How to Use Super Intelligence for Critical Thinking | Claims, Evidence, Assumptions and Independent Judgment

The Critical-Thinking Stack

A strong critical-thinking process can be organised into layers. Start with the claim, then source, evidence, inference, assumptions, alternatives, uncertainty and finally judgment. Each layer asks a different question.

• Claim: What exactly is being asserted?
• Source: Who or what is making the claim?
• Evidence: What observations or records support it?
• Inference: How does the conclusion follow from the evidence?
• Assumptions: What must be true for the inference to work?
• Alternatives: What other explanations fit the same evidence?
• Uncertainty: What is missing, weak or contested?
• Judgment: What level of confidence or action is justified?

The stack protects users from skipping directly from claim to judgment. SI can help populate each layer, but the user should inspect the transitions.

Argument Mapping

Argument mapping turns a paragraph into claims and relationships. Identify the main conclusion, supporting reasons, evidence and objections.

Example: “Schools should use more online learning because digital tools improve access.” The main conclusion concerns school policy. The reason is improved access. Evidence would need to show which students gain access, what form of learning, what infrastructure exists and whether access improves actual learning outcomes.

Mapping separates rhetorical flow from logical structure. A persuasive paragraph may contain one unsupported bridge that becomes obvious when represented as a map.

Premise Quality

A valid inference can still produce a weak conclusion when its premises are false or unsupported. Evaluate premise quality separately from logical structure.

Ask whether each premise is directly supported, inferred, assumed or value-based. Weak premises should lower confidence even if the argument form is valid.

SI can classify premises, but evidence verification remains external.

Validity and Soundness

Validity asks whether the conclusion follows from the premises if the premises are true. Soundness adds the requirement that the premises are actually true or well supported.

This distinction helps when an argument is logically neat but factually weak. For example: “All students learn best in groups; Mei is a student; therefore Mei learns best in groups.” The form may be valid, but the first premise is unsupported and overgeneralised.

Critical thinking checks both structure and evidence.

Deductive and Inductive Reasoning

Deductive reasoning aims for conclusions that necessarily follow from premises. Inductive reasoning supports conclusions probabilistically from patterns, samples or observations.

Many real-world claims are inductive. Evidence can strengthen them without proving them absolutely. SI answers often blur this distinction by using strong language for probabilistic conclusions.

Ask what type of reasoning the task requires and match the conclusion strength accordingly.

Abductive Reasoning

Abduction asks for the best explanation of observed evidence. It is common in diagnosis, debugging, history and science.

Abductive reasoning requires comparing several explanations. The best explanation may be the one that fits more evidence with fewer unsupported assumptions, but it remains revisable.

Use SI to generate hypotheses, then test them rather than accepting the most elegant narrative.

Analogical Reasoning

Analogies transfer structure from one case to another. They can accelerate understanding but also import irrelevant assumptions.

Evaluate analogy by asking which relationships truly match and where the cases differ. “An organisation is like a machine” highlights components and coordination but can hide human incentives and interpretation.

A good analogy includes its own failure boundary.

Causal Reasoning

Causal claims require more than sequence or correlation. Ask whether the cause precedes the effect, whether plausible alternatives exist, whether interventions change outcomes and whether the mechanism is coherent.

Observational evidence can support causal hypotheses but may not establish them. Study design matters.

SI can help list confounders and causal pathways, but the evidence must come from the actual research or experiment.

Necessary and Sufficient Conditions

A necessary condition must be present for an outcome; a sufficient condition guarantees the outcome under the stated model. Confusing them creates bad arguments.

Example: oxygen is necessary for ordinary human survival but not sufficient for health. Passing one prerequisite may be necessary for a course without guaranteeing success.

Ask whether the argument treats a contributor as a guarantee or a requirement as the whole explanation.

Base Rates

Base rates are background frequencies. Ignoring them can make rare explanations appear more likely than they are.

If a diagnostic test is highly accurate but the condition is extremely rare, false positives can still matter. In everyday reasoning, dramatic cases can distort perception of commonness.

Ask what happens in the wider population before interpreting an isolated signal.

Survivorship Bias

Survivorship bias occurs when the visible cases are those that succeeded or remained observable while failures disappeared.

Business advice based only on successful founders can omit similar strategies used by failed firms. Study habits reported by top students may omit students who used the same habits without success.

