade 9 learners.

The 50-Second Route

When discussing AI, separate the system, data, task, output, evaluation and responsibility. Ask what the tool is designed to do, what information it uses, what it produces, how the output is checked, what risks exist, and which human remains responsible for the final decision.

• AI system: the overall computational system performing tasks associated with intelligent behaviour.
• Model: a learned or designed computational representation used to produce predictions or outputs.
• Training data: information used to fit or develop a model.
• Prompt/input: information given to a system.
• Output: the system’s response or generated result.
• Generative AI: AI designed to generate synthetic text, images, audio, video or other content.
• Evaluation: testing output quality, accuracy, usefulness and risk.
• Human oversight: people reviewing, directing or intervening in system use.
• Accountability: responsibility for decisions and consequences.
• Verification: checking claims and outputs rather than assuming correctness.

AI literacy improves when students can distinguish what a system generates from what a human has verified.

AI Is a Broad Category

Artificial intelligence is an umbrella term covering many systems and tasks. A recommendation system, computer-vision model, speech recogniser and generative language model may all be called AI, but they behave differently. Students should avoid treating AI as one single machine or mind.

The most useful vocabulary separates systems by purpose and method. Some models classify, some predict, some rank, some generate. A generative system produces new synthetic content based on learned statistical patterns. That capability does not guarantee factual accuracy, reasoning quality or appropriateness.

Clear vocabulary prevents anthropomorphic shortcuts from replacing explanation.

Generative AI and Confabulation

NIST defines generative AI as a class of AI models that generate derived synthetic content such as text, images, video or audio. Generative outputs can be useful, creative and fluent, but fluency should not be confused with truth.

NIST’s generative-AI risk profile uses the term confabulation for confidently generated erroneous or false content. The word hallucination is common in public discussion, but confabulation usefully reminds students that the issue is output accuracy, not a human mental state.

The practical lesson is verification. Students should independently check claims, quotations, references, dates and specialised information before relying on generated output.

40 AI, Technology and Digital Literacy Vocabulary Units

1. artificial intelligence

Meaning: computer systems performing tasks associated with perception, learning, reasoning, communication or action. Natural phrase: AI system; artificial intelligence application. AI-literacy job: core. Precision note: AI is a broad category, not one single technology. Create one school or everyday example where the term applies and one example that is often confused with it.

2. generative AI

Meaning: AI models designed to generate synthetic content. Natural phrase: generative AI model; GenAI output. AI-literacy job: core. Precision note: Generation is different from retrieval or classification. Create one school or everyday example where the term applies and one example that is often confused with it.

3. machine learning

Meaning: methods in which systems learn patterns from data for tasks. Natural phrase: machine-learning model; train a model. AI-literacy job: technique. Precision note: Not every software rule is machine learning. Create one school or everyday example where the term applies and one example that is often confused with it.

4. model

Meaning: computational representation used to make predictions or generate outputs. Natural phrase: AI model; trained model. AI-literacy job: system. Precision note: A model is not the same as the full product or service. Create one school or everyday example where the term applies and one example that is often confused with it.

5. algorithm

Meaning: set of procedures or computational steps. Natural phrase: ranking algorithm; algorithmic process. AI-literacy job: system. Precision note: Algorithm is broader than AI. Create one school or everyday example where the term applies and one example that is often confused with it.

6. training data

Meaning: data used to fit or develop a model. Natural phrase: training dataset; training-data quality. AI-literacy job: data. Precision note: Data quality and representativeness matter. Create one school or everyday example where the term applies and one example that is often confused with it.

7. dataset

Meaning: organised collection of data. Natural phrase: large dataset; labelled dataset. AI-literacy job: data. Precision note: A dataset is not automatically unbiased or accurate. Create one school or everyday example where the term applies and one example that is often confused with it.

8. input

Meaning: information given to a system. Natural phrase: user input; input data. AI-literacy job: interaction. Precision note: Input may include text, image, audio or structured data. Create one school or everyday example where the term applies and one example that is often confused with it.

9. prompt

Meaning: instruction or context given to a generative system. Natural phrase: write a prompt; prompt design. AI-literacy job: interaction. Precision note: Prompt quality can affect output but cannot guarantee truth. Create one school or everyday example where the term applies and one example that is often confused with it.

10. output

Meaning: result produced by a system. Natural phrase: model output; generated output. AI-literacy job: interaction. Precision note: Output still needs evaluation. Create one school or everyday example where the term applies and one example that is often confused with it.

11. token

Meaning: small unit used in text processing by many language models. Natural phrase: input tokens; token sequence. AI-literacy job: technical. Precision note: Tokens are not always whole words. Create one school or everyday example where the term applies and one example that is often confused with it.

12. parameter

Meaning: learned or configured value influencing model behaviour. Natural phrase: model parameter; parameter count. AI-literacy job: technical. Precision note: Parameter count alone does not determine usefulness. Create one school or everyday example where the term applies and one example that is often confused with it.

13. inference

Meaning: running a trained model to produce a prediction/output. Natural phrase: model inference; inference time. AI-literacy job: technical. Precision note: Different from reading-comprehension inference despite same word. Create one school or everyday example where the term applies and one example that is often confused with it.

14. classification

Meaning: assigning inputs to categories. Natural phrase: image classification; classify text. AI-literacy job: task. Precision note: Different from generation. Create one school or everyday example where the term applies and one example that is often confused with it.

15. prediction

Meaning: estimating an outcome or value. Natural phrase: predictive model; make a prediction. AI-literacy job: task. Precision note: Prediction is probabilistic, not certainty. Create one school or everyday example where the term applies and one example that is often confused with it.

16. recommendation system

Meaning: system ranking or suggesting items to users. Natural phrase: recommendation algorithm; recommended content. AI-literacy job: task. Precision note: Visibility may reflect ranking criteria, not truth or importance. Create one school or everyday example where the term applies and one example that is often confused with it.

17. automation

Meaning: use of systems to perform tasks with reduced direct human action. Natural phrase: automate a task; automation. AI-literacy job: application. Precision note: Automation does not necessarily mean AI. Create one school or everyday example where the term applies and one example that is often confused with it.

18. synthetic content

Meaning: content generated computationally rather than directly captured from reality. Natural phrase: synthetic image; synthetic media. AI-literacy job: output. Precision note: Synthetic does not automatically mean deceptive. Create one school or everyday example where the term applies and one example that is often confused with it.

19. deepfake

Meaning: synthetic/manipulated media made to resemble real people/events. Natural phrase: deepfake video; detect deepfake. AI-literacy job: risk. Precision note: Verification and disclosure matter. Create one school or everyday example where the term applies and one example that is often confused with it.

20. confabulation

Meaning: generated false or erroneous content presented confidently. Natural phrase: AI confabulation; verify confabulated output. AI-literacy job: risk. Precision note: Often colloquially called hallucination. Create one school or everyday example where the term applies and one example that is often confused with it.

21. hallucination

Meaning: common informal term for AI-generated false or unsupported content. Natural phrase: AI hallucination; hallucinated citation. AI-literacy job: risk. Precision note: Use with awareness that technical bodies may prefer confabulation. Create one school or everyday example where the term applies and one example that is often confused with it.

22. bias

Meaning: systematic tendency that can create skewed outcomes. Natural phrase: algorithmic bias; data bias. AI-literacy job: risk. Precision note: Bias can enter through data, design, measurement or deployment. Create one school or everyday example where the term applies and one example that is often confused with it.

23. fairness

Meaning: principles and methods for avoiding unjustified disparities. Natural phrase: AI fairness; fairness assessment. AI-literacy job: ethics. Precision note: Different fairness definitions can conflict. Create one school or everyday example where the term applies and one example that is often confused with it.

24. transparency

Meaning: openness about system purpose, limits, data or decision processes. Natural phrase: AI transparency; transparent use. AI-literacy job: ethics. Precision note: Transparency can support informed evaluation. Create one school or everyday example where the term applies and one example that is often confused with it.

25. explainability

Meaning: ability to provide understandable reasons or accounts of system behaviour. Natural phrase: model explainability; explainable AI. AI-literacy job: ethics. Precision note: What counts as an explanation depends on audience and task. Create one school or everyday example where the term applies and one example that is often confused with it.

26. privacy

Meaning: control/protection of personal information. Natural phrase: data privacy; privacy risk. AI-literacy job: ethics. Precision note: Students should consider what data they provide. Create one school or everyday example where the term applies and one example that is often confused with it.

27. personal data

Meaning: information relating to an identifiable person. Natural phrase: personal-data protection; share personal data. AI-literacy job: data. Precision note: Avoid unnecessary disclosure to tools. Create one school or everyday example where the term applies and one example that is often confused with it.

28. security

Meaning: protection against unauthorised access, misuse or attack. Natural phrase: AI security; cybersecurity. AI-literacy job: risk. Precision note: Security and privacy overlap but are not identical. Create one school or everyday example where the term applies and one example that is often confused with it.

29. accountability

Meaning: responsibility for decisions and consequences. Natural phrase: human accountability; accountability mechanism. AI-literacy job: ethics. Precision note: Responsibility should not disappear because AI was involved. Create one school or everyday example where the term applies and one example that is often confused with it.

30. human oversight

Meaning: human review, direction or intervention in system operation. Natural phrase: human oversight; human review. AI-literacy job: governance. Precision note: Important for consequential decisions. Create one school or everyday example where the term applies and one example that is often confused with it.

31. human agency

Meaning: ability of people to make meaningful choices and retain control. Natural phrase: preserve human agency; human-centred design. AI-literacy job: ethics. Precision note: UNESCO emphasises human-centred AI literacy. Create one school or everyday example where the term applies and one example that is often confused with it.

32. plagiarism

Meaning: using another’s work or ideas without appropriate acknowledgement. Natural phrase: avoid plagiarism; plagiarism policy. AI-literacy job: school use. Precision note: AI use can create attribution questions; follow school rules. Create one school or everyday example where the term applies and one example that is often confused with it.

33. attribution

Meaning: identifying source or contribution. Natural phrase: proper attribution; attribute a source. AI-literacy job: school use. Precision note: Attribution practices depend on task and institution. Create one school or everyday example where the term applies and one example that is often confused with it.

34. citation

Meaning: reference identifying a source. Natural phrase: verify citation; cite a source. AI-literacy job: school use. Precision note: Generated citations may be false; verify them. Create one school or everyday example where the term applies and one example that is often confused with it.

35. provenance

Meaning: record of origin/history of content or data. Natural phrase: content provenance; data provenance. AI-literacy job: verification. Precision note: Useful for synthetic media and sources. Create one school or everyday example where the term applies and one example that is often confused with it.

36. verification

Meaning: checking accuracy/authenticity. Natural phrase: verify output; independent verification. AI-literacy job: practice. Precision note: Never treat fluent output as self-verifying. Create one school or everyday example where the term applies and one example that is often confused with it.

37. evaluation

Meaning: systematic assessment of quality or performance. Natural phrase: model evaluation; evaluate accuracy. AI-literacy job: practice. Precision note: Metrics must match task and risk. Create one school or everyday example where the term applies and one example that is often confused with it.

38. accuracy

Meaning: degree to which output is correct under defined criteria. Natural phrase: accuracy rate; factual accuracy. AI-literacy job: quality. Precision note: Correctness depends on task definition. Create one school or everyday example where the term applies and one example that is often confused with it.

39. reliability

Meaning: consistency/dependability of performance. Natural phrase: reliable system; reliability testing. AI-literacy job: quality. Precision note: Reliable output can still be systematically wrong. Create one school or everyday example where the term applies and one example that is often confused with it.

40. responsible use

Meaning: use that considers accuracy, privacy, fairness, rules and human responsibility. Natural phrase: responsible AI use; responsible practice. AI-literacy job: ethics. Precision note: Context determines appropriate use. Create one school or everyday example where the term applies and one example that is often confused with it.

Eight AI-Literacy Cycles

AI Literacy Cycle 1

artificial intelligence — Describe a simple AI use case and identify where the target appears in the system. Keep “computer systems performing tasks associated with perception, learning, reasoning, communication or action” as the meaning anchor, “AI system; artificial intelligence application” as natural language and “core” as the conceptual role. AI is a broad category, not one single technology.

generative AI — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “AI models designed to generate synthetic content” as the meaning anchor, “generative AI model; GenAI output” as natural language and “core” as the conceptual role. Generation is different from retrieval or classification.

machine learning — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “methods in which systems learn patterns from data for tasks” as the meaning anchor, “machine-learning model; train a model” as natural language and “technique” as the conceptual role. Not every software rule is machine learning.

model — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “computational representation used to make predictions or generate outputs” as the meaning anchor, “AI model; trained model” as natural language and “system” as the conceptual role. A model is not the same as the full product or service.

algorithm — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “set of procedures or computational steps” as the meaning anchor, “ranking algorithm; algorithmic process” as natural language and “system” as the conceptual role. Algorithm is broader than AI.

training data — Use the target in a media-literacy context involving synthetic content or online claims. Keep “data used to fit or develop a model” as the meaning anchor, “training dataset; training-data quality” as natural language and “data” as the conceptual role. Data quality and representativeness matter.

dataset — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “organised collection of data” as the meaning anchor, “large dataset; labelled dataset” as natural language and “data” as the conceptual role. A dataset is not automatically unbiased or accurate.

input — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “information given to a system” as the meaning anchor, “user input; input data” as natural language and “interaction” as the conceptual role. Input may include text, image, audio or structured data.

prompt — Describe a simple AI use case and identify where the target appears in the system. Keep “instruction or context given to a generative system” as the meaning anchor, “write a prompt; prompt design” as natural language and “interaction” as the conceptual role. Prompt quality can affect output but cannot guarantee truth.

output — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “result produced by a system” as the meaning anchor, “model output; generated output” as natural language and “interaction” as the conceptual role. Output still needs evaluation.

token — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “small unit used in text processing by many language models” as the meaning anchor, “input tokens; token sequence” as natural language and “technical” as the conceptual role. Tokens are not always whole words.

parameter — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “learned or configured value influencing model behaviour” as the meaning anchor, “model parameter; parameter count” as natural language and “technical” as the conceptual role. Parameter count alone does not determine usefulness.

inference — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “running a trained model to produce a prediction/output” as the meaning anchor, “model inference; inference time” as natural language and “technical” as the conceptual role. Different from reading-comprehension inference despite same word.

classification — Use the target in a media-literacy context involving synthetic content or online claims. Keep “assigning inputs to categories” as the meaning anchor, “image classification; classify text” as natural language and “task” as the conceptual role. Different from generation.

prediction — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “estimating an outcome or value” as the meaning anchor, “predictive model; make a prediction” as natural language and “task” as the conceptual role. Prediction is probabilistic, not certainty.

recommendation system — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “system ranking or suggesting items to users” as the meaning anchor, “recommendation algorithm; recommended content” as natural language and “task” as the conceptual role. Visibility may reflect ranking criteria, not truth or importance.

automation — Describe a simple AI use case and identify where the target appears in the system. Keep “use of systems to perform tasks with reduced direct human action” as the meaning anchor, “automate a task; automation” as natural language and “application” as the conceptual role. Automation does not necessarily mean AI.

synthetic content — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “content generated computationally rather than directly captured from reality” as the meaning anchor, “synthetic image; synthetic media” as natural language and “output” as the conceptual role. Synthetic does not automatically mean deceptive.

deepfake — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “synthetic/manipulated media made to resemble real people/events” as the meaning anchor, “deepfake video; detect deepfake” as natural language and “risk” as the conceptual role. Verification and disclosure matter.

confabulation — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “generated false or erroneous content presented confidently” as the meaning anchor, “AI confabulation; verify confabulated output” as natural language and “risk” as the conceptual role. Often colloquially called hallucination.

hallucination — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “common informal term for AI-generated false or unsupported content” as the meaning anchor, “AI hallucination; hallucinated citation” as natural language and “risk” as the conceptual role. Use with awareness that technical bodies may prefer confabulation.

bias — Use the target in a media-literacy context involving synthetic content or online claims. Keep “systematic tendency that can create skewed outcomes” as the meaning anchor, “algorithmic bias; data bias” as natural language and “risk” as the conceptual role. Bias can enter through data, design, measurement or deployment.

fairness — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “principles and methods for avoiding unjustified disparities” as the meaning anchor, “AI fairness; fairness assessment” as natural language and “ethics” as the conceptual role. Different fairness definitions can conflict.

transparency — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “openness about system purpose, limits, data or decision processes” as the meaning anchor, “AI transparency; transparent use” as natural language and “ethics” as the conceptual role. Transparency can support informed evaluation.

explainability — Describe a simple AI use case and identify where the target appears in the system. Keep “ability to provide understandable reasons or accounts of system behaviour” as the meaning anchor, “model explainability; explainable AI” as natural language and “ethics” as the conceptual role. What counts as an explanation depends on audience and task.

privacy — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “control/protection of personal information” as the meaning anchor, “data privacy; privacy risk” as natural language and “ethics” as the conceptual role. Students should consider what data they provide.

personal data — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “information relating to an identifiable person” as the meaning anchor, “personal-data protection; share personal data” as natural language and “data” as the conceptual role. Avoid unnecessary disclosure to tools.

security — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “protection against unauthorised access, misuse or attack” as the meaning anchor, “AI security; cybersecurity” as natural language and “risk” as the conceptual role. Security and privacy overlap but are not identical.

accountability — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “responsibility for decisions and consequences” as the meaning anchor, “human accountability; accountability mechanism” as natural language and “ethics” as the conceptual role. Responsibility should not disappear because AI was involved.

human oversight — Use the target in a media-literacy context involving synthetic content or online claims. Keep “human review, direction or intervention in system operation” as the meaning anchor, “human oversight; human review” as natural language and “governance” as the conceptual role. Important for consequential decisions.

human agency — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “ability of people to make meaningful choices and retain control” as the meaning anchor, “preserve human agency; human-centred design” as natural language and “ethics” as the conceptual role. UNESCO emphasises human-centred AI literacy.

plagiarism — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “using another’s work or ideas without appropriate acknowledgement” as the meaning anchor, “avoid plagiarism; plagiarism policy” as natural language and “school use” as the conceptual role. AI use can create attribution questions; follow school rules.

attribution — Describe a simple AI use case and identify where the target appears in the system. Keep “identifying source or contribution” as the meaning anchor, “proper attribution; attribute a source” as natural language and “school use” as the conceptual role. Attribution practices depend on task and institution.

citation — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “reference identifying a source” as the meaning anchor, “verify citation; cite a source” as natural language and “school use” as the conceptual role. Generated citations may be false; verify them.

provenance — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “record of origin/history of content or data” as the meaning anchor, “content provenance; data provenance” as natural language and “verification” as the conceptual role. Useful for synthetic media and sources.

verification — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “checking accuracy/authenticity” as the meaning anchor, “verify output; independent verification” as natural language and “practice” as the conceptual role. Never treat fluent output as self-verifying.

evaluation — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “systematic assessment of quality or performance” as the meaning anchor, “model evaluation; evaluate accuracy” as natural language and “practice” as the conceptual role. Metrics must match task and risk.

accuracy — Use the target in a media-literacy context involving synthetic content or online claims. Keep “degree to which output is correct under defined criteria” as the meaning anchor, “accuracy rate; factual accuracy” as natural language and “quality” as the conceptual role. Correctness depends on task definition.

reliability — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “consistency/dependability of performance” as the meaning anchor, “reliable system; reliability testing” as natural language and “quality” as the conceptual role. Reliable output can still be systematically wrong.

responsible use — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “use that considers accuracy, privacy, fairness, rules and human responsibility” as the meaning anchor, “responsible AI use; responsible practice” as natural language and “ethics” as the conceptual role. Context determines appropriate use.

