A transferable teaching pattern · ECON 2316 example
Argue With the AI
The AI defends a claim that sounds right but hides a flaw. The student's job:
use course theory to catch it. Learning happens in the evaluation, not the answer.
The pattern (any discipline): give students an AI that confidently states a defensible-but-wrong
position → they must rebut it with course concepts → they grade the AI's reasoning.
Here it's micro theory; swap the claim for your field.
AI
Perfect price discrimination eliminates deadweight loss — output reaches the competitive level.
So it's
efficient, which means it's the
best outcome for society.
Student
Careful — you conflated
efficiency with
welfare.
Yes, perfect price discrimination removes the deadweight loss, so it's allocatively efficient.
But it does that by transferring
all consumer surplus to the monopolist — consumers are left
with zero. "Efficient" only means the surplus isn't wasted; it says nothing about who gets it or
whether that's
good. A distributional judgment is separate from efficiency.
AI
You're right. I overstated it — efficiency (no DWL) is not the same as desirability. The equity
question is distinct, and I collapsed the two.
AI-THINK
Student evaluates and corrects AI output using theory — the core AI-readiness skill.
AI-DATA
Sees a concrete limit of LLMs: fluent, confident, and still wrong.
AI-ETHICS
Efficiency vs. equity — models can smuggle value judgments into "neutral" claims.
AI-CREATE
The human adds the judgment the AI lacked — the value-add that stays theirs.
Adapt it: History — AI defends a tidy but contested narrative; students cite sources.
Philosophy — AI argues an ethics position; students find the hidden premise.
Policy — AI recommends a "efficient" policy; students surface who bears the cost.
Same pattern, your discipline.