Ask who or what is missing from the sample.

Selection Bias

Selection bias occurs when inclusion in the data is related to the outcome or characteristics being studied.

Online surveys, voluntary programmes and self-selected communities often differ from the wider population.

SI can flag possible selection mechanisms, but the study design or data collection process determines whether they are real.

Availability Bias

Memorable examples can seem more common than they are. Recent or vivid incidents receive disproportionate weight.

Use data rather than memory when frequency matters. Ask SI to help find relevant statistics, then verify source and denominator.

Critical thinking replaces ease of recall with evidence proportionality.

Anchoring

The first number or explanation can shape later judgment even when it is arbitrary. AI responses can create anchors because the first answer often becomes the frame for every follow-up.

To reduce anchoring, generate an independent analysis before revealing a preferred answer, or ask for several alternatives under different assumptions.

Then compare against evidence rather than against the first generated suggestion.

Framing Effects

The same information can produce different reactions depending on presentation. “90% survival” and “10% mortality” describe the same rate but frame it differently.

When a decision feels obvious, restate key quantities in complementary forms. Ask whether the conclusion changes when the framing changes.

A robust decision should not depend entirely on emotionally loaded presentation.

Sunk-Cost Reasoning

Past investment can influence whether people continue a failing plan, even when the past cost cannot be recovered.

Critical thinking asks what decision would be made if starting from the current state today. Future costs and benefits matter more than unrecoverable past expenditure.

This does not mean history is irrelevant; switching costs, commitments and learning can still matter. Separate them from sunk cost itself.

Status-Quo Bias

Existing arrangements can feel safer merely because they already exist. Ask whether the status quo would be chosen if it were one option among several today.

At the same time, change has real switching costs and uncertainty. Critical thinking does not assume change is better.

Compare status quo and alternatives using the same criteria.

Motivated Reasoning

People interpret evidence in ways that protect identity, preference or desired outcomes. AI can amplify this when users ask leading questions.

Use neutral prompts: “What evidence supports and weakens this claim?” rather than “Explain why my plan is correct.”

Record what evidence would change your view before searching. This reduces moving the goalposts after evidence appears.

Argument From Authority

Expert testimony can be valuable, but the strength depends on expertise, domain relevance, evidence and consensus. Authority alone does not make every claim correct.

Ask whether the source has relevant expertise, whether the statement falls inside that expertise and whether the claim is supported by evidence.

Critical thinking respects expertise without treating it as infallible.

Argument From Popularity

A widely held belief can be true or false. Popularity is evidence about belief prevalence, not necessarily about underlying reality.

Community consensus can matter for norms, user experience or practical adoption, but separate those questions from factual truth.

Use the right evidence for the right claim.

Straw-Man Reasoning

A straw man weakens or distorts an opposing position so it becomes easier to attack.

Ask SI to restate the opposing argument in a form its strongest advocate would recognise before criticising it.

Then critique the strongest relevant version, not a caricature.

Steel-Manning With Caution

Steel-manning attempts to construct the strongest plausible version of an argument. It can reduce unfair critique.

However, do not invent a stronger argument that the actual source never made and then attribute it back to them. Label the reconstruction as your interpretation.

Critical thinking preserves both fairness and source fidelity.

False Equivalence

Two positions can both have weaknesses without having equal evidence. Critical thinking should acknowledge disagreement without pretending every side has identical support.

Compare quality, quantity and directness of evidence.

Neutral analysis means fair treatment of evidence, not forced symmetry.

Moving the Goalposts

A claim becomes difficult to test when its success condition changes after evidence appears.

Define acceptance criteria before evaluation. If the criterion legitimately changes, record why.

This protects both research and everyday decision making from post-hoc rationalisation.

Cherry-Picking

Cherry-picking selects evidence that supports a conclusion while ignoring relevant contrary evidence.

Use a transparent search strategy, inclusion criteria or source map for important research.

Ask SI which relevant evidence would challenge the current set, then verify it.

Correlation and Causation

Correlation shows variables vary together. Causation claims one influences another. Confounding, reverse causation and coincidence can produce correlation.

When SI uses causal verbs, inspect the study design. Experiments, natural experiments and stronger causal methods can provide more support than simple association.

Use language strength that matches evidence strength.

