Claim
What did AI say?
AI & Education ResearcherAll work ↗
Case study 01Proposed community program
Can nature teach children to question AI?
A family science experience where AI gives the first answer—and nature holds the evidence.
Try the Evidence Lab ↓“The AI gave us an answer. Now we need to find the evidence.”
AudienceChildren + caregivers
Format90-minute family session
Program plan5 public sessions, iterated
SettingThe Arboretum at Penn State
My roleStrategy · curriculum · facilitation
01 The learning problem
Children are learning to ask AI questions. Are they learning to question its answers?
My work with teachers and parents surfaced a timely challenge: even young children are encountering AI systems that answer quickly and confidently. Typical AI-literacy lessons often add more screen time or more rules. Neither necessarily builds judgment.
Nature offers a different learning environment. It makes uncertainty concrete. A photograph can hide a mushroom's underside; two species can look similar; a field guide can conflict with a confident digital answer. Learners have to slow down and ask, How do I know?
Design challengeHow might we help children question AI without teaching AI literacy through even more screen time?
Learning goal
Evaluate an AI-generated claim using observation, evidence, and questioning—rather than accepting it at face value.
Epistemic agencyCalibrated trustEmbodied inquiryFamily sensemaking
02 The design response
A repeatable reasoning routine
Move from a confident claim to an explainable decision.
Every activity follows the same five moves. The routine gives children a portable strategy they can use beyond mushrooms—and beyond AI.
Observe
What do we see?
Check
What else can help?
Decide
Agree, disagree, or need more evidence?
Explain
Why?
The 90-minute experience
Four stations create productive friction between AI confidence and physical evidence.
First-picture trap
AI makes a confident claim from a partial image. Families identify what the image cannot show.
Zoom in
Magnifiers and microscopes reveal texture, pores, gills, and other previously hidden clues.
Context matters
Families compare habitat, scale, season, underside, and surrounding evidence.
Check sources
Field guides and expert-reviewed cards are compared with the original AI claim.
Interactive prototype
Mushroom Evidence Lab
Try one compressed version of the family investigation. This is about evidence—not edible mushroom identification.
AI
AI claim · 94% confident
“This mushroom has gills underneath its cap.”
The image shows only the top of the mushroom.
Observe + check
What evidence should decide?
Decide
Based on the evidence, what do you think?
03 Key design decisions
AI is the object of inquiry, not the delivery platform.
Prepared, expert-reviewed claim cards remove Wi-Fi dependence and let facilitators control scientific accuracy.
Caregivers model how uncertainty sounds.
The no-drop-off format turns verification into a shared routine that can continue at home.
“Need more evidence” is a success state.
Learners practice calibrated trust instead of being rewarded only for immediate certainty.
The experience never teaches edibility.
No collecting or tasting; cultivated specimens, models, and prepared cases ensure a complete experience in any weather.
04 Evidence & iteration
Design status
A rigorous plan, presented without invented outcomes.
This case documents a proposed community program and its design rationale. The curriculum architecture, session flow, material system, safety plan, and iterative facilitation strategy are complete. Learner-impact claims will be added only after implementation.
Design memos
Record intended learning mechanisms, anticipated breakdowns, and why each revision is made.
Facilitator notes
Capture critical incidents: what children notice, what evidence changes a decision, and where families get stuck.
Artifacts + revisions
Compare passports, explanations, station timing, and case revisions across five public sessions.
Low-burden documentation
No video or interviews; any research use would follow Penn State review. The community program can proceed independently.
What this project demonstrates
Learning design is not the delivery of correct answers. It is the careful design of what learners do when an answer might be wrong.
Curriculum strategyInformal science learningAI literacyFamily learningEvaluation planning