AI Literacy Cycle 2

artificial intelligence — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “computer systems performing tasks associated with perception, learning, reasoning, communication or action” as the meaning anchor, “AI system; artificial intelligence application” as natural language and “core” as the conceptual role. AI is a broad category, not one single technology.

generative AI — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “AI models designed to generate synthetic content” as the meaning anchor, “generative AI model; GenAI output” as natural language and “core” as the conceptual role. Generation is different from retrieval or classification.

machine learning — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “methods in which systems learn patterns from data for tasks” as the meaning anchor, “machine-learning model; train a model” as natural language and “technique” as the conceptual role. Not every software rule is machine learning.

model — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “computational representation used to make predictions or generate outputs” as the meaning anchor, “AI model; trained model” as natural language and “system” as the conceptual role. A model is not the same as the full product or service.

algorithm — Use the target in a media-literacy context involving synthetic content or online claims. Keep “set of procedures or computational steps” as the meaning anchor, “ranking algorithm; algorithmic process” as natural language and “system” as the conceptual role. Algorithm is broader than AI.

training data — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “data used to fit or develop a model” as the meaning anchor, “training dataset; training-data quality” as natural language and “data” as the conceptual role. Data quality and representativeness matter.

dataset — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “organised collection of data” as the meaning anchor, “large dataset; labelled dataset” as natural language and “data” as the conceptual role. A dataset is not automatically unbiased or accurate.

input — Describe a simple AI use case and identify where the target appears in the system. Keep “information given to a system” as the meaning anchor, “user input; input data” as natural language and “interaction” as the conceptual role. Input may include text, image, audio or structured data.

prompt — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “instruction or context given to a generative system” as the meaning anchor, “write a prompt; prompt design” as natural language and “interaction” as the conceptual role. Prompt quality can affect output but cannot guarantee truth.

output — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “result produced by a system” as the meaning anchor, “model output; generated output” as natural language and “interaction” as the conceptual role. Output still needs evaluation.

token — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “small unit used in text processing by many language models” as the meaning anchor, “input tokens; token sequence” as natural language and “technical” as the conceptual role. Tokens are not always whole words.

parameter — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “learned or configured value influencing model behaviour” as the meaning anchor, “model parameter; parameter count” as natural language and “technical” as the conceptual role. Parameter count alone does not determine usefulness.

inference — Use the target in a media-literacy context involving synthetic content or online claims. Keep “running a trained model to produce a prediction/output” as the meaning anchor, “model inference; inference time” as natural language and “technical” as the conceptual role. Different from reading-comprehension inference despite same word.

classification — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “assigning inputs to categories” as the meaning anchor, “image classification; classify text” as natural language and “task” as the conceptual role. Different from generation.

prediction — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “estimating an outcome or value” as the meaning anchor, “predictive model; make a prediction” as natural language and “task” as the conceptual role. Prediction is probabilistic, not certainty.

recommendation system — Describe a simple AI use case and identify where the target appears in the system. Keep “system ranking or suggesting items to users” as the meaning anchor, “recommendation algorithm; recommended content” as natural language and “task” as the conceptual role. Visibility may reflect ranking criteria, not truth or importance.

automation — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “use of systems to perform tasks with reduced direct human action” as the meaning anchor, “automate a task; automation” as natural language and “application” as the conceptual role. Automation does not necessarily mean AI.

synthetic content — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “content generated computationally rather than directly captured from reality” as the meaning anchor, “synthetic image; synthetic media” as natural language and “output” as the conceptual role. Synthetic does not automatically mean deceptive.

deepfake — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “synthetic/manipulated media made to resemble real people/events” as the meaning anchor, “deepfake video; detect deepfake” as natural language and “risk” as the conceptual role. Verification and disclosure matter.

confabulation — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “generated false or erroneous content presented confidently” as the meaning anchor, “AI confabulation; verify confabulated output” as natural language and “risk” as the conceptual role. Often colloquially called hallucination.

hallucination — Use the target in a media-literacy context involving synthetic content or online claims. Keep “common informal term for AI-generated false or unsupported content” as the meaning anchor, “AI hallucination; hallucinated citation” as natural language and “risk” as the conceptual role. Use with awareness that technical bodies may prefer confabulation.

bias — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “systematic tendency that can create skewed outcomes” as the meaning anchor, “algorithmic bias; data bias” as natural language and “risk” as the conceptual role. Bias can enter through data, design, measurement or deployment.

fairness — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “principles and methods for avoiding unjustified disparities” as the meaning anchor, “AI fairness; fairness assessment” as natural language and “ethics” as the conceptual role. Different fairness definitions can conflict.

transparency — Describe a simple AI use case and identify where the target appears in the system. Keep “openness about system purpose, limits, data or decision processes” as the meaning anchor, “AI transparency; transparent use” as natural language and “ethics” as the conceptual role. Transparency can support informed evaluation.

explainability — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “ability to provide understandable reasons or accounts of system behaviour” as the meaning anchor, “model explainability; explainable AI” as natural language and “ethics” as the conceptual role. What counts as an explanation depends on audience and task.

privacy — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “control/protection of personal information” as the meaning anchor, “data privacy; privacy risk” as natural language and “ethics” as the conceptual role. Students should consider what data they provide.

personal data — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “information relating to an identifiable person” as the meaning anchor, “personal-data protection; share personal data” as natural language and “data” as the conceptual role. Avoid unnecessary disclosure to tools.

security — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “protection against unauthorised access, misuse or attack” as the meaning anchor, “AI security; cybersecurity” as natural language and “risk” as the conceptual role. Security and privacy overlap but are not identical.

accountability — Use the target in a media-literacy context involving synthetic content or online claims. Keep “responsibility for decisions and consequences” as the meaning anchor, “human accountability; accountability mechanism” as natural language and “ethics” as the conceptual role. Responsibility should not disappear because AI was involved.

human oversight — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “human review, direction or intervention in system operation” as the meaning anchor, “human oversight; human review” as natural language and “governance” as the conceptual role. Important for consequential decisions.

human agency — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “ability of people to make meaningful choices and retain control” as the meaning anchor, “preserve human agency; human-centred design” as natural language and “ethics” as the conceptual role. UNESCO emphasises human-centred AI literacy.

plagiarism — Describe a simple AI use case and identify where the target appears in the system. Keep “using another’s work or ideas without appropriate acknowledgement” as the meaning anchor, “avoid plagiarism; plagiarism policy” as natural language and “school use” as the conceptual role. AI use can create attribution questions; follow school rules.

attribution — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “identifying source or contribution” as the meaning anchor, “proper attribution; attribute a source” as natural language and “school use” as the conceptual role. Attribution practices depend on task and institution.

citation — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “reference identifying a source” as the meaning anchor, “verify citation; cite a source” as natural language and “school use” as the conceptual role. Generated citations may be false; verify them.

provenance — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “record of origin/history of content or data” as the meaning anchor, “content provenance; data provenance” as natural language and “verification” as the conceptual role. Useful for synthetic media and sources.

verification — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “checking accuracy/authenticity” as the meaning anchor, “verify output; independent verification” as natural language and “practice” as the conceptual role. Never treat fluent output as self-verifying.

evaluation — Use the target in a media-literacy context involving synthetic content or online claims. Keep “systematic assessment of quality or performance” as the meaning anchor, “model evaluation; evaluate accuracy” as natural language and “practice” as the conceptual role. Metrics must match task and risk.

accuracy — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “degree to which output is correct under defined criteria” as the meaning anchor, “accuracy rate; factual accuracy” as natural language and “quality” as the conceptual role. Correctness depends on task definition.

reliability — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “consistency/dependability of performance” as the meaning anchor, “reliable system; reliability testing” as natural language and “quality” as the conceptual role. Reliable output can still be systematically wrong.

responsible use — Describe a simple AI use case and identify where the target appears in the system. Keep “use that considers accuracy, privacy, fairness, rules and human responsibility” as the meaning anchor, “responsible AI use; responsible practice” as natural language and “ethics” as the conceptual role. Context determines appropriate use.

AI Literacy Cycle 3

artificial intelligence — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “computer systems performing tasks associated with perception, learning, reasoning, communication or action” as the meaning anchor, “AI system; artificial intelligence application” as natural language and “core” as the conceptual role. AI is a broad category, not one single technology.

generative AI — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “AI models designed to generate synthetic content” as the meaning anchor, “generative AI model; GenAI output” as natural language and “core” as the conceptual role. Generation is different from retrieval or classification.

machine learning — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “methods in which systems learn patterns from data for tasks” as the meaning anchor, “machine-learning model; train a model” as natural language and “technique” as the conceptual role. Not every software rule is machine learning.

model — Use the target in a media-literacy context involving synthetic content or online claims. Keep “computational representation used to make predictions or generate outputs” as the meaning anchor, “AI model; trained model” as natural language and “system” as the conceptual role. A model is not the same as the full product or service.

algorithm — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “set of procedures or computational steps” as the meaning anchor, “ranking algorithm; algorithmic process” as natural language and “system” as the conceptual role. Algorithm is broader than AI.

training data — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “data used to fit or develop a model” as the meaning anchor, “training dataset; training-data quality” as natural language and “data” as the conceptual role. Data quality and representativeness matter.

dataset — Describe a simple AI use case and identify where the target appears in the system. Keep “organised collection of data” as the meaning anchor, “large dataset; labelled dataset” as natural language and “data” as the conceptual role. A dataset is not automatically unbiased or accurate.

input — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “information given to a system” as the meaning anchor, “user input; input data” as natural language and “interaction” as the conceptual role. Input may include text, image, audio or structured data.

prompt — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “instruction or context given to a generative system” as the meaning anchor, “write a prompt; prompt design” as natural language and “interaction” as the conceptual role. Prompt quality can affect output but cannot guarantee truth.

output — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “result produced by a system” as the meaning anchor, “model output; generated output” as natural language and “interaction” as the conceptual role. Output still needs evaluation.

token — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “small unit used in text processing by many language models” as the meaning anchor, “input tokens; token sequence” as natural language and “technical” as the conceptual role. Tokens are not always whole words.

parameter — Use the target in a media-literacy context involving synthetic content or online claims. Keep “learned or configured value influencing model behaviour” as the meaning anchor, “model parameter; parameter count” as natural language and “technical” as the conceptual role. Parameter count alone does not determine usefulness.

inference — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “running a trained model to produce a prediction/output” as the meaning anchor, “model inference; inference time” as natural language and “technical” as the conceptual role. Different from reading-comprehension inference despite same word.

classification — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “assigning inputs to categories” as the meaning anchor, “image classification; classify text” as natural language and “task” as the conceptual role. Different from generation.

prediction — Describe a simple AI use case and identify where the target appears in the system. Keep “estimating an outcome or value” as the meaning anchor, “predictive model; make a prediction” as natural language and “task” as the conceptual role. Prediction is probabilistic, not certainty.

recommendation system — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “system ranking or suggesting items to users” as the meaning anchor, “recommendation algorithm; recommended content” as natural language and “task” as the conceptual role. Visibility may reflect ranking criteria, not truth or importance.

automation — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “use of systems to perform tasks with reduced direct human action” as the meaning anchor, “automate a task; automation” as natural language and “application” as the conceptual role. Automation does not necessarily mean AI.

synthetic content — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “content generated computationally rather than directly captured from reality” as the meaning anchor, “synthetic image; synthetic media” as natural language and “output” as the conceptual role. Synthetic does not automatically mean deceptive.

deepfake — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “synthetic/manipulated media made to resemble real people/events” as the meaning anchor, “deepfake video; detect deepfake” as natural language and “risk” as the conceptual role. Verification and disclosure matter.

confabulation — Use the target in a media-literacy context involving synthetic content or online claims. Keep “generated false or erroneous content presented confidently” as the meaning anchor, “AI confabulation; verify confabulated output” as natural language and “risk” as the conceptual role. Often colloquially called hallucination.

hallucination — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “common informal term for AI-generated false or unsupported content” as the meaning anchor, “AI hallucination; hallucinated citation” as natural language and “risk” as the conceptual role. Use with awareness that technical bodies may prefer confabulation.

bias — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “systematic tendency that can create skewed outcomes” as the meaning anchor, “algorithmic bias; data bias” as natural language and “risk” as the conceptual role. Bias can enter through data, design, measurement or deployment.

fairness — Describe a simple AI use case and identify where the target appears in the system. Keep “principles and methods for avoiding unjustified disparities” as the meaning anchor, “AI fairness; fairness assessment” as natural language and “ethics” as the conceptual role. Different fairness definitions can conflict.

transparency — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “openness about system purpose, limits, data or decision processes” as the meaning anchor, “AI transparency; transparent use” as natural language and “ethics” as the conceptual role. Transparency can support informed evaluation.

explainability — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “ability to provide understandable reasons or accounts of system behaviour” as the meaning anchor, “model explainability; explainable AI” as natural language and “ethics” as the conceptual role. What counts as an explanation depends on audience and task.

privacy — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “control/protection of personal information” as the meaning anchor, “data privacy; privacy risk” as natural language and “ethics” as the conceptual role. Students should consider what data they provide.

personal data — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “information relating to an identifiable person” as the meaning anchor, “personal-data protection; share personal data” as natural language and “data” as the conceptual role. Avoid unnecessary disclosure to tools.

security — Use the target in a media-literacy context involving synthetic content or online claims. Keep “protection against unauthorised access, misuse or attack” as the meaning anchor, “AI security; cybersecurity” as natural language and “risk” as the conceptual role. Security and privacy overlap but are not identical.

accountability — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “responsibility for decisions and consequences” as the meaning anchor, “human accountability; accountability mechanism” as natural language and “ethics” as the conceptual role. Responsibility should not disappear because AI was involved.

human oversight — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “human review, direction or intervention in system operation” as the meaning anchor, “human oversight; human review” as natural language and “governance” as the conceptual role. Important for consequential decisions.

human agency — Describe a simple AI use case and identify where the target appears in the system. Keep “ability of people to make meaningful choices and retain control” as the meaning anchor, “preserve human agency; human-centred design” as natural language and “ethics” as the conceptual role. UNESCO emphasises human-centred AI literacy.

plagiarism — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “using another’s work or ideas without appropriate acknowledgement” as the meaning anchor, “avoid plagiarism; plagiarism policy” as natural language and “school use” as the conceptual role. AI use can create attribution questions; follow school rules.

attribution — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “identifying source or contribution” as the meaning anchor, “proper attribution; attribute a source” as natural language and “school use” as the conceptual role. Attribution practices depend on task and institution.

citation — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “reference identifying a source” as the meaning anchor, “verify citation; cite a source” as natural language and “school use” as the conceptual role. Generated citations may be false; verify them.

provenance — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “record of origin/history of content or data” as the meaning anchor, “content provenance; data provenance” as natural language and “verification” as the conceptual role. Useful for synthetic media and sources.

verification — Use the target in a media-literacy context involving synthetic content or online claims. Keep “checking accuracy/authenticity” as the meaning anchor, “verify output; independent verification” as natural language and “practice” as the conceptual role. Never treat fluent output as self-verifying.

evaluation — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “systematic assessment of quality or performance” as the meaning anchor, “model evaluation; evaluate accuracy” as natural language and “practice” as the conceptual role. Metrics must match task and risk.

accuracy — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “degree to which output is correct under defined criteria” as the meaning anchor, “accuracy rate; factual accuracy” as natural language and “quality” as the conceptual role. Correctness depends on task definition.

reliability — Describe a simple AI use case and identify where the target appears in the system. Keep “consistency/dependability of performance” as the meaning anchor, “reliable system; reliability testing” as natural language and “quality” as the conceptual role. Reliable output can still be systematically wrong.

responsible use — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “use that considers accuracy, privacy, fairness, rules and human responsibility” as the meaning anchor, “responsible AI use; responsible practice” as natural language and “ethics” as the conceptual role. Context determines appropriate use.

