Null vs active control

AI-personalized computer-assisted learning did no better than non-personalized software for 8,647 rural Chinese primary students.

Stanford REAP and collaborators randomized 315 classes in 105 Sichuan schools to AI-personalized CAL, non-personalized CAL, or non-CAL activities; the AI arm showed no improvement in any ability tercile and may have been worse than plain CAL for the middle tercile.

Sources

primarySSRN: Putting AI to the Test — Evidence from a Large-Scale RCT in Rural Chinapapers.ssrn.com
primaryCESifo working paper versionifo.de
primaryAEA RCT Registry AEARCTR-0007888: Adaptive versus Non-adaptive EdTechsocialscienceregistry.org

Editor's notes

Intervention dates come from the AEA registry (AEARCTR-0007888, PI Loyalka), whose sample matches the paper exactly (105 schools, 315 classes, 8,647 students, three class-level arms): intervention 4 April to 14 June 2021. Window is pre on either reading below. verify: a secondary account of a 2025 talk calls it a year-long trial, so confirm the intervention period against the paper's methods. The SSRN page returned 403 and the CESifo PDF would not extract for the run. verify: the registry was first registered on 28 June 2021, after the intervention ended, so the primary endpoint was not strictly preregistered. Evidence A is arguable; the editor should decide whether it stays.

History

Oct 23, 2026Re-graded window to pre: the AEA registry (AEARCTR-0007888) dates the intervention 4 April to 14 June 2021; added the registry as a primary source.
Oct 22, 2026Added.

Cite this entry

Permalink https://becauseofai.org/entry/china-ai-cal-null

“AI-personalized computer-assisted learning did no better than non-personalized software for 8,647 rural Chinese primary students.” Because of AI, updated Oct 23, 2026. https://becauseofai.org/entry/china-ai-cal-null