Regression to the Mean

Extreme results often move closer to average on later measurement even without intervention. This can make ineffective actions look beneficial.

Critical thinking asks whether improvement would have been expected partly from natural variation.

This matters in education, medicine, performance and operations when interventions follow unusually bad outcomes.

Confounding

A confounder influences both the supposed cause and outcome, creating a misleading relationship.

For example, motivation may influence both tutoring participation and test performance.

Ask what variables could plausibly affect both sides and whether the design accounts for them.

Measurement Error

Bad measurement weakens every later inference. If “productivity” is measured only by messages sent, the metric may reward activity rather than useful output.

Inspect operational definitions, instrument reliability and data collection.

SI can analyse the metric only after the metric itself is understood.

Goodhart’s Law as a Critical-Thinking Warning

When a measure becomes a target, people may optimise the measure rather than the underlying goal. This is a common systems problem.

If schools optimise test scores, firms optimise clicks or support teams optimise ticket closure, quality can shift in unintended ways.

Use multiple measures and receiver outcomes to reduce metric gaming.

Statistical Significance Versus Practical Significance

A statistically detectable effect can be too small to matter operationally. A practically important effect can be uncertain in a small sample.

Inspect effect size, uncertainty, sample size and real-world consequence rather than one threshold.

Critical thinking connects statistical evidence to the decision context.

Relative Risk Versus Absolute Risk

Relative changes can sound dramatic without the baseline. A risk doubling from 1 in 10,000 to 2 in 10,000 differs from doubling from 10% to 20%.

Ask for absolute numbers when risk communication matters.

This is another denominator and baseline check.

A Critical-Thinking Source Matrix

• Claim being evaluated.
• Source type.
• Authority for this claim.
• Date and population.
• Direct evidence.
• Method limitations.
• Conflicts with other sources.
• Interpretation versus source fact.
• Confidence level justified.
• Next evidence needed.

The matrix turns vague trust judgments into inspectable source reasoning.

A Critical-Thinking Conversation Protocol

When using SI interactively, assign roles over time. First Map: identify claims and evidence. Then Challenge: find assumptions and alternatives. Then Verify: locate sources or calculations. Finally Conclude: state bounded judgment.

Do not ask the model to perform every role in one opaque response if the task matters. Staging makes errors easier to locate.

Keep your own provisional view separate so you can compare rather than merely follow the generated frame.

A Worked Case Study: Product Claim

Claim: “This app doubles productivity.” Ask: how is productivity measured? What is the baseline? Who was studied? Was there a control group? Over what duration?

If evidence is a company survey of selected users, the claim may need narrowing. The critical-thinking output might become: “Some surveyed users reported substantial time savings; this does not establish a universal doubling of productivity.”

The repair is not cynical rejection. It is matching wording to evidence.

A Worked Case Study: School Intervention

Claim: “Small classes cause better results.” Gather study design, age group, teacher quality, subject and outcome. Alternative mechanisms may include more feedback or selection effects.

If evidence differs across contexts, preserve the variation rather than producing one global rule.

Policy decisions may still favour smaller classes for reasons beyond test scores; those value judgments should be separated from the empirical claim.

A Worked Case Study: Workplace Automation

Claim: “Automating triage will save staff time.” Baseline current triage time, correction rate and downstream errors. Test a pilot.

If automation saves initial sorting time but doubles correction work, the original claim fails at system level.

Critical thinking uses end-to-end evidence rather than a local metric.

A Worked Case Study: Historical Interpretation

Two historians explain the same event differently. Compare primary sources, evidence selection, definitions and causal emphasis.

The disagreement may reflect incomplete evidence rather than one side being irrational. Critical thinking can map what each explanation accounts for and what it leaves unexplained.

Avoid asking SI for a winner before understanding the historiography.

A Worked Case Study: Personal Decision

Claim: “I should quit this project because progress is slow.” Separate current progress, objective, sunk cost, opportunity cost and remaining value.

Slow progress may indicate poor strategy, insufficient time, a wrong objective or simply a difficult but worthwhile project.

A critical-thinking process converts emotional conclusion into explicit decision factors.

