SSRN

The Generative AI Learning Penalty: Evidence from Chinese Secondary Education by David Strömberg, Victor Lei, and Yanhui Wu. Generative AI promises productivity gains but what happens when students use it for schoolwork? Using an extraordinary 30‑month panel of 26,811 students in grades 7–12, this paper provides one of the clearest empirical answers to date. The authors combine monthly closed‑book exams, high‑stakes entrance‑exam results, homework scores, and completion times across nine subjects, exploiting staggered AI adoption in a difference‑in‑differences design.

The pattern is stark. AI adoption raises homework scores by 18% and cuts completion time by 30%, yet monthly exam scores fall by 20% within six months. High‑stakes entrance‑exam scores drop by 18–24%, with the full penalty emerging only after about two years. Losses are largest in social sciences, followed by STEM and languages, and are especially pronounced for junior students, high achievers, and boys.

Crucially, the learning penalty is concentrated among the ~80% of AI users who appear to outsource homework, evidenced by unusually short completion times paired with high homework scores. Students who maintain normal homework effort experience only small losses.

Read: http://spkl.io/6183716SJ

work is expanding & often leaves workers vulnerable. This article argues that ’s Charter section 2(d) should provide platform gig workers with a constitutional right to bargain collectively.

Read: http://spkl.io/61857EMBP
Subscribe: http://spkl.io/61887EMBr

Peter A. Allard School of Law