AI Literacy Cycle 4

artificial intelligence — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “computer systems performing tasks associated with perception, learning, reasoning, communication or action” as the meaning anchor, “AI system; artificial intelligence application” as natural language and “core” as the conceptual role. AI is a broad category, not one single technology.

generative AI — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “AI models designed to generate synthetic content” as the meaning anchor, “generative AI model; GenAI output” as natural language and “core” as the conceptual role. Generation is different from retrieval or classification.

machine learning — Use the target in a media-literacy context involving synthetic content or online claims. Keep “methods in which systems learn patterns from data for tasks” as the meaning anchor, “machine-learning model; train a model” as natural language and “technique” as the conceptual role. Not every software rule is machine learning.

model — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “computational representation used to make predictions or generate outputs” as the meaning anchor, “AI model; trained model” as natural language and “system” as the conceptual role. A model is not the same as the full product or service.

algorithm — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “set of procedures or computational steps” as the meaning anchor, “ranking algorithm; algorithmic process” as natural language and “system” as the conceptual role. Algorithm is broader than AI.

training data — Describe a simple AI use case and identify where the target appears in the system. Keep “data used to fit or develop a model” as the meaning anchor, “training dataset; training-data quality” as natural language and “data” as the conceptual role. Data quality and representativeness matter.

dataset — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “organised collection of data” as the meaning anchor, “large dataset; labelled dataset” as natural language and “data” as the conceptual role. A dataset is not automatically unbiased or accurate.

input — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “information given to a system” as the meaning anchor, “user input; input data” as natural language and “interaction” as the conceptual role. Input may include text, image, audio or structured data.

prompt — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “instruction or context given to a generative system” as the meaning anchor, “write a prompt; prompt design” as natural language and “interaction” as the conceptual role. Prompt quality can affect output but cannot guarantee truth.

output — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “result produced by a system” as the meaning anchor, “model output; generated output” as natural language and “interaction” as the conceptual role. Output still needs evaluation.

token — Use the target in a media-literacy context involving synthetic content or online claims. Keep “small unit used in text processing by many language models” as the meaning anchor, “input tokens; token sequence” as natural language and “technical” as the conceptual role. Tokens are not always whole words.

parameter — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “learned or configured value influencing model behaviour” as the meaning anchor, “model parameter; parameter count” as natural language and “technical” as the conceptual role. Parameter count alone does not determine usefulness.

inference — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “running a trained model to produce a prediction/output” as the meaning anchor, “model inference; inference time” as natural language and “technical” as the conceptual role. Different from reading-comprehension inference despite same word.

classification — Describe a simple AI use case and identify where the target appears in the system. Keep “assigning inputs to categories” as the meaning anchor, “image classification; classify text” as natural language and “task” as the conceptual role. Different from generation.

prediction — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “estimating an outcome or value” as the meaning anchor, “predictive model; make a prediction” as natural language and “task” as the conceptual role. Prediction is probabilistic, not certainty.

recommendation system — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “system ranking or suggesting items to users” as the meaning anchor, “recommendation algorithm; recommended content” as natural language and “task” as the conceptual role. Visibility may reflect ranking criteria, not truth or importance.

automation — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “use of systems to perform tasks with reduced direct human action” as the meaning anchor, “automate a task; automation” as natural language and “application” as the conceptual role. Automation does not necessarily mean AI.

synthetic content — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “content generated computationally rather than directly captured from reality” as the meaning anchor, “synthetic image; synthetic media” as natural language and “output” as the conceptual role. Synthetic does not automatically mean deceptive.

deepfake — Use the target in a media-literacy context involving synthetic content or online claims. Keep “synthetic/manipulated media made to resemble real people/events” as the meaning anchor, “deepfake video; detect deepfake” as natural language and “risk” as the conceptual role. Verification and disclosure matter.

confabulation — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “generated false or erroneous content presented confidently” as the meaning anchor, “AI confabulation; verify confabulated output” as natural language and “risk” as the conceptual role. Often colloquially called hallucination.

hallucination — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “common informal term for AI-generated false or unsupported content” as the meaning anchor, “AI hallucination; hallucinated citation” as natural language and “risk” as the conceptual role. Use with awareness that technical bodies may prefer confabulation.

bias — Describe a simple AI use case and identify where the target appears in the system. Keep “systematic tendency that can create skewed outcomes” as the meaning anchor, “algorithmic bias; data bias” as natural language and “risk” as the conceptual role. Bias can enter through data, design, measurement or deployment.

fairness — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “principles and methods for avoiding unjustified disparities” as the meaning anchor, “AI fairness; fairness assessment” as natural language and “ethics” as the conceptual role. Different fairness definitions can conflict.

transparency — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “openness about system purpose, limits, data or decision processes” as the meaning anchor, “AI transparency; transparent use” as natural language and “ethics” as the conceptual role. Transparency can support informed evaluation.

explainability — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “ability to provide understandable reasons or accounts of system behaviour” as the meaning anchor, “model explainability; explainable AI” as natural language and “ethics” as the conceptual role. What counts as an explanation depends on audience and task.

privacy — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “control/protection of personal information” as the meaning anchor, “data privacy; privacy risk” as natural language and “ethics” as the conceptual role. Students should consider what data they provide.

personal data — Use the target in a media-literacy context involving synthetic content or online claims. Keep “information relating to an identifiable person” as the meaning anchor, “personal-data protection; share personal data” as natural language and “data” as the conceptual role. Avoid unnecessary disclosure to tools.

security — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “protection against unauthorised access, misuse or attack” as the meaning anchor, “AI security; cybersecurity” as natural language and “risk” as the conceptual role. Security and privacy overlap but are not identical.

accountability — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “responsibility for decisions and consequences” as the meaning anchor, “human accountability; accountability mechanism” as natural language and “ethics” as the conceptual role. Responsibility should not disappear because AI was involved.

human oversight — Describe a simple AI use case and identify where the target appears in the system. Keep “human review, direction or intervention in system operation” as the meaning anchor, “human oversight; human review” as natural language and “governance” as the conceptual role. Important for consequential decisions.

human agency — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “ability of people to make meaningful choices and retain control” as the meaning anchor, “preserve human agency; human-centred design” as natural language and “ethics” as the conceptual role. UNESCO emphasises human-centred AI literacy.

plagiarism — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “using another’s work or ideas without appropriate acknowledgement” as the meaning anchor, “avoid plagiarism; plagiarism policy” as natural language and “school use” as the conceptual role. AI use can create attribution questions; follow school rules.

attribution — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “identifying source or contribution” as the meaning anchor, “proper attribution; attribute a source” as natural language and “school use” as the conceptual role. Attribution practices depend on task and institution.

citation — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “reference identifying a source” as the meaning anchor, “verify citation; cite a source” as natural language and “school use” as the conceptual role. Generated citations may be false; verify them.

provenance — Use the target in a media-literacy context involving synthetic content or online claims. Keep “record of origin/history of content or data” as the meaning anchor, “content provenance; data provenance” as natural language and “verification” as the conceptual role. Useful for synthetic media and sources.

verification — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “checking accuracy/authenticity” as the meaning anchor, “verify output; independent verification” as natural language and “practice” as the conceptual role. Never treat fluent output as self-verifying.

evaluation — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “systematic assessment of quality or performance” as the meaning anchor, “model evaluation; evaluate accuracy” as natural language and “practice” as the conceptual role. Metrics must match task and risk.

accuracy — Describe a simple AI use case and identify where the target appears in the system. Keep “degree to which output is correct under defined criteria” as the meaning anchor, “accuracy rate; factual accuracy” as natural language and “quality” as the conceptual role. Correctness depends on task definition.

reliability — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “consistency/dependability of performance” as the meaning anchor, “reliable system; reliability testing” as natural language and “quality” as the conceptual role. Reliable output can still be systematically wrong.

responsible use — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “use that considers accuracy, privacy, fairness, rules and human responsibility” as the meaning anchor, “responsible AI use; responsible practice” as natural language and “ethics” as the conceptual role. Context determines appropriate use.

AI Literacy Cycle 5

artificial intelligence — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “computer systems performing tasks associated with perception, learning, reasoning, communication or action” as the meaning anchor, “AI system; artificial intelligence application” as natural language and “core” as the conceptual role. AI is a broad category, not one single technology.

generative AI — Use the target in a media-literacy context involving synthetic content or online claims. Keep “AI models designed to generate synthetic content” as the meaning anchor, “generative AI model; GenAI output” as natural language and “core” as the conceptual role. Generation is different from retrieval or classification.

machine learning — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “methods in which systems learn patterns from data for tasks” as the meaning anchor, “machine-learning model; train a model” as natural language and “technique” as the conceptual role. Not every software rule is machine learning.

model — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “computational representation used to make predictions or generate outputs” as the meaning anchor, “AI model; trained model” as natural language and “system” as the conceptual role. A model is not the same as the full product or service.

algorithm — Describe a simple AI use case and identify where the target appears in the system. Keep “set of procedures or computational steps” as the meaning anchor, “ranking algorithm; algorithmic process” as natural language and “system” as the conceptual role. Algorithm is broader than AI.

training data — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “data used to fit or develop a model” as the meaning anchor, “training dataset; training-data quality” as natural language and “data” as the conceptual role. Data quality and representativeness matter.

dataset — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “organised collection of data” as the meaning anchor, “large dataset; labelled dataset” as natural language and “data” as the conceptual role. A dataset is not automatically unbiased or accurate.

input — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “information given to a system” as the meaning anchor, “user input; input data” as natural language and “interaction” as the conceptual role. Input may include text, image, audio or structured data.

prompt — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “instruction or context given to a generative system” as the meaning anchor, “write a prompt; prompt design” as natural language and “interaction” as the conceptual role. Prompt quality can affect output but cannot guarantee truth.

output — Use the target in a media-literacy context involving synthetic content or online claims. Keep “result produced by a system” as the meaning anchor, “model output; generated output” as natural language and “interaction” as the conceptual role. Output still needs evaluation.

token — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “small unit used in text processing by many language models” as the meaning anchor, “input tokens; token sequence” as natural language and “technical” as the conceptual role. Tokens are not always whole words.

parameter — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “learned or configured value influencing model behaviour” as the meaning anchor, “model parameter; parameter count” as natural language and “technical” as the conceptual role. Parameter count alone does not determine usefulness.

inference — Describe a simple AI use case and identify where the target appears in the system. Keep “running a trained model to produce a prediction/output” as the meaning anchor, “model inference; inference time” as natural language and “technical” as the conceptual role. Different from reading-comprehension inference despite same word.

classification — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “assigning inputs to categories” as the meaning anchor, “image classification; classify text” as natural language and “task” as the conceptual role. Different from generation.

prediction — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “estimating an outcome or value” as the meaning anchor, “predictive model; make a prediction” as natural language and “task” as the conceptual role. Prediction is probabilistic, not certainty.

recommendation system — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “system ranking or suggesting items to users” as the meaning anchor, “recommendation algorithm; recommended content” as natural language and “task” as the conceptual role. Visibility may reflect ranking criteria, not truth or importance.

automation — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “use of systems to perform tasks with reduced direct human action” as the meaning anchor, “automate a task; automation” as natural language and “application” as the conceptual role. Automation does not necessarily mean AI.

synthetic content — Use the target in a media-literacy context involving synthetic content or online claims. Keep “content generated computationally rather than directly captured from reality” as the meaning anchor, “synthetic image; synthetic media” as natural language and “output” as the conceptual role. Synthetic does not automatically mean deceptive.

deepfake — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “synthetic/manipulated media made to resemble real people/events” as the meaning anchor, “deepfake video; detect deepfake” as natural language and “risk” as the conceptual role. Verification and disclosure matter.

confabulation — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “generated false or erroneous content presented confidently” as the meaning anchor, “AI confabulation; verify confabulated output” as natural language and “risk” as the conceptual role. Often colloquially called hallucination.

hallucination — Describe a simple AI use case and identify where the target appears in the system. Keep “common informal term for AI-generated false or unsupported content” as the meaning anchor, “AI hallucination; hallucinated citation” as natural language and “risk” as the conceptual role. Use with awareness that technical bodies may prefer confabulation.

bias — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “systematic tendency that can create skewed outcomes” as the meaning anchor, “algorithmic bias; data bias” as natural language and “risk” as the conceptual role. Bias can enter through data, design, measurement or deployment.

fairness — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “principles and methods for avoiding unjustified disparities” as the meaning anchor, “AI fairness; fairness assessment” as natural language and “ethics” as the conceptual role. Different fairness definitions can conflict.

transparency — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “openness about system purpose, limits, data or decision processes” as the meaning anchor, “AI transparency; transparent use” as natural language and “ethics” as the conceptual role. Transparency can support informed evaluation.

explainability — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “ability to provide understandable reasons or accounts of system behaviour” as the meaning anchor, “model explainability; explainable AI” as natural language and “ethics” as the conceptual role. What counts as an explanation depends on audience and task.

privacy — Use the target in a media-literacy context involving synthetic content or online claims. Keep “control/protection of personal information” as the meaning anchor, “data privacy; privacy risk” as natural language and “ethics” as the conceptual role. Students should consider what data they provide.

personal data — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “information relating to an identifiable person” as the meaning anchor, “personal-data protection; share personal data” as natural language and “data” as the conceptual role. Avoid unnecessary disclosure to tools.

security — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “protection against unauthorised access, misuse or attack” as the meaning anchor, “AI security; cybersecurity” as natural language and “risk” as the conceptual role. Security and privacy overlap but are not identical.

accountability — Describe a simple AI use case and identify where the target appears in the system. Keep “responsibility for decisions and consequences” as the meaning anchor, “human accountability; accountability mechanism” as natural language and “ethics” as the conceptual role. Responsibility should not disappear because AI was involved.

human oversight — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “human review, direction or intervention in system operation” as the meaning anchor, “human oversight; human review” as natural language and “governance” as the conceptual role. Important for consequential decisions.

human agency — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “ability of people to make meaningful choices and retain control” as the meaning anchor, “preserve human agency; human-centred design” as natural language and “ethics” as the conceptual role. UNESCO emphasises human-centred AI literacy.

plagiarism — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “using another’s work or ideas without appropriate acknowledgement” as the meaning anchor, “avoid plagiarism; plagiarism policy” as natural language and “school use” as the conceptual role. AI use can create attribution questions; follow school rules.

attribution — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “identifying source or contribution” as the meaning anchor, “proper attribution; attribute a source” as natural language and “school use” as the conceptual role. Attribution practices depend on task and institution.

citation — Use the target in a media-literacy context involving synthetic content or online claims. Keep “reference identifying a source” as the meaning anchor, “verify citation; cite a source” as natural language and “school use” as the conceptual role. Generated citations may be false; verify them.

provenance — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “record of origin/history of content or data” as the meaning anchor, “content provenance; data provenance” as natural language and “verification” as the conceptual role. Useful for synthetic media and sources.

verification — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “checking accuracy/authenticity” as the meaning anchor, “verify output; independent verification” as natural language and “practice” as the conceptual role. Never treat fluent output as self-verifying.

evaluation — Describe a simple AI use case and identify where the target appears in the system. Keep “systematic assessment of quality or performance” as the meaning anchor, “model evaluation; evaluate accuracy” as natural language and “practice” as the conceptual role. Metrics must match task and risk.

accuracy — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “degree to which output is correct under defined criteria” as the meaning anchor, “accuracy rate; factual accuracy” as natural language and “quality” as the conceptual role. Correctness depends on task definition.

reliability — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “consistency/dependability of performance” as the meaning anchor, “reliable system; reliability testing” as natural language and “quality” as the conceptual role. Reliable output can still be systematically wrong.

responsible use — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “use that considers accuracy, privacy, fairness, rules and human responsibility” as the meaning anchor, “responsible AI use; responsible practice” as natural language and “ethics” as the conceptual role. Context determines appropriate use.

