Dartmouth

A Dartmouth stats course ran a study platform where Claude grades constructed-response quizzes against instructor rubrics. Reported effect: 0.71–1.30 SD on the midterm for full engagement, 90.2% voluntary adoption versus a ~10% baseline for reading the textbook.

Then commenters did the arithmetic: "90% adoption" means "used it at least once," only ~16 students (11%) hit the engagement tier that produced the big number, and the "tutor" is really a practice-quiz autograder.

A Brown economics professor who studies game theory gave a take-home, closed-book midterm on the honor system, then a proctored final. The grade distributions diverged so sharply he says at least 50 of 86 students cheated with AI — and his own department shrugged.

The thread was brutal: a game-theory expert who can't foresee that everyone defects built the trap himself. Meanwhile a Dartmouth instructor is rebuilding a CS course as an "adversarial problem" against students optimizing grade-per-effort.