Critical-Thinking Practice Set

• Rewrite one broad claim narrowly.
• Identify one hidden assumption.
• Find one plausible alternative explanation.
• Locate the denominator behind one percentage.
• Check one source’s population and date.
• Convert one causal claim into association language and ask what evidence would justify causality.
• Steel-man one opposing position without misattribution.
• State one piece of evidence that would change your mind.
• Write one bounded conclusion.
• Identify where a human value judgment enters the decision.

The Critical-Thinking Maintenance Rule

Critical-thinking habits can decay under time pressure or repeated agreement from AI. Maintain them with occasional adversarial reviews and independent source checks.

For recurring high-consequence workflows, build the key questions into a checklist or schema rather than relying on memory.

Retire questions that never alter decisions, but preserve those that catch important historical failures.

The Final Critical-Thinking Gate

Before accepting an important SI-assisted conclusion, explain the claim, strongest supporting evidence, strongest limitation, main alternative explanation and what would change your mind.

Then verify the decisive fact or calculation independently. If the conclusion depends mostly on a value judgment, state that judgment explicitly rather than dressing it as fact.

Finally, ask whether another reasonable reader with the same evidence could disagree. If yes, identify where the disagreement enters: evidence weighting, assumptions or values.

A mature critical-thinking process does not eliminate uncertainty or disagreement. It makes the structure of both visible enough for independent judgment.

Critical Thinking Under Time Pressure

Time pressure makes people rely on familiar frames, vivid evidence and first answers. SI can help by quickly structuring the question, but speed increases the risk that an unsupported frame becomes the default.

Use a minimum critical-thinking protocol for urgent work: state the decision, identify the strongest fact, identify the largest unknown, name one alternative explanation and define one stop condition.

This protocol is intentionally smaller than the full framework. The goal is to preserve the highest-value checks when there is no time for exhaustive analysis.

Critical Thinking Under Information Overload

When there are too many sources, the problem becomes selection rather than access. Define the claim first, then retrieve only evidence capable of changing the conclusion.

Use source matrices, dates and scope. Remove duplicate articles that all repeat one underlying source.

SI can summarise a large collection, but the user should know which original records the synthesis depends on.

Critical Thinking Under Uncertainty

Uncertainty should affect both conclusion strength and action size. When evidence is weak, reversible actions and information-gathering steps may be more appropriate than large commitments.

State uncertainty explicitly rather than hiding it behind broad caveats. Which variable is unknown? How much could it change the decision?

This makes uncertainty operational rather than rhetorical.

Critical Thinking and Decision Thresholds

Not every decision requires certainty. Define the threshold of evidence needed for the action.

A reversible low-cost experiment can proceed with less certainty than an irreversible high-consequence commitment. The evidence standard should scale with consequence.

SI can help compare thresholds and scenarios, while the responsible human sets the final action standard.

Critical Thinking and Receiver Harm

A conclusion can be logically defensible yet communicated in a way that misleads the receiver. Ask what a reader is likely to infer beyond the literal wording.

For example, a school report stating that one group scored higher may be interpreted as a causal claim even if the report only describes averages. Add scope and limitation where the receiver could reasonably overread the evidence.

Receiver-centred critical thinking therefore includes communication, not only internal analysis.

Critical Thinking in Group Discussions

Groups can amplify shared assumptions and social pressure. Use structured roles: one person states evidence, another identifies assumptions, another seeks alternatives and a decision owner closes the discussion.

SI can act as a challenger or note-taker, but it should not become an invisible source of authority simply because its wording sounds neutral.

Record disagreements that remain unresolved rather than editing them into artificial consensus.

Critical Thinking in Long SI Conversations

Long conversations accumulate earlier claims and frames. Periodically reset the current critical-thinking state: active claim, accepted evidence, rejected assumptions, open questions and next check.

Mark superseded claims explicitly. A later correction should not coexist ambiguously with an earlier error.

This connects critical thinking with the state-management discipline from Article 20.

Critical Thinking and Tool Use

Different tools answer different parts of a critical-thinking problem. Search finds current evidence. Code reproduces calculations. Files provide source text. SI organises interpretation.

Do not ask one tool to substitute for another. A generated claim about a calculation is weaker than the actual calculation; a generated description of a policy is weaker than the current official record.

Tool routing is part of epistemic discipline.

Critical Thinking and Model Disagreement

Two AI systems may produce different answers. Disagreement between models is not evidence that one must be correct.

