Universal Learning Diagnostics (ULD) can be run with the help of AI and large language models (LLMs) such as ChatGPT, Gemini, or similar systems. When used correctly, AI can assist with guided questioning, controlled variation, and structured retesting.
However, AI is not the diagnostic authority.
This page explains how to run ULD safely with AI, what AI can and cannot do, and the constraints users must understand to avoid misdiagnosis.
Who Universal Learning Diagnostics (ULD) Is For
ULD is designed for humans who care about reliable learning outcomes, especially when progress has stalled or performance is inconsistent. It applies across age, sector, and context.
Students and Learners
ULD is for:
- school students who study hard but underperform in exams
- learners who freeze under time pressure despite understanding
- students who struggle when questions are rephrased or unfamiliar
- individuals experiencing repeated learning plateaus
ULD helps identify why learning breaks, not just where grades fall.
Parents and Caregivers
ULD is for:
- parents seeking clarity beyond grades and effort
- families trying to understand inconsistent performance
- caregivers deciding when intervention is actually needed
- those who want diagnosis before committing to tutoring or remediation
ULD provides a structured way to understand learning bottlenecks without blame.
Teachers, Tutors, and Educators
ULD is for:
- educators diagnosing why teaching “doesn’t stick”
- tutors working with students who regress or plateau
- teachers distinguishing misunderstanding from pressure-related collapse
- instructors seeking evidence-based intervention targeting
ULD supports diagnosis without replacing pedagogy or professional judgement.
Educational Institutions and Schools
ULD is for:
- schools investigating systemic learning breakdowns
- institutions evaluating training effectiveness
- academic leaders seeking consistent diagnostic standards
- programmes aiming to reduce silent failure and hidden plateaus
ULD provides a universal diagnostic layer independent of syllabus or curriculum.
Adult Learners and Professionals
ULD is for:
- adults experiencing skill plateaus in work or training
- professionals who perform well privately but fail in high-stakes settings
- individuals reskilling or transitioning careers
- lifelong learners seeking stable capability, not just credentials
ULD reframes adult learning failure as diagnosable and recoverable.
Training and Learning & Development (L&D) Teams
ULD is for:
- corporate training programmes with poor transfer outcomes
- L&D teams evaluating why training doesn’t show up in performance
- organisations diagnosing capability gaps without overtraining
- teams seeking recovery rather than repeated retraining
ULD helps isolate whether training fails due to Depth, Load, or Transfer.
Coaches and Performance Specialists
ULD is for:
- academic, executive, or performance coaches
- practitioners working with pressure, execution, and consistency
- specialists diagnosing performance collapse under constraint
- coaches needing a structured diagnostic framework
ULD supports coaching with measurable diagnostics and verification.
Researchers and System Designers
ULD is for:
- researchers studying learning failure modes
- system designers building education or training platforms
- analysts evaluating learning stability across populations
- designers of diagnostic or assessment frameworks
ULD offers a first-principles capability model that is falsifiable and stable.
AI-Assisted Self-Learners (With Boundaries)
ULD is for:
- disciplined self-learners using AI tools responsibly
- users seeking early diagnostic insight before guided use
- individuals willing to retest and verify results
ULD is not for casual self-labelling or one-shot judgement.
Who ULD Is Not For
ULD is not designed for:
- intelligence ranking or labelling
- personality assessment
- one-off testing without verification
- justification for exclusion or judgement
- replacing teaching, mentoring, or human decision-making
ULD diagnoses capability states, not identity or potential.
Final Scope Statement
Universal Learning Diagnostics is for anyone who wants to answer one question accurately:
“Why is learning failing here — and what exactly needs to be repaired?”
When that question matters, ULD applies.
What AI Can Do in ULD
AI can act as a diagnostic assistant, not a judge.
Used correctly, AI can help by:
- generating structured diagnostic questions
- removing prompts and scaffolding gradually
- varying formats and contexts for Transfer testing
- simulating time or constraint conditions
- tracking responses across retests
AI is useful for process control, not for deciding outcomes.