AI Literacy Cycle 6

artificial intelligence — Use the target in a media-literacy context involving synthetic content or online claims. Keep “computer systems performing tasks associated with perception, learning, reasoning, communication or action” as the meaning anchor, “AI system; artificial intelligence application” as natural language and “core” as the conceptual role. AI is a broad category, not one single technology.

generative AI — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “AI models designed to generate synthetic content” as the meaning anchor, “generative AI model; GenAI output” as natural language and “core” as the conceptual role. Generation is different from retrieval or classification.

machine learning — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “methods in which systems learn patterns from data for tasks” as the meaning anchor, “machine-learning model; train a model” as natural language and “technique” as the conceptual role. Not every software rule is machine learning.

model — Describe a simple AI use case and identify where the target appears in the system. Keep “computational representation used to make predictions or generate outputs” as the meaning anchor, “AI model; trained model” as natural language and “system” as the conceptual role. A model is not the same as the full product or service.

algorithm — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “set of procedures or computational steps” as the meaning anchor, “ranking algorithm; algorithmic process” as natural language and “system” as the conceptual role. Algorithm is broader than AI.

training data — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “data used to fit or develop a model” as the meaning anchor, “training dataset; training-data quality” as natural language and “data” as the conceptual role. Data quality and representativeness matter.

dataset — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “organised collection of data” as the meaning anchor, “large dataset; labelled dataset” as natural language and “data” as the conceptual role. A dataset is not automatically unbiased or accurate.

input — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “information given to a system” as the meaning anchor, “user input; input data” as natural language and “interaction” as the conceptual role. Input may include text, image, audio or structured data.

prompt — Use the target in a media-literacy context involving synthetic content or online claims. Keep “instruction or context given to a generative system” as the meaning anchor, “write a prompt; prompt design” as natural language and “interaction” as the conceptual role. Prompt quality can affect output but cannot guarantee truth.

output — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “result produced by a system” as the meaning anchor, “model output; generated output” as natural language and “interaction” as the conceptual role. Output still needs evaluation.

token — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “small unit used in text processing by many language models” as the meaning anchor, “input tokens; token sequence” as natural language and “technical” as the conceptual role. Tokens are not always whole words.

parameter — Describe a simple AI use case and identify where the target appears in the system. Keep “learned or configured value influencing model behaviour” as the meaning anchor, “model parameter; parameter count” as natural language and “technical” as the conceptual role. Parameter count alone does not determine usefulness.

inference — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “running a trained model to produce a prediction/output” as the meaning anchor, “model inference; inference time” as natural language and “technical” as the conceptual role. Different from reading-comprehension inference despite same word.

classification — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “assigning inputs to categories” as the meaning anchor, “image classification; classify text” as natural language and “task” as the conceptual role. Different from generation.

prediction — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “estimating an outcome or value” as the meaning anchor, “predictive model; make a prediction” as natural language and “task” as the conceptual role. Prediction is probabilistic, not certainty.

recommendation system — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “system ranking or suggesting items to users” as the meaning anchor, “recommendation algorithm; recommended content” as natural language and “task” as the conceptual role. Visibility may reflect ranking criteria, not truth or importance.

automation — Use the target in a media-literacy context involving synthetic content or online claims. Keep “use of systems to perform tasks with reduced direct human action” as the meaning anchor, “automate a task; automation” as natural language and “application” as the conceptual role. Automation does not necessarily mean AI.

synthetic content — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “content generated computationally rather than directly captured from reality” as the meaning anchor, “synthetic image; synthetic media” as natural language and “output” as the conceptual role. Synthetic does not automatically mean deceptive.

deepfake — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “synthetic/manipulated media made to resemble real people/events” as the meaning anchor, “deepfake video; detect deepfake” as natural language and “risk” as the conceptual role. Verification and disclosure matter.

confabulation — Describe a simple AI use case and identify where the target appears in the system. Keep “generated false or erroneous content presented confidently” as the meaning anchor, “AI confabulation; verify confabulated output” as natural language and “risk” as the conceptual role. Often colloquially called hallucination.

hallucination — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “common informal term for AI-generated false or unsupported content” as the meaning anchor, “AI hallucination; hallucinated citation” as natural language and “risk” as the conceptual role. Use with awareness that technical bodies may prefer confabulation.

bias — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “systematic tendency that can create skewed outcomes” as the meaning anchor, “algorithmic bias; data bias” as natural language and “risk” as the conceptual role. Bias can enter through data, design, measurement or deployment.

fairness — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “principles and methods for avoiding unjustified disparities” as the meaning anchor, “AI fairness; fairness assessment” as natural language and “ethics” as the conceptual role. Different fairness definitions can conflict.

transparency — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “openness about system purpose, limits, data or decision processes” as the meaning anchor, “AI transparency; transparent use” as natural language and “ethics” as the conceptual role. Transparency can support informed evaluation.

explainability — Use the target in a media-literacy context involving synthetic content or online claims. Keep “ability to provide understandable reasons or accounts of system behaviour” as the meaning anchor, “model explainability; explainable AI” as natural language and “ethics” as the conceptual role. What counts as an explanation depends on audience and task.

privacy — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “control/protection of personal information” as the meaning anchor, “data privacy; privacy risk” as natural language and “ethics” as the conceptual role. Students should consider what data they provide.

personal data — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “information relating to an identifiable person” as the meaning anchor, “personal-data protection; share personal data” as natural language and “data” as the conceptual role. Avoid unnecessary disclosure to tools.

security — Describe a simple AI use case and identify where the target appears in the system. Keep “protection against unauthorised access, misuse or attack” as the meaning anchor, “AI security; cybersecurity” as natural language and “risk” as the conceptual role. Security and privacy overlap but are not identical.

accountability — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “responsibility for decisions and consequences” as the meaning anchor, “human accountability; accountability mechanism” as natural language and “ethics” as the conceptual role. Responsibility should not disappear because AI was involved.

human oversight — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “human review, direction or intervention in system operation” as the meaning anchor, “human oversight; human review” as natural language and “governance” as the conceptual role. Important for consequential decisions.

human agency — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “ability of people to make meaningful choices and retain control” as the meaning anchor, “preserve human agency; human-centred design” as natural language and “ethics” as the conceptual role. UNESCO emphasises human-centred AI literacy.

plagiarism — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “using another’s work or ideas without appropriate acknowledgement” as the meaning anchor, “avoid plagiarism; plagiarism policy” as natural language and “school use” as the conceptual role. AI use can create attribution questions; follow school rules.

attribution — Use the target in a media-literacy context involving synthetic content or online claims. Keep “identifying source or contribution” as the meaning anchor, “proper attribution; attribute a source” as natural language and “school use” as the conceptual role. Attribution practices depend on task and institution.

citation — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “reference identifying a source” as the meaning anchor, “verify citation; cite a source” as natural language and “school use” as the conceptual role. Generated citations may be false; verify them.

provenance — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “record of origin/history of content or data” as the meaning anchor, “content provenance; data provenance” as natural language and “verification” as the conceptual role. Useful for synthetic media and sources.

verification — Describe a simple AI use case and identify where the target appears in the system. Keep “checking accuracy/authenticity” as the meaning anchor, “verify output; independent verification” as natural language and “practice” as the conceptual role. Never treat fluent output as self-verifying.

evaluation — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “systematic assessment of quality or performance” as the meaning anchor, “model evaluation; evaluate accuracy” as natural language and “practice” as the conceptual role. Metrics must match task and risk.

accuracy — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “degree to which output is correct under defined criteria” as the meaning anchor, “accuracy rate; factual accuracy” as natural language and “quality” as the conceptual role. Correctness depends on task definition.

reliability — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “consistency/dependability of performance” as the meaning anchor, “reliable system; reliability testing” as natural language and “quality” as the conceptual role. Reliable output can still be systematically wrong.

responsible use — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “use that considers accuracy, privacy, fairness, rules and human responsibility” as the meaning anchor, “responsible AI use; responsible practice” as natural language and “ethics” as the conceptual role. Context determines appropriate use.

AI Literacy Cycle 7

artificial intelligence — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “computer systems performing tasks associated with perception, learning, reasoning, communication or action” as the meaning anchor, “AI system; artificial intelligence application” as natural language and “core” as the conceptual role. AI is a broad category, not one single technology.

generative AI — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “AI models designed to generate synthetic content” as the meaning anchor, “generative AI model; GenAI output” as natural language and “core” as the conceptual role. Generation is different from retrieval or classification.

machine learning — Describe a simple AI use case and identify where the target appears in the system. Keep “methods in which systems learn patterns from data for tasks” as the meaning anchor, “machine-learning model; train a model” as natural language and “technique” as the conceptual role. Not every software rule is machine learning.

model — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “computational representation used to make predictions or generate outputs” as the meaning anchor, “AI model; trained model” as natural language and “system” as the conceptual role. A model is not the same as the full product or service.

algorithm — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “set of procedures or computational steps” as the meaning anchor, “ranking algorithm; algorithmic process” as natural language and “system” as the conceptual role. Algorithm is broader than AI.

training data — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “data used to fit or develop a model” as the meaning anchor, “training dataset; training-data quality” as natural language and “data” as the conceptual role. Data quality and representativeness matter.

dataset — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “organised collection of data” as the meaning anchor, “large dataset; labelled dataset” as natural language and “data” as the conceptual role. A dataset is not automatically unbiased or accurate.

input — Use the target in a media-literacy context involving synthetic content or online claims. Keep “information given to a system” as the meaning anchor, “user input; input data” as natural language and “interaction” as the conceptual role. Input may include text, image, audio or structured data.

prompt — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “instruction or context given to a generative system” as the meaning anchor, “write a prompt; prompt design” as natural language and “interaction” as the conceptual role. Prompt quality can affect output but cannot guarantee truth.

output — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “result produced by a system” as the meaning anchor, “model output; generated output” as natural language and “interaction” as the conceptual role. Output still needs evaluation.

token — Describe a simple AI use case and identify where the target appears in the system. Keep “small unit used in text processing by many language models” as the meaning anchor, “input tokens; token sequence” as natural language and “technical” as the conceptual role. Tokens are not always whole words.

parameter — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “learned or configured value influencing model behaviour” as the meaning anchor, “model parameter; parameter count” as natural language and “technical” as the conceptual role. Parameter count alone does not determine usefulness.

inference — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “running a trained model to produce a prediction/output” as the meaning anchor, “model inference; inference time” as natural language and “technical” as the conceptual role. Different from reading-comprehension inference despite same word.

classification — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “assigning inputs to categories” as the meaning anchor, “image classification; classify text” as natural language and “task” as the conceptual role. Different from generation.

prediction — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “estimating an outcome or value” as the meaning anchor, “predictive model; make a prediction” as natural language and “task” as the conceptual role. Prediction is probabilistic, not certainty.

recommendation system — Use the target in a media-literacy context involving synthetic content or online claims. Keep “system ranking or suggesting items to users” as the meaning anchor, “recommendation algorithm; recommended content” as natural language and “task” as the conceptual role. Visibility may reflect ranking criteria, not truth or importance.

automation — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “use of systems to perform tasks with reduced direct human action” as the meaning anchor, “automate a task; automation” as natural language and “application” as the conceptual role. Automation does not necessarily mean AI.

synthetic content — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “content generated computationally rather than directly captured from reality” as the meaning anchor, “synthetic image; synthetic media” as natural language and “output” as the conceptual role. Synthetic does not automatically mean deceptive.

deepfake — Describe a simple AI use case and identify where the target appears in the system. Keep “synthetic/manipulated media made to resemble real people/events” as the meaning anchor, “deepfake video; detect deepfake” as natural language and “risk” as the conceptual role. Verification and disclosure matter.

confabulation — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “generated false or erroneous content presented confidently” as the meaning anchor, “AI confabulation; verify confabulated output” as natural language and “risk” as the conceptual role. Often colloquially called hallucination.

hallucination — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “common informal term for AI-generated false or unsupported content” as the meaning anchor, “AI hallucination; hallucinated citation” as natural language and “risk” as the conceptual role. Use with awareness that technical bodies may prefer confabulation.

bias — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “systematic tendency that can create skewed outcomes” as the meaning anchor, “algorithmic bias; data bias” as natural language and “risk” as the conceptual role. Bias can enter through data, design, measurement or deployment.

fairness — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “principles and methods for avoiding unjustified disparities” as the meaning anchor, “AI fairness; fairness assessment” as natural language and “ethics” as the conceptual role. Different fairness definitions can conflict.

transparency — Use the target in a media-literacy context involving synthetic content or online claims. Keep “openness about system purpose, limits, data or decision processes” as the meaning anchor, “AI transparency; transparent use” as natural language and “ethics” as the conceptual role. Transparency can support informed evaluation.

explainability — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “ability to provide understandable reasons or accounts of system behaviour” as the meaning anchor, “model explainability; explainable AI” as natural language and “ethics” as the conceptual role. What counts as an explanation depends on audience and task.

privacy — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “control/protection of personal information” as the meaning anchor, “data privacy; privacy risk” as natural language and “ethics” as the conceptual role. Students should consider what data they provide.

personal data — Describe a simple AI use case and identify where the target appears in the system. Keep “information relating to an identifiable person” as the meaning anchor, “personal-data protection; share personal data” as natural language and “data” as the conceptual role. Avoid unnecessary disclosure to tools.

security — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “protection against unauthorised access, misuse or attack” as the meaning anchor, “AI security; cybersecurity” as natural language and “risk” as the conceptual role. Security and privacy overlap but are not identical.

accountability — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “responsibility for decisions and consequences” as the meaning anchor, “human accountability; accountability mechanism” as natural language and “ethics” as the conceptual role. Responsibility should not disappear because AI was involved.

human oversight — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “human review, direction or intervention in system operation” as the meaning anchor, “human oversight; human review” as natural language and “governance” as the conceptual role. Important for consequential decisions.

human agency — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “ability of people to make meaningful choices and retain control” as the meaning anchor, “preserve human agency; human-centred design” as natural language and “ethics” as the conceptual role. UNESCO emphasises human-centred AI literacy.

plagiarism — Use the target in a media-literacy context involving synthetic content or online claims. Keep “using another’s work or ideas without appropriate acknowledgement” as the meaning anchor, “avoid plagiarism; plagiarism policy” as natural language and “school use” as the conceptual role. AI use can create attribution questions; follow school rules.

attribution — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “identifying source or contribution” as the meaning anchor, “proper attribution; attribute a source” as natural language and “school use” as the conceptual role. Attribution practices depend on task and institution.

citation — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “reference identifying a source” as the meaning anchor, “verify citation; cite a source” as natural language and “school use” as the conceptual role. Generated citations may be false; verify them.

provenance — Describe a simple AI use case and identify where the target appears in the system. Keep “record of origin/history of content or data” as the meaning anchor, “content provenance; data provenance” as natural language and “verification” as the conceptual role. Useful for synthetic media and sources.

verification — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “checking accuracy/authenticity” as the meaning anchor, “verify output; independent verification” as natural language and “practice” as the conceptual role. Never treat fluent output as self-verifying.

evaluation — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “systematic assessment of quality or performance” as the meaning anchor, “model evaluation; evaluate accuracy” as natural language and “practice” as the conceptual role. Metrics must match task and risk.

accuracy — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “degree to which output is correct under defined criteria” as the meaning anchor, “accuracy rate; factual accuracy” as natural language and “quality” as the conceptual role. Correctness depends on task definition.

reliability — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “consistency/dependability of performance” as the meaning anchor, “reliable system; reliability testing” as natural language and “quality” as the conceptual role. Reliable output can still be systematically wrong.

responsible use — Use the target in a media-literacy context involving synthetic content or online claims. Keep “use that considers accuracy, privacy, fairness, rules and human responsibility” as the meaning anchor, “responsible AI use; responsible practice” as natural language and “ethics” as the conceptual role. Context determines appropriate use.

AI Literacy Cycle 8

artificial intelligence — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “computer systems performing tasks associated with perception, learning, reasoning, communication or action” as the meaning anchor, “AI system; artificial intelligence application” as natural language and “core” as the conceptual role. AI is a broad category, not one single technology.

generative AI — Describe a simple AI use case and identify where the target appears in the system. Keep “AI models designed to generate synthetic content” as the meaning anchor, “generative AI model; GenAI output” as natural language and “core” as the conceptual role. Generation is different from retrieval or classification.

machine learning — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “methods in which systems learn patterns from data for tasks” as the meaning anchor, “machine-learning model; train a model” as natural language and “technique” as the conceptual role. Not every software rule is machine learning.

model — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “computational representation used to make predictions or generate outputs” as the meaning anchor, “AI model; trained model” as natural language and “system” as the conceptual role. A model is not the same as the full product or service.

algorithm — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “set of procedures or computational steps” as the meaning anchor, “ranking algorithm; algorithmic process” as natural language and “system” as the conceptual role. Algorithm is broader than AI.

training data — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “data used to fit or develop a model” as the meaning anchor, “training dataset; training-data quality” as natural language and “data” as the conceptual role. Data quality and representativeness matter.

dataset — Use the target in a media-literacy context involving synthetic content or online claims. Keep “organised collection of data” as the meaning anchor, “large dataset; labelled dataset” as natural language and “data” as the conceptual role. A dataset is not automatically unbiased or accurate.

input — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “information given to a system” as the meaning anchor, “user input; input data” as natural language and “interaction” as the conceptual role. Input may include text, image, audio or structured data.

prompt — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “instruction or context given to a generative system” as the meaning anchor, “write a prompt; prompt design” as natural language and “interaction” as the conceptual role. Prompt quality can affect output but cannot guarantee truth.

output — Describe a simple AI use case and identify where the target appears in the system. Keep “result produced by a system” as the meaning anchor, “model output; generated output” as natural language and “interaction” as the conceptual role. Output still needs evaluation.

token — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “small unit used in text processing by many language models” as the meaning anchor, “input tokens; token sequence” as natural language and “technical” as the conceptual role. Tokens are not always whole words.

parameter — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “learned or configured value influencing model behaviour” as the meaning anchor, “model parameter; parameter count” as natural language and “technical” as the conceptual role. Parameter count alone does not determine usefulness.

inference — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “running a trained model to produce a prediction/output” as the meaning anchor, “model inference; inference time” as natural language and “technical” as the conceptual role. Different from reading-comprehension inference despite same word.

classification — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “assigning inputs to categories” as the meaning anchor, “image classification; classify text” as natural language and “task” as the conceptual role. Different from generation.

prediction — Use the target in a media-literacy context involving synthetic content or online claims. Keep “estimating an outcome or value” as the meaning anchor, “predictive model; make a prediction” as natural language and “task” as the conceptual role. Prediction is probabilistic, not certainty.

recommendation system — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “system ranking or suggesting items to users” as the meaning anchor, “recommendation algorithm; recommended content” as natural language and “task” as the conceptual role. Visibility may reflect ranking criteria, not truth or importance.

automation — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “use of systems to perform tasks with reduced direct human action” as the meaning anchor, “automate a task; automation” as natural language and “application” as the conceptual role. Automation does not necessarily mean AI.

synthetic content — Describe a simple AI use case and identify where the target appears in the system. Keep “content generated computationally rather than directly captured from reality” as the meaning anchor, “synthetic image; synthetic media” as natural language and “output” as the conceptual role. Synthetic does not automatically mean deceptive.