Compare their sources, assumptions and interpretation. If both rely on the same unsupported premise, apparent diversity does not create independent confirmation.

Use disagreement as a signal to inspect the evidence path.

Critical Thinking and Consensus Between Models

Agreement between several models can still be wrong when they share similar training patterns, sources or prompts.

Independent verification means different evidence or methods, not simply more generated votes.

For factual claims, return to source. For calculation, reproduce. For code, test. For decisions, inspect criteria and values.

Critical Thinking and Human Expertise

A qualified expert can identify assumptions that generic reasoning misses. Their judgment is especially important when the domain contains tacit knowledge, regulation or high consequence.

Prepare expert review efficiently: claim, evidence, assumptions, unresolved question and what decision depends on it.

SI can improve the briefing, but the expert’s authority should be represented accurately rather than inflated or ignored.

A Critical-Thinking Handoff

For important analysis, create a handoff containing Claim, Evidence, Assumptions, Alternatives, Unknowns, Bounded Conclusion and Next Check.

Another authorised person should be able to inspect the reasoning without reading the entire conversation.

This handoff turns critical thinking into shared, reviewable work rather than private intuition.

A Critical-Thinking Regression Set

• One ordinary factual claim.
• One causal claim.
• One percentage with a hidden denominator.
• One source outside the relevant population.
• One leading question.
• One false binary.
• One claim with genuinely mixed evidence.
• One value judgment presented as factual necessity.

Use the set to practise or evaluate a recurring SI-assisted analysis workflow.

A Final Critical-Thinking Portability Test

Take the framework into a new domain and use a different tool or no SI for the first pass. The core questions—claim, evidence, assumptions, alternatives, scope and uncertainty—should still apply.

Then use SI for critique. Compare what it found with your own analysis. Any gap becomes a learning target.

The framework is portable when it improves independent judgment rather than requiring one particular prompt or model.

A Final Critical-Thinking Governance Gate

Before an important conclusion becomes action, confirm who owns the decision, what evidence standard applies, which uncertainty remains and what would trigger review.

If the conclusion affects other people, check whether the receiver could reasonably interpret the evidence more strongly than intended.

Then preserve the analysis in a form that can be revisited if new evidence appears. Critical thinking is not a one-time performance; it is a method for keeping conclusions proportionate to evidence over time.

The Final Evidence-to-Action Check

Before acting on a conclusion, write one sentence for each of four layers: what the evidence directly shows, what you infer from it, what uncertainty remains and what action is justified despite that uncertainty.

If the action statement is stronger than the inference, or the inference is stronger than the evidence, step back. The gap between layers is where overclaiming usually enters.

Then state the review trigger: what new evidence, outcome or changed condition would cause the decision to be reconsidered? This keeps critical thinking alive after the initial analysis rather than treating one conclusion as permanent.

The strongest SI-assisted critical thinking therefore ends with proportionate action, explicit uncertainty and a path for revision when reality changes.

A Final Critical-Thinking Transfer Exercise

Take one claim from education, one from business and one from technology. For each, write the claim precisely, identify its strongest evidence, expose one assumption, generate one alternative explanation and state the narrowest justified conclusion.

Then compare your analysis across the three domains. The surface facts should change, but the critical-thinking structure should remain stable. Where the method breaks, identify which domain-specific expertise is missing rather than pretending the generic framework is enough.

This final transfer exercise protects the purpose of the article: a reusable reasoning method that strengthens independent judgment across different kinds of problems.

Critical Thinking as a Repeatable Operating Habit

The final goal is not to run a large framework every time. With practice, the core checks become compact: what is the claim, what is the evidence, what is assumed, what else could explain it and what level of confidence is justified?

For important work, make those checks visible in the output. For ordinary low-stakes reasoning, the habit can remain lightweight. The skill is proportionality: enough scrutiny to match the consequence without turning every small question into a research project.

This is the durable value of SI-assisted critical thinking—faster access to counterarguments and structure while the user becomes increasingly capable of running the reasoning process independently.

The Final Independent-Judgment Check

Before accepting an SI-assisted conclusion, write your own one-sentence judgment without copying the generated wording. State what the evidence supports and what it does not support.

If you cannot explain the conclusion independently, return to the evidence map. Critical thinking is complete only when the user can own the reasoning rather than merely recognise a persuasive answer.