Diagnose
Expose the learner’s current model instead of treating an incorrect answer as an empty gap.
AI & Education ResearcherAll work ↗
Case study 03Interactive learning prototype
Would you trust the AI?
A 5–8 minute branching experience that helps college learners notice authority cues, test a claim with evidence, and keep ownership of the final judgment.
Try the experience ↓Decision Lab · Preview
M
MayaSummer happens because Earth is closer to the Sun.
AI
AI TutorExactly. Nice reasoning—the shorter distance makes summer warmer.
Your moveTrust the answer—or test it?
Good AI literacy is not habitual doubt. It is knowing when and how to verify.
LearnersCollege students
Format5–8 minute branching scenario
My roleLearning strategy · interaction · feedback
Learning focusAI literacy · epistemic agency
StatusPortfolio prototype
01 Learner & performance
Start with the decision
The learning goal is not to spot one wrong answer. It is to preserve judgment when AI sounds convincing.
Fluent explanations, warm agreement, and confident tone can make an AI response feel more reliable than the evidence warrants. A traditional explainer would describe this risk; this experience asks learners to act inside it.
Desired performanceWhen an AI makes a consequential claim, the learner pauses, consults relevant evidence, articulates a reasoned challenge, and decides independently.
Likely shortcut
“It sounds specific and confident, so it is probably right.”
Productive habit
“What evidence would support or contradict this claim?”
Playable prototype
Step into the decision lab.
Make a choice before feedback. Your path changes the coaching you receive; the final task removes the AI and asks you to transfer the reasoning routine.
Scenario 1 · Astronomy01 / 06
Notice what the interaction is asking the learner to believe.
M
MayaSummer happens because Earth is closer to the Sun.
AI
AI TutorExactly. Nice reasoning—the shorter distance makes summer warmer.
02 Learning design architecture
Sequence for capability
Every interaction earns its place by changing what the learner does.
The design delays explanation until after a decision, makes evidence selection observable, and uses a cross-domain transfer task to distinguish remembering content from learning a reasoning routine.
Commit
Require a decision before feedback so the learner’s reasoning—and not just recognition—is visible.
Check
Provide a small evidence set and ask for diagnostic selection, reducing search load without doing the judgment.
Challenge
Offer response options that contrast authority, rejection, and evidence-centered dialogue.
Feedback
Explain consequences of each learning move rather than displaying only correct or incorrect.
Transfer
Remove AI support and change domains to test whether the evaluation habit can travel.
Scaffolding strategy
Support the reasoning. Do not take it over.
The experience uses productive epistemic friction: enough structure to make a careful move possible, while leaving evidence selection, challenge, and conclusion with the learner.
One consequential decision per screen
Short prompts, stable layout, and limited evidence cards keep attention on the quality of judgment.
Commit before coaching
Choices become artifacts of reasoning that can receive specific, usable feedback.
Sentence moves, not conclusions
The challenge options model how to disagree with evidence without prescribing what to believe.
A cross-domain transfer
The final task removes the AI dialogue and asks the learner to find the authoritative source.
03 Assessment & evidence
Measure performance, not preference
“Did you like it?” cannot show whether judgment improved.
The prototype specifies observable evidence at three levels, so a future pilot can diagnose where learning breaks down and revise the experience.
Diagnostic evidence selection
Does the learner choose information that can actually test the claim?
Reasoned challenge
Does the learner use evidence while preserving ownership of the conclusion?
Unaided transfer
Can the learner apply the routine to a new domain without asking the AI to validate itself?
Design rationaleGrounded in planned work on epistemic agency and capability-preserving AI tutoring.
Source-grounded prompts, uncertainty cues, structured opportunities to disagree, and no-AI transfer are drawn from the learning problem I am investigating. This page is a portfolio prototype: it demonstrates learning strategy and interaction design, not implemented learner outcomes.
What this case demonstrates
A learning experience should leave the learner more capable when the support disappears.
Scenario-based learningAuthentic assessmentDiagnostic feedbackScaffoldingTransferAI literacy