What AI Cannot Do in ULD
AI cannot:
- verify truthfulness of user input
- detect emotional or physiological states reliably
- replace real-world pressure or stakes
- guarantee accuracy if inputs are incomplete or misleading
- diagnose capability without honest responses
ULD results generated through AI are only as reliable as the human inputs and testing discipline.
Step-by-Step: Running ULD With AI
Step 1: Define the Task Clearly
Before using AI, specify:
- the exact skill or task
- the format (written, spoken, problem-solving, etc.)
- the environment (timed, untimed, private, public)
Vague tasks produce unreliable diagnostics.
Step 2: Run Depth Diagnostics First
Ask the AI to:
- request explanations in the user’s own words
- remove examples and templates
- ask for reconstruction from first principles
- probe error reasoning
If the user cannot explain or rebuild calmly, Depth is the bottleneck.
Depth:
https://edukatesg.com/uld-depth/
Step 3: Introduce Controlled Load
Ask the AI to:
- impose time limits
- increase step density
- reduce pauses or retries
- simulate mild performance pressure
If performance collapses only under constraint, Load is the bottleneck.
Load:
https://edukatesg.com/uld-load/
Step 4: Test Transfer Through Variation
Ask the AI to:
- reframe the task
- change context or format
- remove surface cues
- apply the same structure elsewhere
If performance collapses under novelty, Transfer is the bottleneck.
Transfer:
https://edukatesg.com/uld-transfer/
Verification Is Mandatory
AI-assisted ULD must always include:
- multiple retests
- simplified and constrained versions
- varied task formats
One-shot diagnostics are invalid.
Scoring and verification:
https://edukatesg.com/uld-scoring/
Critical Constraints When Using AI
Constraint 1: Time Is Artificial
AI-simulated time pressure is not the same as real-world stakes. True Load testing may require:
- live environments
- real deadlines
- human observation
AI can approximate, not replace, reality.
Constraint 2: Accuracy Depends on Honesty
AI cannot detect whether:
- effort is genuine
- answers are copied
- difficulty is understated
- failures are hidden
Dishonest inputs produce unreliable diagnostics.
Constraint 3: AI May Over-Explain
AI tends to:
- scaffold excessively
- hint unintentionally
- correct too early
Users must explicitly instruct AI to:
- withhold hints
- delay feedback
- avoid teaching during diagnosis
Otherwise, Depth and Transfer signals become distorted.
Constraint 4: AI Can Hallucinate Authority
AI may appear confident even when uncertain.
ULD users must treat AI outputs as hypotheses, not conclusions. Diagnosis is confirmed only through:
- repeated testing
- consistent results
- verification under constraint
What AI-Assisted ULD Is Best Used For
AI-assisted ULD is most suitable for:
- early self-diagnosis
- low-stakes exploration
- identifying likely bottlenecks
- preparing for guided diagnosis
High-stakes decisions should involve guided or supervised use.
Usage boundaries:
https://edukatesg.com/uld-usage-boundaries/
What AI-Assisted ULD Should Not Be Used For
AI-assisted ULD should not be used as:
- a final judgement of ability
- a basis for exclusion or ranking
- a replacement for real performance testing
- an identity or intelligence label
Misuse increases error and reduces trust.
Responsible Use and Human Abuse Risk
AI does not prevent misuse.
ULD can be abused if a human:
- selectively inputs data
- ignores retest requirements
- uses AI outputs to justify preconceived conclusions
- labels themselves or others based on partial diagnostics
Such abuse reflects human misuse, not system failure.
ULD explicitly requires verification, repetition, and separation of diagnosis from identity to reduce this risk.
When to Escalate Beyond AI
Escalate to guided or real-world testing when:
- stakes are high
- results are inconsistent
- multiple axes appear weak
- recovery attempts fail
AI is a tool—not the final authority.
Where to Go Next
If you are new to ULD:
https://edukatesg.com/why-smart-humans-get-stuck/
If you want the full diagnostic system:
https://edukatesg.com/uld/
If you want the recovery modes:
https://edukatesg.com/uld-recovery-modes/
Final Statement
AI can assist ULD.
It cannot replace discipline, honesty, or verification.
Universal Learning Diagnostics remains a human-centred diagnostic system, whether run with AI or without it.