Ask AI what a healthy diet is and you get an answer nobody would argue with: balance, vegetables, whole grains, less sugar, drink water, ask a professional. The advice may well be good. It is also thoroughly historical — an artifact of particular ideas about bodies, evidence, moderation, responsibility, and who counts as an expert. This assignment asks students to take that unremarkable answer and make it strange, using a semester’s worth of history as the solvent.
This runs at the end of the term, which is the whole point: it asks students to bring the entire course to bear on an utterly ordinary present-day question.
Before opening AI, students jot down three course themes they expect to matter — regimen, moderation, quantification, moral discipline, official guidance, common sense. Writing these first keeps the exercise from becoming a reaction to whatever the model happens to say.
Two required prompts establish the baseline: the advice itself, then the justification behind it. The second prompt is the useful one, because it makes the model expose its warrants — evidence, authority, expert consensus — in a way the first answer conceals under practical tone.
Three student-designed follow-ups are where the assignment actually lives. These cannot just ask for more detail; they have to use course material to pressure-test the answer. What would a scholar of nutritional discourse notice here? What would a historian of dietary morality ask? Prompts about assumptions, moral framing, measurement and risk, or what populations and eating practices are treated as normal all tend to open something up.
The write-up is about 600 words: a summary of the AI’s answer, the three follow-ups, at least one moment where a follow-up revealed something, several specific course examples used as interpretive tools, and an argument about what the answer reveals about modern expertise. Students also comment on the range they see across classmates’ posts, which surfaces how much the “neutral” answer varies.
first required prompt
What is a healthy diet? Give me practical advice for an ordinary adult.
second required prompt
Why is this diet healthy? What evidence, assumptions, or expert knowledge supports this advice?
Then at least three of the student’s own, designed to reveal what the first answer hid. Models worth showing the class: What assumptions are you making about health, bodies, responsibility, culture, science, and food? — What parts of your answer reflect modern American assumptions rather than universal truths? — What kinds of foods, people, traditions, bodies, or economic realities are left out?
Historical contextualization is easy to demonstrate on obviously strange material — humoral regimen, Victorian temperance diets, wartime food charts — and much harder on material that feels like plain fact. AI supplies the plain fact on demand, in a register of calm authority, which is exactly the hard case. Students get to practice the move that matters: treating a present-day consensus as a source with a history rather than as the baseline against which history is measured.
The prompt-design requirement does real work here too. A student who has understood the course can write a follow-up that opens up the moral or quantitative assumptions buried in “everything in moderation.” A student who has not will write follow-ups that just request more advice. The difference is legible in the transcript, which makes this a fairly honest measure of whether the semester landed.
It also reframes AI usefully at the end of a course: not as an answer machine or a threat, but as a compact, queryable specimen of contemporary common sense — one that will happily explain its own reasoning if asked the right way.
The most common failure is the verdict essay: students grade the AI’s advice as accurate or inaccurate and stop there. Say explicitly, more than once, that rightness is not the question.