deepfake — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “synthetic/manipulated media made to resemble real people/events” as the meaning anchor, “deepfake video; detect deepfake” as natural language and “risk” as the conceptual role. Verification and disclosure matter.

confabulation — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “generated false or erroneous content presented confidently” as the meaning anchor, “AI confabulation; verify confabulated output” as natural language and “risk” as the conceptual role. Often colloquially called hallucination.

hallucination — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “common informal term for AI-generated false or unsupported content” as the meaning anchor, “AI hallucination; hallucinated citation” as natural language and “risk” as the conceptual role. Use with awareness that technical bodies may prefer confabulation.

bias — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “systematic tendency that can create skewed outcomes” as the meaning anchor, “algorithmic bias; data bias” as natural language and “risk” as the conceptual role. Bias can enter through data, design, measurement or deployment.

fairness — Use the target in a media-literacy context involving synthetic content or online claims. Keep “principles and methods for avoiding unjustified disparities” as the meaning anchor, “AI fairness; fairness assessment” as natural language and “ethics” as the conceptual role. Different fairness definitions can conflict.

transparency — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “openness about system purpose, limits, data or decision processes” as the meaning anchor, “AI transparency; transparent use” as natural language and “ethics” as the conceptual role. Transparency can support informed evaluation.

explainability — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “ability to provide understandable reasons or accounts of system behaviour” as the meaning anchor, “model explainability; explainable AI” as natural language and “ethics” as the conceptual role. What counts as an explanation depends on audience and task.

privacy — Describe a simple AI use case and identify where the target appears in the system. Keep “control/protection of personal information” as the meaning anchor, “data privacy; privacy risk” as natural language and “ethics” as the conceptual role. Students should consider what data they provide.

personal data — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “information relating to an identifiable person” as the meaning anchor, “personal-data protection; share personal data” as natural language and “data” as the conceptual role. Avoid unnecessary disclosure to tools.

security — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “protection against unauthorised access, misuse or attack” as the meaning anchor, “AI security; cybersecurity” as natural language and “risk” as the conceptual role. Security and privacy overlap but are not identical.

accountability — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “responsibility for decisions and consequences” as the meaning anchor, “human accountability; accountability mechanism” as natural language and “ethics” as the conceptual role. Responsibility should not disappear because AI was involved.

human oversight — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “human review, direction or intervention in system operation” as the meaning anchor, “human oversight; human review” as natural language and “governance” as the conceptual role. Important for consequential decisions.

human agency — Use the target in a media-literacy context involving synthetic content or online claims. Keep “ability of people to make meaningful choices and retain control” as the meaning anchor, “preserve human agency; human-centred design” as natural language and “ethics” as the conceptual role. UNESCO emphasises human-centred AI literacy.

plagiarism — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “using another’s work or ideas without appropriate acknowledgement” as the meaning anchor, “avoid plagiarism; plagiarism policy” as natural language and “school use” as the conceptual role. AI use can create attribution questions; follow school rules.

attribution — Move the term into a non-generative AI example so students do not equate all AI with chatbots. Keep “identifying source or contribution” as the meaning anchor, “proper attribution; attribute a source” as natural language and “school use” as the conceptual role. Attribution practices depend on task and institution.

citation — Describe a simple AI use case and identify where the target appears in the system. Keep “reference identifying a source” as the meaning anchor, “verify citation; cite a source” as natural language and “school use” as the conceptual role. Generated citations may be false; verify them.

provenance — Create a misconception about the target, then correct it in language a Grade 9 student could explain to a peer. Keep “record of origin/history of content or data” as the meaning anchor, “content provenance; data provenance” as natural language and “verification” as the conceptual role. Useful for synthetic media and sources.

verification — Write a verification task where a generated output contains one plausible falsehood. Explain how the target helps diagnose it. Keep “checking accuracy/authenticity” as the meaning anchor, “verify output; independent verification” as natural language and “practice” as the conceptual role. Never treat fluent output as self-verifying.

evaluation — Create a classroom-use scenario and decide what human responsibility remains even if AI assists. Keep “systematic assessment of quality or performance” as the meaning anchor, “model evaluation; evaluate accuracy” as natural language and “practice” as the conceptual role. Metrics must match task and risk.

accuracy — Compare the target with a neighbouring term. State the boundary that keeps them distinct. Keep “degree to which output is correct under defined criteria” as the meaning anchor, “accuracy rate; factual accuracy” as natural language and “quality” as the conceptual role. Correctness depends on task definition.

reliability — Use the target in a media-literacy context involving synthetic content or online claims. Keep “consistency/dependability of performance” as the meaning anchor, “reliable system; reliability testing” as natural language and “quality” as the conceptual role. Reliable output can still be systematically wrong.

responsible use — Write a short argument about AI use in school using the target accurately, one benefit and one limitation. Keep “use that considers accuracy, privacy, fairness, rules and human responsibility” as the meaning anchor, “responsible AI use; responsible practice” as natural language and “ethics” as the conceptual role. Context determines appropriate use.

A Human-Centred AI Vocabulary Framework

Human-centred mindset

Useful vocabulary: human agency, human oversight, accessibility, inclusion, responsibility. Keep people and social purposes central. Build one discussion question using at least two terms without reducing the topic to “AI good or bad.”

Ethics of AI

Useful vocabulary: fairness, privacy, transparency, accountability, bias, safety. Ask who may benefit, be excluded or bear risk. Build one discussion question using at least two terms without reducing the topic to “AI good or bad.”

AI techniques and applications

Useful vocabulary: model, algorithm, training data, machine learning, inference, classification, generation. Understand enough mechanism to avoid magical thinking. Build one discussion question using at least two terms without reducing the topic to “AI good or bad.”

AI system design

Useful vocabulary: input, output, evaluation, feedback, constraints, deployment, monitoring. Think about systems rather than isolated outputs. Build one discussion question using at least two terms without reducing the topic to “AI good or bad.”

AI Vocabulary for Schoolwork

Students need language for responsible academic use: brainstorm, draft, revise, verify, attribute, cite, paraphrase, evaluate, disclose. These ordinary academic verbs become especially important when generative tools are involved.

A sensible workflow is to define the task first, check school rules, decide which parts require the student’s own unaided work, use tools only where appropriate, verify generated claims and retain responsibility for the final submission.

The vocabulary of responsible use should support agency rather than encourage automatic delegation.

Prompt Literacy

A prompt is not magic wording. It is input that shapes the context available to a generative system. Clear prompts can specify task, audience, constraints, format and evidence requirements, but they cannot guarantee factuality or safety.

Students should distinguish prompt quality from output truth. A beautifully structured prompt can still produce a wrong answer. Verification remains separate.

Prompt literacy is therefore part of communication literacy: define purpose clearly, provide relevant context, evaluate the response and revise the interaction.

A 10-Week AI and Technology Vocabulary Programme

Week 1: AI versus ordinary software

Study AI versus ordinary software using plain-language examples and one current real-world case. Require students to define terms, distinguish neighbours, evaluate one claim and explain one human responsibility. Avoid treating vocabulary as product branding; focus on transferable concepts.

Week 2: models, algorithms and data

Study models, algorithms and data using plain-language examples and one current real-world case. Require students to define terms, distinguish neighbours, evaluate one claim and explain one human responsibility. Avoid treating vocabulary as product branding; focus on transferable concepts.

Week 3: generative AI and synthetic content

Study generative AI and synthetic content using plain-language examples and one current real-world case. Require students to define terms, distinguish neighbours, evaluate one claim and explain one human responsibility. Avoid treating vocabulary as product branding; focus on transferable concepts.

Week 4: prompts, inputs and outputs

Study prompts, inputs and outputs using plain-language examples and one current real-world case. Require students to define terms, distinguish neighbours, evaluate one claim and explain one human responsibility. Avoid treating vocabulary as product branding; focus on transferable concepts.

Week 5: confabulation and verification

Study confabulation and verification using plain-language examples and one current real-world case. Require students to define terms, distinguish neighbours, evaluate one claim and explain one human responsibility. Avoid treating vocabulary as product branding; focus on transferable concepts.

Week 6: bias, fairness and representation

Study bias, fairness and representation using plain-language examples and one current real-world case. Require students to define terms, distinguish neighbours, evaluate one claim and explain one human responsibility. Avoid treating vocabulary as product branding; focus on transferable concepts.

Week 7: privacy, security and personal data

Study privacy, security and personal data using plain-language examples and one current real-world case. Require students to define terms, distinguish neighbours, evaluate one claim and explain one human responsibility. Avoid treating vocabulary as product branding; focus on transferable concepts.

Week 8: transparency, explainability and accountability

Study transparency, explainability and accountability using plain-language examples and one current real-world case. Require students to define terms, distinguish neighbours, evaluate one claim and explain one human responsibility. Avoid treating vocabulary as product branding; focus on transferable concepts.

Week 9: AI in schoolwork and plagiarism/attribution

Study AI in schoolwork and plagiarism/attribution using plain-language examples and one current real-world case. Require students to define terms, distinguish neighbours, evaluate one claim and explain one human responsibility. Avoid treating vocabulary as product branding; focus on transferable concepts.

Week 10: human-centred responsible use

Study human-centred responsible use using plain-language examples and one current real-world case. Require students to define terms, distinguish neighbours, evaluate one claim and explain one human responsibility. Avoid treating vocabulary as product branding; focus on transferable concepts.

Frequently Asked Questions

What is generative AI?

AI models designed to generate synthetic content such as text, images, audio or video.

Is all AI generative AI?

No. AI also includes classification, prediction, recommendation, perception and other tasks.

What is a model?

A computational representation used to produce predictions, classifications or generated outputs.

What is training data?

Data used to fit or develop a model.

What is an AI hallucination?

A common term for generated false or unsupported content; NIST uses the term confabulation in its generative-AI risk profile.

What is algorithmic bias?

Systematic skew in outcomes that can arise from data, design, measurement, objectives or deployment.

Why is verification important?

Generative output can be fluent but wrong, incomplete or fabricated.

What does responsible AI use mean for students?

Following school rules, protecting privacy, verifying outputs, acknowledging sources appropriately and retaining responsibility for submitted work.

Is automation the same as AI?

No. Many automated systems use fixed rules and do not involve AI.

How does AI vocabulary help English?

It supports precise contemporary reading, media literacy, argument, explanation and discussion of technology.

Routes Into eduKateSG

• Secondary 3 English Vocabulary apex.
• Media Literacy, Misinformation and Source Credibility.
• Vocabulary Transfer Across Subjects.
• Independent Vocabulary Learning.
• Vocabulary Learning Hub.

References

UNESCO’s AI Competency Framework for Students organises student AI learning around a human-centred mindset, ethics of AI, AI techniques and applications, and AI system design. NIST’s current glossary defines generative AI as AI models that generate synthetic content, while its Generative AI Profile uses confabulation for confidently generated erroneous or false content. Common Sense Education’s Grades 9–12 AI lesson also foregrounds artificial intelligence, generative AI and plagiarism as student-facing vocabulary.

Final Principle

Secondary 3 AI vocabulary should reduce magical thinking. Students should be able to say what the system is, what data and input it uses, what output it produces, what can go wrong, how the output is verified and who remains responsible. The strongest AI literacy is not knowing the most buzzwords. It is knowing which distinction matters before making a judgement.

Extended AI Literacy Laboratory

AI Literacy Laboratory Cycle 1

artificial intelligence — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “computer systems performing tasks associated with perception, learning, reasoning, communication or action”, phrase “AI system; artificial intelligence application” and role “core”. AI is a broad category, not one single technology.

generative AI — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “AI models designed to generate synthetic content”, phrase “generative AI model; GenAI output” and role “core”. Generation is different from retrieval or classification.

machine learning — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “methods in which systems learn patterns from data for tasks”, phrase “machine-learning model; train a model” and role “technique”. Not every software rule is machine learning.

model — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “computational representation used to make predictions or generate outputs”, phrase “AI model; trained model” and role “system”. A model is not the same as the full product or service.

algorithm — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “set of procedures or computational steps”, phrase “ranking algorithm; algorithmic process” and role “system”. Algorithm is broader than AI.

training data — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “data used to fit or develop a model”, phrase “training dataset; training-data quality” and role “data”. Data quality and representativeness matter.

dataset — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “organised collection of data”, phrase “large dataset; labelled dataset” and role “data”. A dataset is not automatically unbiased or accurate.

input — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “information given to a system”, phrase “user input; input data” and role “interaction”. Input may include text, image, audio or structured data.

prompt — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “instruction or context given to a generative system”, phrase “write a prompt; prompt design” and role “interaction”. Prompt quality can affect output but cannot guarantee truth.

output — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “result produced by a system”, phrase “model output; generated output” and role “interaction”. Output still needs evaluation.

token — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “small unit used in text processing by many language models”, phrase “input tokens; token sequence” and role “technical”. Tokens are not always whole words.

parameter — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “learned or configured value influencing model behaviour”, phrase “model parameter; parameter count” and role “technical”. Parameter count alone does not determine usefulness.

inference — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “running a trained model to produce a prediction/output”, phrase “model inference; inference time” and role “technical”. Different from reading-comprehension inference despite same word.

classification — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “assigning inputs to categories”, phrase “image classification; classify text” and role “task”. Different from generation.

prediction — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “estimating an outcome or value”, phrase “predictive model; make a prediction” and role “task”. Prediction is probabilistic, not certainty.

recommendation system — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “system ranking or suggesting items to users”, phrase “recommendation algorithm; recommended content” and role “task”. Visibility may reflect ranking criteria, not truth or importance.

automation — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “use of systems to perform tasks with reduced direct human action”, phrase “automate a task; automation” and role “application”. Automation does not necessarily mean AI.

synthetic content — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “content generated computationally rather than directly captured from reality”, phrase “synthetic image; synthetic media” and role “output”. Synthetic does not automatically mean deceptive.

deepfake — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “synthetic/manipulated media made to resemble real people/events”, phrase “deepfake video; detect deepfake” and role “risk”. Verification and disclosure matter.

confabulation — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “generated false or erroneous content presented confidently”, phrase “AI confabulation; verify confabulated output” and role “risk”. Often colloquially called hallucination.

hallucination — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “common informal term for AI-generated false or unsupported content”, phrase “AI hallucination; hallucinated citation” and role “risk”. Use with awareness that technical bodies may prefer confabulation.

bias — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “systematic tendency that can create skewed outcomes”, phrase “algorithmic bias; data bias” and role “risk”. Bias can enter through data, design, measurement or deployment.

fairness — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “principles and methods for avoiding unjustified disparities”, phrase “AI fairness; fairness assessment” and role “ethics”. Different fairness definitions can conflict.

transparency — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “openness about system purpose, limits, data or decision processes”, phrase “AI transparency; transparent use” and role “ethics”. Transparency can support informed evaluation.

explainability — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “ability to provide understandable reasons or accounts of system behaviour”, phrase “model explainability; explainable AI” and role “ethics”. What counts as an explanation depends on audience and task.

privacy — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “control/protection of personal information”, phrase “data privacy; privacy risk” and role “ethics”. Students should consider what data they provide.

personal data — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “information relating to an identifiable person”, phrase “personal-data protection; share personal data” and role “data”. Avoid unnecessary disclosure to tools.

security — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “protection against unauthorised access, misuse or attack”, phrase “AI security; cybersecurity” and role “risk”. Security and privacy overlap but are not identical.

accountability — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “responsibility for decisions and consequences”, phrase “human accountability; accountability mechanism” and role “ethics”. Responsibility should not disappear because AI was involved.

human oversight — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “human review, direction or intervention in system operation”, phrase “human oversight; human review” and role “governance”. Important for consequential decisions.

human agency — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “ability of people to make meaningful choices and retain control”, phrase “preserve human agency; human-centred design” and role “ethics”. UNESCO emphasises human-centred AI literacy.

plagiarism — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “using another’s work or ideas without appropriate acknowledgement”, phrase “avoid plagiarism; plagiarism policy” and role “school use”. AI use can create attribution questions; follow school rules.

attribution — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “identifying source or contribution”, phrase “proper attribution; attribute a source” and role “school use”. Attribution practices depend on task and institution.

citation — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “reference identifying a source”, phrase “verify citation; cite a source” and role “school use”. Generated citations may be false; verify them.

provenance — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “record of origin/history of content or data”, phrase “content provenance; data provenance” and role “verification”. Useful for synthetic media and sources.

verification — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “checking accuracy/authenticity”, phrase “verify output; independent verification” and role “practice”. Never treat fluent output as self-verifying.

evaluation — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “systematic assessment of quality or performance”, phrase “model evaluation; evaluate accuracy” and role “practice”. Metrics must match task and risk.

accuracy — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “degree to which output is correct under defined criteria”, phrase “accuracy rate; factual accuracy” and role “quality”. Correctness depends on task definition.

reliability — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “consistency/dependability of performance”, phrase “reliable system; reliability testing” and role “quality”. Reliable output can still be systematically wrong.

responsible use — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “use that considers accuracy, privacy, fairness, rules and human responsibility”, phrase “responsible AI use; responsible practice” and role “ethics”. Context determines appropriate use.

