Building AI literacy into a science course
AI reached classrooms faster than the habits students need to use it well. So I built the habits into the course itself: purpose-built AI partners, a documentation protocol, and one requirement that does most of the work.
The problem
Most schools met AI with a ban or a shrug, and neither teaches a student anything. I wanted students to treat AI the way a scientist treats any source: openly, critically, and with their own judgment intact. That meant building the literacy into the course, not bolting a policy onto the side of it.
A familiar discipline
None of this is entirely new. For years we taught students to treat Wikipedia the same way: a good place to start research, never a source to cite in a final reference list. The discipline was the point. AI gets the same treatment, with documentation added. Use it to think, evaluate it like any other source, and never mistake fluency for accuracy. Naming AI as a source that is usually right but never guaranteed is what conditions a student to keep checking.
The approach: AI partners, not a chatbot
Instead of pointing students at a general chatbot, I built a set of AI partners, each scoped to a single job. A partner with a defined role drifts less and teaches more than an open prompt, and every partner output is still a source the student has to evaluate.
The three rules
The documentation protocol
Every AI output is treated as a source, and the same evaluation framework applies. Students complete a short documentation template with every deliverable: the tools used, representative prompts, what the AI contributed, what they contributed, and an evaluation of the AI itself. Honest disclosure carries no penalty; hidden use is an integrity issue. Students also learn the failure modes to watch for: hallucinated citations, confident wrongness, training-cutoff gaps, sycophancy, and domain weakness.
Every student must name one specific thing the AI got wrong. It is required. If you cannot find one, you were not checking carefully enough. That single requirement turns passive users into critical ones.
Try it: spot what the AI got wrong
Here is a confident AI answer about a real earthquake. One specific fact is wrong. Finding it is the move every student makes, every time. Click the claim you would check.
The 1964 Great Alaska earthquake struck near Anchorage on 27 March 1964. It registered a moment magnitude of 9.0, making it the most powerful earthquake in North American history. The shaking lasted about four and a half minutes and triggered a tsunami that reached the coast of California.
That is the whole skill: treat the AI as a source, verify the confident specifics, and document what it got wrong.
Could I explain to my teacher exactly what I did, and would I be comfortable doing so?
Why it travels
The framework is subject-agnostic. I developed it in Earth and Environmental Science, but the partners, the protocol, and the documentation habit transfer to any course, and I carry them forward into new material. In the end it is less about a tool and more about a stance: AI as a source to be evaluated, never a substitute for judgment.