Subject deep dive · Science

Open · the fabrication risk

Do Ask AI to explain a phenomenon and cite studies; show that it can invent confident, wrong specifics and fake citations. Tie back to "how AI fails" from the summer session.

Focus · research with AI without outsourcing judgment

• Where it helps: generating hypotheses to test, designing investigations, explaining phenomena at a chosen reading level, producing sample datasets to analyze, and scaffolding lab-report structure.
• Where it fails: inventing "data," fabricating citations, and giving explanations students accept without checking. In science, an unverified claim is the whole problem.
• The move: AI produces a claim, dataset, or explanation; students test it against evidence, an observation, a real dataset, a source. The verification is the science.

Watch Keep the discipline of evidence central. AI is the hypothesis generator, not the authority.

Hands-on · redesign an inquiry (laptops open)

Each teacher rebuilds one activity so verification is required, choosing one of:

• Test the explanation — students get an AI explanation of a phenomenon and design an observation or experiment that would confirm or refute it.
• Check the data — students receive an AI-generated dataset (which may contain errors) and validate it against a real measurement or known value.
• Vet the sources — students take AI's cited "studies" and verify whether they exist and say what AI claims.

Do Draft the activity in the shared doc, including the disclosure line from September. Pairs swap and sharpen.

Watch One finished, runnable activity per teacher, not a menu of ideas.

Share-out + commit

Do A couple of teachers show their redesign. Commitment: run it this month. Preview: February is about what students are actually doing with AI.