AI Literacy Laboratory Cycle 2

artificial intelligence — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “computer systems performing tasks associated with perception, learning, reasoning, communication or action”, phrase “AI system; artificial intelligence application” and role “core”. AI is a broad category, not one single technology.

generative AI — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “AI models designed to generate synthetic content”, phrase “generative AI model; GenAI output” and role “core”. Generation is different from retrieval or classification.

machine learning — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “methods in which systems learn patterns from data for tasks”, phrase “machine-learning model; train a model” and role “technique”. Not every software rule is machine learning.

model — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “computational representation used to make predictions or generate outputs”, phrase “AI model; trained model” and role “system”. A model is not the same as the full product or service.

algorithm — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “set of procedures or computational steps”, phrase “ranking algorithm; algorithmic process” and role “system”. Algorithm is broader than AI.

training data — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “data used to fit or develop a model”, phrase “training dataset; training-data quality” and role “data”. Data quality and representativeness matter.

dataset — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “organised collection of data”, phrase “large dataset; labelled dataset” and role “data”. A dataset is not automatically unbiased or accurate.

input — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “information given to a system”, phrase “user input; input data” and role “interaction”. Input may include text, image, audio or structured data.

prompt — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “instruction or context given to a generative system”, phrase “write a prompt; prompt design” and role “interaction”. Prompt quality can affect output but cannot guarantee truth.

output — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “result produced by a system”, phrase “model output; generated output” and role “interaction”. Output still needs evaluation.

token — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “small unit used in text processing by many language models”, phrase “input tokens; token sequence” and role “technical”. Tokens are not always whole words.

parameter — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “learned or configured value influencing model behaviour”, phrase “model parameter; parameter count” and role “technical”. Parameter count alone does not determine usefulness.

inference — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “running a trained model to produce a prediction/output”, phrase “model inference; inference time” and role “technical”. Different from reading-comprehension inference despite same word.

classification — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “assigning inputs to categories”, phrase “image classification; classify text” and role “task”. Different from generation.

prediction — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “estimating an outcome or value”, phrase “predictive model; make a prediction” and role “task”. Prediction is probabilistic, not certainty.

recommendation system — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “system ranking or suggesting items to users”, phrase “recommendation algorithm; recommended content” and role “task”. Visibility may reflect ranking criteria, not truth or importance.

automation — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “use of systems to perform tasks with reduced direct human action”, phrase “automate a task; automation” and role “application”. Automation does not necessarily mean AI.

synthetic content — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “content generated computationally rather than directly captured from reality”, phrase “synthetic image; synthetic media” and role “output”. Synthetic does not automatically mean deceptive.

deepfake — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “synthetic/manipulated media made to resemble real people/events”, phrase “deepfake video; detect deepfake” and role “risk”. Verification and disclosure matter.

confabulation — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “generated false or erroneous content presented confidently”, phrase “AI confabulation; verify confabulated output” and role “risk”. Often colloquially called hallucination.

hallucination — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “common informal term for AI-generated false or unsupported content”, phrase “AI hallucination; hallucinated citation” and role “risk”. Use with awareness that technical bodies may prefer confabulation.

bias — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “systematic tendency that can create skewed outcomes”, phrase “algorithmic bias; data bias” and role “risk”. Bias can enter through data, design, measurement or deployment.

fairness — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “principles and methods for avoiding unjustified disparities”, phrase “AI fairness; fairness assessment” and role “ethics”. Different fairness definitions can conflict.

transparency — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “openness about system purpose, limits, data or decision processes”, phrase “AI transparency; transparent use” and role “ethics”. Transparency can support informed evaluation.

explainability — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “ability to provide understandable reasons or accounts of system behaviour”, phrase “model explainability; explainable AI” and role “ethics”. What counts as an explanation depends on audience and task.

privacy — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “control/protection of personal information”, phrase “data privacy; privacy risk” and role “ethics”. Students should consider what data they provide.

personal data — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “information relating to an identifiable person”, phrase “personal-data protection; share personal data” and role “data”. Avoid unnecessary disclosure to tools.

security — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “protection against unauthorised access, misuse or attack”, phrase “AI security; cybersecurity” and role “risk”. Security and privacy overlap but are not identical.

accountability — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “responsibility for decisions and consequences”, phrase “human accountability; accountability mechanism” and role “ethics”. Responsibility should not disappear because AI was involved.

human oversight — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “human review, direction or intervention in system operation”, phrase “human oversight; human review” and role “governance”. Important for consequential decisions.

human agency — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “ability of people to make meaningful choices and retain control”, phrase “preserve human agency; human-centred design” and role “ethics”. UNESCO emphasises human-centred AI literacy.

plagiarism — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “using another’s work or ideas without appropriate acknowledgement”, phrase “avoid plagiarism; plagiarism policy” and role “school use”. AI use can create attribution questions; follow school rules.

attribution — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “identifying source or contribution”, phrase “proper attribution; attribute a source” and role “school use”. Attribution practices depend on task and institution.

citation — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “reference identifying a source”, phrase “verify citation; cite a source” and role “school use”. Generated citations may be false; verify them.

provenance — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “record of origin/history of content or data”, phrase “content provenance; data provenance” and role “verification”. Useful for synthetic media and sources.

verification — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “checking accuracy/authenticity”, phrase “verify output; independent verification” and role “practice”. Never treat fluent output as self-verifying.

evaluation — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “systematic assessment of quality or performance”, phrase “model evaluation; evaluate accuracy” and role “practice”. Metrics must match task and risk.

accuracy — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “degree to which output is correct under defined criteria”, phrase “accuracy rate; factual accuracy” and role “quality”. Correctness depends on task definition.

reliability — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “consistency/dependability of performance”, phrase “reliable system; reliability testing” and role “quality”. Reliable output can still be systematically wrong.

responsible use — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “use that considers accuracy, privacy, fairness, rules and human responsibility”, phrase “responsible AI use; responsible practice” and role “ethics”. Context determines appropriate use.

AI Literacy Laboratory Cycle 3

artificial intelligence — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “computer systems performing tasks associated with perception, learning, reasoning, communication or action”, phrase “AI system; artificial intelligence application” and role “core”. AI is a broad category, not one single technology.

generative AI — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “AI models designed to generate synthetic content”, phrase “generative AI model; GenAI output” and role “core”. Generation is different from retrieval or classification.

machine learning — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “methods in which systems learn patterns from data for tasks”, phrase “machine-learning model; train a model” and role “technique”. Not every software rule is machine learning.

model — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “computational representation used to make predictions or generate outputs”, phrase “AI model; trained model” and role “system”. A model is not the same as the full product or service.

algorithm — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “set of procedures or computational steps”, phrase “ranking algorithm; algorithmic process” and role “system”. Algorithm is broader than AI.

training data — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “data used to fit or develop a model”, phrase “training dataset; training-data quality” and role “data”. Data quality and representativeness matter.

dataset — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “organised collection of data”, phrase “large dataset; labelled dataset” and role “data”. A dataset is not automatically unbiased or accurate.

input — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “information given to a system”, phrase “user input; input data” and role “interaction”. Input may include text, image, audio or structured data.

prompt — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “instruction or context given to a generative system”, phrase “write a prompt; prompt design” and role “interaction”. Prompt quality can affect output but cannot guarantee truth.

output — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “result produced by a system”, phrase “model output; generated output” and role “interaction”. Output still needs evaluation.

token — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “small unit used in text processing by many language models”, phrase “input tokens; token sequence” and role “technical”. Tokens are not always whole words.

parameter — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “learned or configured value influencing model behaviour”, phrase “model parameter; parameter count” and role “technical”. Parameter count alone does not determine usefulness.

inference — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “running a trained model to produce a prediction/output”, phrase “model inference; inference time” and role “technical”. Different from reading-comprehension inference despite same word.

classification — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “assigning inputs to categories”, phrase “image classification; classify text” and role “task”. Different from generation.

prediction — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “estimating an outcome or value”, phrase “predictive model; make a prediction” and role “task”. Prediction is probabilistic, not certainty.

recommendation system — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “system ranking or suggesting items to users”, phrase “recommendation algorithm; recommended content” and role “task”. Visibility may reflect ranking criteria, not truth or importance.

automation — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “use of systems to perform tasks with reduced direct human action”, phrase “automate a task; automation” and role “application”. Automation does not necessarily mean AI.

synthetic content — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “content generated computationally rather than directly captured from reality”, phrase “synthetic image; synthetic media” and role “output”. Synthetic does not automatically mean deceptive.

deepfake — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “synthetic/manipulated media made to resemble real people/events”, phrase “deepfake video; detect deepfake” and role “risk”. Verification and disclosure matter.

confabulation — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “generated false or erroneous content presented confidently”, phrase “AI confabulation; verify confabulated output” and role “risk”. Often colloquially called hallucination.

hallucination — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “common informal term for AI-generated false or unsupported content”, phrase “AI hallucination; hallucinated citation” and role “risk”. Use with awareness that technical bodies may prefer confabulation.

bias — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “systematic tendency that can create skewed outcomes”, phrase “algorithmic bias; data bias” and role “risk”. Bias can enter through data, design, measurement or deployment.

fairness — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “principles and methods for avoiding unjustified disparities”, phrase “AI fairness; fairness assessment” and role “ethics”. Different fairness definitions can conflict.

transparency — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “openness about system purpose, limits, data or decision processes”, phrase “AI transparency; transparent use” and role “ethics”. Transparency can support informed evaluation.

explainability — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “ability to provide understandable reasons or accounts of system behaviour”, phrase “model explainability; explainable AI” and role “ethics”. What counts as an explanation depends on audience and task.

privacy — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “control/protection of personal information”, phrase “data privacy; privacy risk” and role “ethics”. Students should consider what data they provide.

personal data — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “information relating to an identifiable person”, phrase “personal-data protection; share personal data” and role “data”. Avoid unnecessary disclosure to tools.

security — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “protection against unauthorised access, misuse or attack”, phrase “AI security; cybersecurity” and role “risk”. Security and privacy overlap but are not identical.

accountability — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “responsibility for decisions and consequences”, phrase “human accountability; accountability mechanism” and role “ethics”. Responsibility should not disappear because AI was involved.

human oversight — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “human review, direction or intervention in system operation”, phrase “human oversight; human review” and role “governance”. Important for consequential decisions.

human agency — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “ability of people to make meaningful choices and retain control”, phrase “preserve human agency; human-centred design” and role “ethics”. UNESCO emphasises human-centred AI literacy.

plagiarism — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “using another’s work or ideas without appropriate acknowledgement”, phrase “avoid plagiarism; plagiarism policy” and role “school use”. AI use can create attribution questions; follow school rules.

attribution — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “identifying source or contribution”, phrase “proper attribution; attribute a source” and role “school use”. Attribution practices depend on task and institution.

citation — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “reference identifying a source”, phrase “verify citation; cite a source” and role “school use”. Generated citations may be false; verify them.

provenance — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “record of origin/history of content or data”, phrase “content provenance; data provenance” and role “verification”. Useful for synthetic media and sources.

verification — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “checking accuracy/authenticity”, phrase “verify output; independent verification” and role “practice”. Never treat fluent output as self-verifying.

evaluation — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “systematic assessment of quality or performance”, phrase “model evaluation; evaluate accuracy” and role “practice”. Metrics must match task and risk.

accuracy — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “degree to which output is correct under defined criteria”, phrase “accuracy rate; factual accuracy” and role “quality”. Correctness depends on task definition.

reliability — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “consistency/dependability of performance”, phrase “reliable system; reliability testing” and role “quality”. Reliable output can still be systematically wrong.

responsible use — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “use that considers accuracy, privacy, fairness, rules and human responsibility”, phrase “responsible AI use; responsible practice” and role “ethics”. Context determines appropriate use.

AI Literacy Laboratory Cycle 4

artificial intelligence — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “computer systems performing tasks associated with perception, learning, reasoning, communication or action”, phrase “AI system; artificial intelligence application” and role “core”. AI is a broad category, not one single technology.

generative AI — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “AI models designed to generate synthetic content”, phrase “generative AI model; GenAI output” and role “core”. Generation is different from retrieval or classification.

machine learning — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “methods in which systems learn patterns from data for tasks”, phrase “machine-learning model; train a model” and role “technique”. Not every software rule is machine learning.

model — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “computational representation used to make predictions or generate outputs”, phrase “AI model; trained model” and role “system”. A model is not the same as the full product or service.

algorithm — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “set of procedures or computational steps”, phrase “ranking algorithm; algorithmic process” and role “system”. Algorithm is broader than AI.

training data — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “data used to fit or develop a model”, phrase “training dataset; training-data quality” and role “data”. Data quality and representativeness matter.

dataset — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “organised collection of data”, phrase “large dataset; labelled dataset” and role “data”. A dataset is not automatically unbiased or accurate.

input — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “information given to a system”, phrase “user input; input data” and role “interaction”. Input may include text, image, audio or structured data.

prompt — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “instruction or context given to a generative system”, phrase “write a prompt; prompt design” and role “interaction”. Prompt quality can affect output but cannot guarantee truth.

output — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “result produced by a system”, phrase “model output; generated output” and role “interaction”. Output still needs evaluation.

token — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “small unit used in text processing by many language models”, phrase “input tokens; token sequence” and role “technical”. Tokens are not always whole words.

parameter — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “learned or configured value influencing model behaviour”, phrase “model parameter; parameter count” and role “technical”. Parameter count alone does not determine usefulness.

inference — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “running a trained model to produce a prediction/output”, phrase “model inference; inference time” and role “technical”. Different from reading-comprehension inference despite same word.

classification — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “assigning inputs to categories”, phrase “image classification; classify text” and role “task”. Different from generation.

prediction — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “estimating an outcome or value”, phrase “predictive model; make a prediction” and role “task”. Prediction is probabilistic, not certainty.

recommendation system — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “system ranking or suggesting items to users”, phrase “recommendation algorithm; recommended content” and role “task”. Visibility may reflect ranking criteria, not truth or importance.

automation — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “use of systems to perform tasks with reduced direct human action”, phrase “automate a task; automation” and role “application”. Automation does not necessarily mean AI.

synthetic content — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “content generated computationally rather than directly captured from reality”, phrase “synthetic image; synthetic media” and role “output”. Synthetic does not automatically mean deceptive.

deepfake — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “synthetic/manipulated media made to resemble real people/events”, phrase “deepfake video; detect deepfake” and role “risk”. Verification and disclosure matter.

confabulation — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “generated false or erroneous content presented confidently”, phrase “AI confabulation; verify confabulated output” and role “risk”. Often colloquially called hallucination.

hallucination — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “common informal term for AI-generated false or unsupported content”, phrase “AI hallucination; hallucinated citation” and role “risk”. Use with awareness that technical bodies may prefer confabulation.

bias — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “systematic tendency that can create skewed outcomes”, phrase “algorithmic bias; data bias” and role “risk”. Bias can enter through data, design, measurement or deployment.

fairness — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “principles and methods for avoiding unjustified disparities”, phrase “AI fairness; fairness assessment” and role “ethics”. Different fairness definitions can conflict.

transparency — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “openness about system purpose, limits, data or decision processes”, phrase “AI transparency; transparent use” and role “ethics”. Transparency can support informed evaluation.

explainability — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “ability to provide understandable reasons or accounts of system behaviour”, phrase “model explainability; explainable AI” and role “ethics”. What counts as an explanation depends on audience and task.

privacy — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “control/protection of personal information”, phrase “data privacy; privacy risk” and role “ethics”. Students should consider what data they provide.

personal data — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “information relating to an identifiable person”, phrase “personal-data protection; share personal data” and role “data”. Avoid unnecessary disclosure to tools.

security — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “protection against unauthorised access, misuse or attack”, phrase “AI security; cybersecurity” and role “risk”. Security and privacy overlap but are not identical.

accountability — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “responsibility for decisions and consequences”, phrase “human accountability; accountability mechanism” and role “ethics”. Responsibility should not disappear because AI was involved.

human oversight — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “human review, direction or intervention in system operation”, phrase “human oversight; human review” and role “governance”. Important for consequential decisions.

human agency — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “ability of people to make meaningful choices and retain control”, phrase “preserve human agency; human-centred design” and role “ethics”. UNESCO emphasises human-centred AI literacy.

plagiarism — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “using another’s work or ideas without appropriate acknowledgement”, phrase “avoid plagiarism; plagiarism policy” and role “school use”. AI use can create attribution questions; follow school rules.

attribution — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “identifying source or contribution”, phrase “proper attribution; attribute a source” and role “school use”. Attribution practices depend on task and institution.

citation — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “reference identifying a source”, phrase “verify citation; cite a source” and role “school use”. Generated citations may be false; verify them.

provenance — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “record of origin/history of content or data”, phrase “content provenance; data provenance” and role “verification”. Useful for synthetic media and sources.

verification — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “checking accuracy/authenticity”, phrase “verify output; independent verification” and role “practice”. Never treat fluent output as self-verifying.

evaluation — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “systematic assessment of quality or performance”, phrase “model evaluation; evaluate accuracy” and role “practice”. Metrics must match task and risk.

accuracy — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “degree to which output is correct under defined criteria”, phrase “accuracy rate; factual accuracy” and role “quality”. Correctness depends on task definition.

reliability — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “consistency/dependability of performance”, phrase “reliable system; reliability testing” and role “quality”. Reliable output can still be systematically wrong.

responsible use — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “use that considers accuracy, privacy, fairness, rules and human responsibility”, phrase “responsible AI use; responsible practice” and role “ethics”. Context determines appropriate use.

AI Literacy Laboratory Cycle 5

artificial intelligence — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “computer systems performing tasks associated with perception, learning, reasoning, communication or action”, phrase “AI system; artificial intelligence application” and role “core”. AI is a broad category, not one single technology.

generative AI — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “AI models designed to generate synthetic content”, phrase “generative AI model; GenAI output” and role “core”. Generation is different from retrieval or classification.

machine learning — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “methods in which systems learn patterns from data for tasks”, phrase “machine-learning model; train a model” and role “technique”. Not every software rule is machine learning.

model — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “computational representation used to make predictions or generate outputs”, phrase “AI model; trained model” and role “system”. A model is not the same as the full product or service.

algorithm — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “set of procedures or computational steps”, phrase “ranking algorithm; algorithmic process” and role “system”. Algorithm is broader than AI.

training data — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “data used to fit or develop a model”, phrase “training dataset; training-data quality” and role “data”. Data quality and representativeness matter.

dataset — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “organised collection of data”, phrase “large dataset; labelled dataset” and role “data”. A dataset is not automatically unbiased or accurate.

input — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “information given to a system”, phrase “user input; input data” and role “interaction”. Input may include text, image, audio or structured data.

prompt — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “instruction or context given to a generative system”, phrase “write a prompt; prompt design” and role “interaction”. Prompt quality can affect output but cannot guarantee truth.

output — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “result produced by a system”, phrase “model output; generated output” and role “interaction”. Output still needs evaluation.

token — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “small unit used in text processing by many language models”, phrase “input tokens; token sequence” and role “technical”. Tokens are not always whole words.

parameter — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “learned or configured value influencing model behaviour”, phrase “model parameter; parameter count” and role “technical”. Parameter count alone does not determine usefulness.

inference — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “running a trained model to produce a prediction/output”, phrase “model inference; inference time” and role “technical”. Different from reading-comprehension inference despite same word.

classification — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “assigning inputs to categories”, phrase “image classification; classify text” and role “task”. Different from generation.

prediction — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “estimating an outcome or value”, phrase “predictive model; make a prediction” and role “task”. Prediction is probabilistic, not certainty.

recommendation system — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “system ranking or suggesting items to users”, phrase “recommendation algorithm; recommended content” and role “task”. Visibility may reflect ranking criteria, not truth or importance.

automation — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “use of systems to perform tasks with reduced direct human action”, phrase “automate a task; automation” and role “application”. Automation does not necessarily mean AI.

synthetic content — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “content generated computationally rather than directly captured from reality”, phrase “synthetic image; synthetic media” and role “output”. Synthetic does not automatically mean deceptive.

deepfake — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “synthetic/manipulated media made to resemble real people/events”, phrase “deepfake video; detect deepfake” and role “risk”. Verification and disclosure matter.

confabulation — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “generated false or erroneous content presented confidently”, phrase “AI confabulation; verify confabulated output” and role “risk”. Often colloquially called hallucination.

hallucination — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “common informal term for AI-generated false or unsupported content”, phrase “AI hallucination; hallucinated citation” and role “risk”. Use with awareness that technical bodies may prefer confabulation.

bias — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “systematic tendency that can create skewed outcomes”, phrase “algorithmic bias; data bias” and role “risk”. Bias can enter through data, design, measurement or deployment.

fairness — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “principles and methods for avoiding unjustified disparities”, phrase “AI fairness; fairness assessment” and role “ethics”. Different fairness definitions can conflict.

transparency — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “openness about system purpose, limits, data or decision processes”, phrase “AI transparency; transparent use” and role “ethics”. Transparency can support informed evaluation.

explainability — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “ability to provide understandable reasons or accounts of system behaviour”, phrase “model explainability; explainable AI” and role “ethics”. What counts as an explanation depends on audience and task.

privacy — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “control/protection of personal information”, phrase “data privacy; privacy risk” and role “ethics”. Students should consider what data they provide.

personal data — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “information relating to an identifiable person”, phrase “personal-data protection; share personal data” and role “data”. Avoid unnecessary disclosure to tools.

security — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “protection against unauthorised access, misuse or attack”, phrase “AI security; cybersecurity” and role “risk”. Security and privacy overlap but are not identical.

accountability — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “responsibility for decisions and consequences”, phrase “human accountability; accountability mechanism” and role “ethics”. Responsibility should not disappear because AI was involved.

human oversight — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “human review, direction or intervention in system operation”, phrase “human oversight; human review” and role “governance”. Important for consequential decisions.

human agency — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “ability of people to make meaningful choices and retain control”, phrase “preserve human agency; human-centred design” and role “ethics”. UNESCO emphasises human-centred AI literacy.

plagiarism — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “using another’s work or ideas without appropriate acknowledgement”, phrase “avoid plagiarism; plagiarism policy” and role “school use”. AI use can create attribution questions; follow school rules.

attribution — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “identifying source or contribution”, phrase “proper attribution; attribute a source” and role “school use”. Attribution practices depend on task and institution.

citation — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “reference identifying a source”, phrase “verify citation; cite a source” and role “school use”. Generated citations may be false; verify them.

provenance — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “record of origin/history of content or data”, phrase “content provenance; data provenance” and role “verification”. Useful for synthetic media and sources.

verification — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “checking accuracy/authenticity”, phrase “verify output; independent verification” and role “practice”. Never treat fluent output as self-verifying.

evaluation — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “systematic assessment of quality or performance”, phrase “model evaluation; evaluate accuracy” and role “practice”. Metrics must match task and risk.

accuracy — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “degree to which output is correct under defined criteria”, phrase “accuracy rate; factual accuracy” and role “quality”. Correctness depends on task definition.

reliability — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “consistency/dependability of performance”, phrase “reliable system; reliability testing” and role “quality”. Reliable output can still be systematically wrong.

responsible use — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “use that considers accuracy, privacy, fairness, rules and human responsibility”, phrase “responsible AI use; responsible practice” and role “ethics”. Context determines appropriate use.

AI Literacy Laboratory Cycle 6

artificial intelligence — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “computer systems performing tasks associated with perception, learning, reasoning, communication or action”, phrase “AI system; artificial intelligence application” and role “core”. AI is a broad category, not one single technology.

generative AI — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “AI models designed to generate synthetic content”, phrase “generative AI model; GenAI output” and role “core”. Generation is different from retrieval or classification.

machine learning — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “methods in which systems learn patterns from data for tasks”, phrase “machine-learning model; train a model” and role “technique”. Not every software rule is machine learning.

model — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “computational representation used to make predictions or generate outputs”, phrase “AI model; trained model” and role “system”. A model is not the same as the full product or service.

algorithm — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “set of procedures or computational steps”, phrase “ranking algorithm; algorithmic process” and role “system”. Algorithm is broader than AI.

training data — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “data used to fit or develop a model”, phrase “training dataset; training-data quality” and role “data”. Data quality and representativeness matter.

dataset — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “organised collection of data”, phrase “large dataset; labelled dataset” and role “data”. A dataset is not automatically unbiased or accurate.

input — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “information given to a system”, phrase “user input; input data” and role “interaction”. Input may include text, image, audio or structured data.

prompt — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “instruction or context given to a generative system”, phrase “write a prompt; prompt design” and role “interaction”. Prompt quality can affect output but cannot guarantee truth.

output — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “result produced by a system”, phrase “model output; generated output” and role “interaction”. Output still needs evaluation.

token — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “small unit used in text processing by many language models”, phrase “input tokens; token sequence” and role “technical”. Tokens are not always whole words.

parameter — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “learned or configured value influencing model behaviour”, phrase “model parameter; parameter count” and role “technical”. Parameter count alone does not determine usefulness.

inference — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “running a trained model to produce a prediction/output”, phrase “model inference; inference time” and role “technical”. Different from reading-comprehension inference despite same word.

classification — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “assigning inputs to categories”, phrase “image classification; classify text” and role “task”. Different from generation.

prediction — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “estimating an outcome or value”, phrase “predictive model; make a prediction” and role “task”. Prediction is probabilistic, not certainty.

recommendation system — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “system ranking or suggesting items to users”, phrase “recommendation algorithm; recommended content” and role “task”. Visibility may reflect ranking criteria, not truth or importance.

automation — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “use of systems to perform tasks with reduced direct human action”, phrase “automate a task; automation” and role “application”. Automation does not necessarily mean AI.

synthetic content — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “content generated computationally rather than directly captured from reality”, phrase “synthetic image; synthetic media” and role “output”. Synthetic does not automatically mean deceptive.

deepfake — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “synthetic/manipulated media made to resemble real people/events”, phrase “deepfake video; detect deepfake” and role “risk”. Verification and disclosure matter.

confabulation — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “generated false or erroneous content presented confidently”, phrase “AI confabulation; verify confabulated output” and role “risk”. Often colloquially called hallucination.

hallucination — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “common informal term for AI-generated false or unsupported content”, phrase “AI hallucination; hallucinated citation” and role “risk”. Use with awareness that technical bodies may prefer confabulation.

bias — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “systematic tendency that can create skewed outcomes”, phrase “algorithmic bias; data bias” and role “risk”. Bias can enter through data, design, measurement or deployment.

fairness — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “principles and methods for avoiding unjustified disparities”, phrase “AI fairness; fairness assessment” and role “ethics”. Different fairness definitions can conflict.

transparency — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “openness about system purpose, limits, data or decision processes”, phrase “AI transparency; transparent use” and role “ethics”. Transparency can support informed evaluation.

explainability — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “ability to provide understandable reasons or accounts of system behaviour”, phrase “model explainability; explainable AI” and role “ethics”. What counts as an explanation depends on audience and task.

privacy — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “control/protection of personal information”, phrase “data privacy; privacy risk” and role “ethics”. Students should consider what data they provide.

personal data — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “information relating to an identifiable person”, phrase “personal-data protection; share personal data” and role “data”. Avoid unnecessary disclosure to tools.

security — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “protection against unauthorised access, misuse or attack”, phrase “AI security; cybersecurity” and role “risk”. Security and privacy overlap but are not identical.

accountability — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “responsibility for decisions and consequences”, phrase “human accountability; accountability mechanism” and role “ethics”. Responsibility should not disappear because AI was involved.

human oversight — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “human review, direction or intervention in system operation”, phrase “human oversight; human review” and role “governance”. Important for consequential decisions.

human agency — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “ability of people to make meaningful choices and retain control”, phrase “preserve human agency; human-centred design” and role “ethics”. UNESCO emphasises human-centred AI literacy.

plagiarism — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “using another’s work or ideas without appropriate acknowledgement”, phrase “avoid plagiarism; plagiarism policy” and role “school use”. AI use can create attribution questions; follow school rules.

attribution — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “identifying source or contribution”, phrase “proper attribution; attribute a source” and role “school use”. Attribution practices depend on task and institution.

citation — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “reference identifying a source”, phrase “verify citation; cite a source” and role “school use”. Generated citations may be false; verify them.

provenance — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “record of origin/history of content or data”, phrase “content provenance; data provenance” and role “verification”. Useful for synthetic media and sources.

verification — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “checking accuracy/authenticity”, phrase “verify output; independent verification” and role “practice”. Never treat fluent output as self-verifying.

evaluation — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “systematic assessment of quality or performance”, phrase “model evaluation; evaluate accuracy” and role “practice”. Metrics must match task and risk.

accuracy — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “degree to which output is correct under defined criteria”, phrase “accuracy rate; factual accuracy” and role “quality”. Correctness depends on task definition.

reliability — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “consistency/dependability of performance”, phrase “reliable system; reliability testing” and role “quality”. Reliable output can still be systematically wrong.

responsible use — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “use that considers accuracy, privacy, fairness, rules and human responsibility”, phrase “responsible AI use; responsible practice” and role “ethics”. Context determines appropriate use.

AI Literacy Laboratory Cycle 7

artificial intelligence — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “computer systems performing tasks associated with perception, learning, reasoning, communication or action”, phrase “AI system; artificial intelligence application” and role “core”. AI is a broad category, not one single technology.

generative AI — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “AI models designed to generate synthetic content”, phrase “generative AI model; GenAI output” and role “core”. Generation is different from retrieval or classification.

machine learning — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “methods in which systems learn patterns from data for tasks”, phrase “machine-learning model; train a model” and role “technique”. Not every software rule is machine learning.

model — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “computational representation used to make predictions or generate outputs”, phrase “AI model; trained model” and role “system”. A model is not the same as the full product or service.

algorithm — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “set of procedures or computational steps”, phrase “ranking algorithm; algorithmic process” and role “system”. Algorithm is broader than AI.

training data — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “data used to fit or develop a model”, phrase “training dataset; training-data quality” and role “data”. Data quality and representativeness matter.

dataset — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “organised collection of data”, phrase “large dataset; labelled dataset” and role “data”. A dataset is not automatically unbiased or accurate.

input — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “information given to a system”, phrase “user input; input data” and role “interaction”. Input may include text, image, audio or structured data.

prompt — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “instruction or context given to a generative system”, phrase “write a prompt; prompt design” and role “interaction”. Prompt quality can affect output but cannot guarantee truth.

output — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “result produced by a system”, phrase “model output; generated output” and role “interaction”. Output still needs evaluation.

token — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “small unit used in text processing by many language models”, phrase “input tokens; token sequence” and role “technical”. Tokens are not always whole words.

parameter — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “learned or configured value influencing model behaviour”, phrase “model parameter; parameter count” and role “technical”. Parameter count alone does not determine usefulness.

inference — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “running a trained model to produce a prediction/output”, phrase “model inference; inference time” and role “technical”. Different from reading-comprehension inference despite same word.

classification — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “assigning inputs to categories”, phrase “image classification; classify text” and role “task”. Different from generation.

prediction — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “estimating an outcome or value”, phrase “predictive model; make a prediction” and role “task”. Prediction is probabilistic, not certainty.

recommendation system — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “system ranking or suggesting items to users”, phrase “recommendation algorithm; recommended content” and role “task”. Visibility may reflect ranking criteria, not truth or importance.

automation — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “use of systems to perform tasks with reduced direct human action”, phrase “automate a task; automation” and role “application”. Automation does not necessarily mean AI.

synthetic content — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “content generated computationally rather than directly captured from reality”, phrase “synthetic image; synthetic media” and role “output”. Synthetic does not automatically mean deceptive.

deepfake — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “synthetic/manipulated media made to resemble real people/events”, phrase “deepfake video; detect deepfake” and role “risk”. Verification and disclosure matter.

confabulation — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “generated false or erroneous content presented confidently”, phrase “AI confabulation; verify confabulated output” and role “risk”. Often colloquially called hallucination.

hallucination — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “common informal term for AI-generated false or unsupported content”, phrase “AI hallucination; hallucinated citation” and role “risk”. Use with awareness that technical bodies may prefer confabulation.

bias — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “systematic tendency that can create skewed outcomes”, phrase “algorithmic bias; data bias” and role “risk”. Bias can enter through data, design, measurement or deployment.

fairness — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “principles and methods for avoiding unjustified disparities”, phrase “AI fairness; fairness assessment” and role “ethics”. Different fairness definitions can conflict.

transparency — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “openness about system purpose, limits, data or decision processes”, phrase “AI transparency; transparent use” and role “ethics”. Transparency can support informed evaluation.

explainability — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “ability to provide understandable reasons or accounts of system behaviour”, phrase “model explainability; explainable AI” and role “ethics”. What counts as an explanation depends on audience and task.

privacy — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “control/protection of personal information”, phrase “data privacy; privacy risk” and role “ethics”. Students should consider what data they provide.

personal data — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “information relating to an identifiable person”, phrase “personal-data protection; share personal data” and role “data”. Avoid unnecessary disclosure to tools.

security — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “protection against unauthorised access, misuse or attack”, phrase “AI security; cybersecurity” and role “risk”. Security and privacy overlap but are not identical.

accountability — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “responsibility for decisions and consequences”, phrase “human accountability; accountability mechanism” and role “ethics”. Responsibility should not disappear because AI was involved.

human oversight — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “human review, direction or intervention in system operation”, phrase “human oversight; human review” and role “governance”. Important for consequential decisions.

human agency — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “ability of people to make meaningful choices and retain control”, phrase “preserve human agency; human-centred design” and role “ethics”. UNESCO emphasises human-centred AI literacy.

plagiarism — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “using another’s work or ideas without appropriate acknowledgement”, phrase “avoid plagiarism; plagiarism policy” and role “school use”. AI use can create attribution questions; follow school rules.

attribution — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “identifying source or contribution”, phrase “proper attribution; attribute a source” and role “school use”. Attribution practices depend on task and institution.

citation — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “reference identifying a source”, phrase “verify citation; cite a source” and role “school use”. Generated citations may be false; verify them.

provenance — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “record of origin/history of content or data”, phrase “content provenance; data provenance” and role “verification”. Useful for synthetic media and sources.

verification — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “checking accuracy/authenticity”, phrase “verify output; independent verification” and role “practice”. Never treat fluent output as self-verifying.

evaluation — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “systematic assessment of quality or performance”, phrase “model evaluation; evaluate accuracy” and role “practice”. Metrics must match task and risk.

accuracy — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “degree to which output is correct under defined criteria”, phrase “accuracy rate; factual accuracy” and role “quality”. Correctness depends on task definition.

reliability — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “consistency/dependability of performance”, phrase “reliable system; reliability testing” and role “quality”. Reliable output can still be systematically wrong.

responsible use — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “use that considers accuracy, privacy, fairness, rules and human responsibility”, phrase “responsible AI use; responsible practice” and role “ethics”. Context determines appropriate use.

AI Literacy Laboratory Cycle 8

artificial intelligence — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “computer systems performing tasks associated with perception, learning, reasoning, communication or action”, phrase “AI system; artificial intelligence application” and role “core”. AI is a broad category, not one single technology.

generative AI — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “AI models designed to generate synthetic content”, phrase “generative AI model; GenAI output” and role “core”. Generation is different from retrieval or classification.

machine learning — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “methods in which systems learn patterns from data for tasks”, phrase “machine-learning model; train a model” and role “technique”. Not every software rule is machine learning.

model — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “computational representation used to make predictions or generate outputs”, phrase “AI model; trained model” and role “system”. A model is not the same as the full product or service.

algorithm — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “set of procedures or computational steps”, phrase “ranking algorithm; algorithmic process” and role “system”. Algorithm is broader than AI.

training data — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “data used to fit or develop a model”, phrase “training dataset; training-data quality” and role “data”. Data quality and representativeness matter.

dataset — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “organised collection of data”, phrase “large dataset; labelled dataset” and role “data”. A dataset is not automatically unbiased or accurate.

input — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “information given to a system”, phrase “user input; input data” and role “interaction”. Input may include text, image, audio or structured data.

prompt — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “instruction or context given to a generative system”, phrase “write a prompt; prompt design” and role “interaction”. Prompt quality can affect output but cannot guarantee truth.

output — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “result produced by a system”, phrase “model output; generated output” and role “interaction”. Output still needs evaluation.

token — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “small unit used in text processing by many language models”, phrase “input tokens; token sequence” and role “technical”. Tokens are not always whole words.

parameter — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “learned or configured value influencing model behaviour”, phrase “model parameter; parameter count” and role “technical”. Parameter count alone does not determine usefulness.

inference — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “running a trained model to produce a prediction/output”, phrase “model inference; inference time” and role “technical”. Different from reading-comprehension inference despite same word.

classification — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “assigning inputs to categories”, phrase “image classification; classify text” and role “task”. Different from generation.

prediction — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “estimating an outcome or value”, phrase “predictive model; make a prediction” and role “task”. Prediction is probabilistic, not certainty.

recommendation system — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “system ranking or suggesting items to users”, phrase “recommendation algorithm; recommended content” and role “task”. Visibility may reflect ranking criteria, not truth or importance.

automation — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “use of systems to perform tasks with reduced direct human action”, phrase “automate a task; automation” and role “application”. Automation does not necessarily mean AI.

synthetic content — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “content generated computationally rather than directly captured from reality”, phrase “synthetic image; synthetic media” and role “output”. Synthetic does not automatically mean deceptive.

deepfake — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “synthetic/manipulated media made to resemble real people/events”, phrase “deepfake video; detect deepfake” and role “risk”. Verification and disclosure matter.

confabulation — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “generated false or erroneous content presented confidently”, phrase “AI confabulation; verify confabulated output” and role “risk”. Often colloquially called hallucination.

hallucination — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “common informal term for AI-generated false or unsupported content”, phrase “AI hallucination; hallucinated citation” and role “risk”. Use with awareness that technical bodies may prefer confabulation.

bias — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “systematic tendency that can create skewed outcomes”, phrase “algorithmic bias; data bias” and role “risk”. Bias can enter through data, design, measurement or deployment.

fairness — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “principles and methods for avoiding unjustified disparities”, phrase “AI fairness; fairness assessment” and role “ethics”. Different fairness definitions can conflict.

transparency — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “openness about system purpose, limits, data or decision processes”, phrase “AI transparency; transparent use” and role “ethics”. Transparency can support informed evaluation.

explainability — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “ability to provide understandable reasons or accounts of system behaviour”, phrase “model explainability; explainable AI” and role “ethics”. What counts as an explanation depends on audience and task.

privacy — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “control/protection of personal information”, phrase “data privacy; privacy risk” and role “ethics”. Students should consider what data they provide.

personal data — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “information relating to an identifiable person”, phrase “personal-data protection; share personal data” and role “data”. Avoid unnecessary disclosure to tools.

security — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “protection against unauthorised access, misuse or attack”, phrase “AI security; cybersecurity” and role “risk”. Security and privacy overlap but are not identical.

accountability — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “responsibility for decisions and consequences”, phrase “human accountability; accountability mechanism” and role “ethics”. Responsibility should not disappear because AI was involved.

human oversight — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “human review, direction or intervention in system operation”, phrase “human oversight; human review” and role “governance”. Important for consequential decisions.

human agency — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “ability of people to make meaningful choices and retain control”, phrase “preserve human agency; human-centred design” and role “ethics”. UNESCO emphasises human-centred AI literacy.

plagiarism — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “using another’s work or ideas without appropriate acknowledgement”, phrase “avoid plagiarism; plagiarism policy” and role “school use”. AI use can create attribution questions; follow school rules.

attribution — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “identifying source or contribution”, phrase “proper attribution; attribute a source” and role “school use”. Attribution practices depend on task and institution.

citation — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “reference identifying a source”, phrase “verify citation; cite a source” and role “school use”. Generated citations may be false; verify them.

provenance — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “record of origin/history of content or data”, phrase “content provenance; data provenance” and role “verification”. Useful for synthetic media and sources.

verification — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “checking accuracy/authenticity”, phrase “verify output; independent verification” and role “practice”. Never treat fluent output as self-verifying.

evaluation — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “systematic assessment of quality or performance”, phrase “model evaluation; evaluate accuracy” and role “practice”. Metrics must match task and risk.

accuracy — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “degree to which output is correct under defined criteria”, phrase “accuracy rate; factual accuracy” and role “quality”. Correctness depends on task definition.

reliability — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “consistency/dependability of performance”, phrase “reliable system; reliability testing” and role “quality”. Reliable output can still be systematically wrong.

responsible use — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “use that considers accuracy, privacy, fairness, rules and human responsibility”, phrase “responsible AI use; responsible practice” and role “ethics”. Context determines appropriate use.

AI Literacy Laboratory Cycle 9

artificial intelligence — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “computer systems performing tasks associated with perception, learning, reasoning, communication or action”, phrase “AI system; artificial intelligence application” and role “core”. AI is a broad category, not one single technology.

generative AI — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “AI models designed to generate synthetic content”, phrase “generative AI model; GenAI output” and role “core”. Generation is different from retrieval or classification.

machine learning — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “methods in which systems learn patterns from data for tasks”, phrase “machine-learning model; train a model” and role “technique”. Not every software rule is machine learning.

model — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “computational representation used to make predictions or generate outputs”, phrase “AI model; trained model” and role “system”. A model is not the same as the full product or service.

algorithm — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “set of procedures or computational steps”, phrase “ranking algorithm; algorithmic process” and role “system”. Algorithm is broader than AI.

training data — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “data used to fit or develop a model”, phrase “training dataset; training-data quality” and role “data”. Data quality and representativeness matter.

dataset — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “organised collection of data”, phrase “large dataset; labelled dataset” and role “data”. A dataset is not automatically unbiased or accurate.

input — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “information given to a system”, phrase “user input; input data” and role “interaction”. Input may include text, image, audio or structured data.

prompt — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “instruction or context given to a generative system”, phrase “write a prompt; prompt design” and role “interaction”. Prompt quality can affect output but cannot guarantee truth.

output — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “result produced by a system”, phrase “model output; generated output” and role “interaction”. Output still needs evaluation.

token — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “small unit used in text processing by many language models”, phrase “input tokens; token sequence” and role “technical”. Tokens are not always whole words.

parameter — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “learned or configured value influencing model behaviour”, phrase “model parameter; parameter count” and role “technical”. Parameter count alone does not determine usefulness.

inference — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “running a trained model to produce a prediction/output”, phrase “model inference; inference time” and role “technical”. Different from reading-comprehension inference despite same word.

classification — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “assigning inputs to categories”, phrase “image classification; classify text” and role “task”. Different from generation.

prediction — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “estimating an outcome or value”, phrase “predictive model; make a prediction” and role “task”. Prediction is probabilistic, not certainty.

recommendation system — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “system ranking or suggesting items to users”, phrase “recommendation algorithm; recommended content” and role “task”. Visibility may reflect ranking criteria, not truth or importance.

automation — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “use of systems to perform tasks with reduced direct human action”, phrase “automate a task; automation” and role “application”. Automation does not necessarily mean AI.

synthetic content — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “content generated computationally rather than directly captured from reality”, phrase “synthetic image; synthetic media” and role “output”. Synthetic does not automatically mean deceptive.

deepfake — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “synthetic/manipulated media made to resemble real people/events”, phrase “deepfake video; detect deepfake” and role “risk”. Verification and disclosure matter.

confabulation — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “generated false or erroneous content presented confidently”, phrase “AI confabulation; verify confabulated output” and role “risk”. Often colloquially called hallucination.

hallucination — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “common informal term for AI-generated false or unsupported content”, phrase “AI hallucination; hallucinated citation” and role “risk”. Use with awareness that technical bodies may prefer confabulation.

bias — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “systematic tendency that can create skewed outcomes”, phrase “algorithmic bias; data bias” and role “risk”. Bias can enter through data, design, measurement or deployment.

fairness — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “principles and methods for avoiding unjustified disparities”, phrase “AI fairness; fairness assessment” and role “ethics”. Different fairness definitions can conflict.

transparency — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “openness about system purpose, limits, data or decision processes”, phrase “AI transparency; transparent use” and role “ethics”. Transparency can support informed evaluation.

explainability — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “ability to provide understandable reasons or accounts of system behaviour”, phrase “model explainability; explainable AI” and role “ethics”. What counts as an explanation depends on audience and task.

privacy — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “control/protection of personal information”, phrase “data privacy; privacy risk” and role “ethics”. Students should consider what data they provide.

personal data — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “information relating to an identifiable person”, phrase “personal-data protection; share personal data” and role “data”. Avoid unnecessary disclosure to tools.

security — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “protection against unauthorised access, misuse or attack”, phrase “AI security; cybersecurity” and role “risk”. Security and privacy overlap but are not identical.

accountability — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “responsibility for decisions and consequences”, phrase “human accountability; accountability mechanism” and role “ethics”. Responsibility should not disappear because AI was involved.

human oversight — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “human review, direction or intervention in system operation”, phrase “human oversight; human review” and role “governance”. Important for consequential decisions.

human agency — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “ability of people to make meaningful choices and retain control”, phrase “preserve human agency; human-centred design” and role “ethics”. UNESCO emphasises human-centred AI literacy.

plagiarism — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “using another’s work or ideas without appropriate acknowledgement”, phrase “avoid plagiarism; plagiarism policy” and role “school use”. AI use can create attribution questions; follow school rules.

attribution — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “identifying source or contribution”, phrase “proper attribution; attribute a source” and role “school use”. Attribution practices depend on task and institution.

citation — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “reference identifying a source”, phrase “verify citation; cite a source” and role “school use”. Generated citations may be false; verify them.

provenance — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “record of origin/history of content or data”, phrase “content provenance; data provenance” and role “verification”. Useful for synthetic media and sources.

verification — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “checking accuracy/authenticity”, phrase “verify output; independent verification” and role “practice”. Never treat fluent output as self-verifying.

evaluation — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “systematic assessment of quality or performance”, phrase “model evaluation; evaluate accuracy” and role “practice”. Metrics must match task and risk.

accuracy — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “degree to which output is correct under defined criteria”, phrase “accuracy rate; factual accuracy” and role “quality”. Correctness depends on task definition.

reliability — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “consistency/dependability of performance”, phrase “reliable system; reliability testing” and role “quality”. Reliable output can still be systematically wrong.

responsible use — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “use that considers accuracy, privacy, fairness, rules and human responsibility”, phrase “responsible AI use; responsible practice” and role “ethics”. Context determines appropriate use.

AI Literacy Laboratory Cycle 10

artificial intelligence — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “computer systems performing tasks associated with perception, learning, reasoning, communication or action”, phrase “AI system; artificial intelligence application” and role “core”. AI is a broad category, not one single technology.

generative AI — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “AI models designed to generate synthetic content”, phrase “generative AI model; GenAI output” and role “core”. Generation is different from retrieval or classification.

machine learning — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “methods in which systems learn patterns from data for tasks”, phrase “machine-learning model; train a model” and role “technique”. Not every software rule is machine learning.

model — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “computational representation used to make predictions or generate outputs”, phrase “AI model; trained model” and role “system”. A model is not the same as the full product or service.

algorithm — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “set of procedures or computational steps”, phrase “ranking algorithm; algorithmic process” and role “system”. Algorithm is broader than AI.

training data — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “data used to fit or develop a model”, phrase “training dataset; training-data quality” and role “data”. Data quality and representativeness matter.

dataset — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “organised collection of data”, phrase “large dataset; labelled dataset” and role “data”. A dataset is not automatically unbiased or accurate.

input — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “information given to a system”, phrase “user input; input data” and role “interaction”. Input may include text, image, audio or structured data.

prompt — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “instruction or context given to a generative system”, phrase “write a prompt; prompt design” and role “interaction”. Prompt quality can affect output but cannot guarantee truth.

output — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “result produced by a system”, phrase “model output; generated output” and role “interaction”. Output still needs evaluation.

token — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “small unit used in text processing by many language models”, phrase “input tokens; token sequence” and role “technical”. Tokens are not always whole words.

parameter — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “learned or configured value influencing model behaviour”, phrase “model parameter; parameter count” and role “technical”. Parameter count alone does not determine usefulness.

inference — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “running a trained model to produce a prediction/output”, phrase “model inference; inference time” and role “technical”. Different from reading-comprehension inference despite same word.

classification — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “assigning inputs to categories”, phrase “image classification; classify text” and role “task”. Different from generation.

prediction — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “estimating an outcome or value”, phrase “predictive model; make a prediction” and role “task”. Prediction is probabilistic, not certainty.

recommendation system — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “system ranking or suggesting items to users”, phrase “recommendation algorithm; recommended content” and role “task”. Visibility may reflect ranking criteria, not truth or importance.

automation — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “use of systems to perform tasks with reduced direct human action”, phrase “automate a task; automation” and role “application”. Automation does not necessarily mean AI.

synthetic content — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “content generated computationally rather than directly captured from reality”, phrase “synthetic image; synthetic media” and role “output”. Synthetic does not automatically mean deceptive.

deepfake — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “synthetic/manipulated media made to resemble real people/events”, phrase “deepfake video; detect deepfake” and role “risk”. Verification and disclosure matter.

confabulation — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “generated false or erroneous content presented confidently”, phrase “AI confabulation; verify confabulated output” and role “risk”. Often colloquially called hallucination.

hallucination — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “common informal term for AI-generated false or unsupported content”, phrase “AI hallucination; hallucinated citation” and role “risk”. Use with awareness that technical bodies may prefer confabulation.

bias — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “systematic tendency that can create skewed outcomes”, phrase “algorithmic bias; data bias” and role “risk”. Bias can enter through data, design, measurement or deployment.

fairness — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “principles and methods for avoiding unjustified disparities”, phrase “AI fairness; fairness assessment” and role “ethics”. Different fairness definitions can conflict.

transparency — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “openness about system purpose, limits, data or decision processes”, phrase “AI transparency; transparent use” and role “ethics”. Transparency can support informed evaluation.

explainability — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “ability to provide understandable reasons or accounts of system behaviour”, phrase “model explainability; explainable AI” and role “ethics”. What counts as an explanation depends on audience and task.

privacy — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “control/protection of personal information”, phrase “data privacy; privacy risk” and role “ethics”. Students should consider what data they provide.

personal data — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “information relating to an identifiable person”, phrase “personal-data protection; share personal data” and role “data”. Avoid unnecessary disclosure to tools.

security — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “protection against unauthorised access, misuse or attack”, phrase “AI security; cybersecurity” and role “risk”. Security and privacy overlap but are not identical.

accountability — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “responsibility for decisions and consequences”, phrase “human accountability; accountability mechanism” and role “ethics”. Responsibility should not disappear because AI was involved.

human oversight — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “human review, direction or intervention in system operation”, phrase “human oversight; human review” and role “governance”. Important for consequential decisions.

human agency — Design a simple system diagram using the target: input, process/model, output, evaluation and human responsibility. Anchor your work in “ability of people to make meaningful choices and retain control”, phrase “preserve human agency; human-centred design” and role “ethics”. UNESCO emphasises human-centred AI literacy.

plagiarism — Write a claim about AI that sounds plausible but overgeneralises. Repair it using the target term precisely. Anchor your work in “using another’s work or ideas without appropriate acknowledgement”, phrase “avoid plagiarism; plagiarism policy” and role “school use”. AI use can create attribution questions; follow school rules.

attribution — Create a generative-AI output containing one confabulated fact. Write the verification steps a student should take. Anchor your work in “identifying source or contribution”, phrase “proper attribution; attribute a source” and role “school use”. Attribution practices depend on task and institution.

citation — Compare an AI system with a non-AI automated system. Explain which vocabulary distinguishes them. Anchor your work in “reference identifying a source”, phrase “verify citation; cite a source” and role “school use”. Generated citations may be false; verify them.

provenance — Write a classroom scenario where the target raises an ethical issue. State what information is needed before judging. Anchor your work in “record of origin/history of content or data”, phrase “content provenance; data provenance” and role “verification”. Useful for synthetic media and sources.

verification — Create a media-literacy example involving synthetic content. Explain which provenance or disclosure evidence matters. Anchor your work in “checking accuracy/authenticity”, phrase “verify output; independent verification” and role “practice”. Never treat fluent output as self-verifying.

evaluation — Use the target in a short argumentative paragraph with one benefit, one risk and one condition. Anchor your work in “systematic assessment of quality or performance”, phrase “model evaluation; evaluate accuracy” and role “practice”. Metrics must match task and risk.

accuracy — Explain the target to a younger student without anthropomorphising the system. Anchor your work in “degree to which output is correct under defined criteria”, phrase “accuracy rate; factual accuracy” and role “quality”. Correctness depends on task definition.

reliability — Write a misuse of the target based on a common buzzword confusion. Correct it. Anchor your work in “consistency/dependability of performance”, phrase “reliable system; reliability testing” and role “quality”. Reliable output can still be systematically wrong.

responsible use — Create a follow-up question that tests whether a student understands the boundary between the target and a neighbouring term. Anchor your work in “use that considers accuracy, privacy, fairness, rules and human responsibility”, phrase “responsible AI use; responsible practice” and role “ethics”. Context determines appropriate use.