AI-induced gains are not purely delegation: treated participants do not perform worse once AI is removed, and lower-education participants retain part of their gain, though a 0.200 SD gap re-emerges¶

Claim · All claims
Evidence · 1 study (2 entries) · causal r3 · q3
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q3 i? Prior AI access raises low-education follow-up scores by 0.171 SD (significant at 5%); the high-education effect of 0.071 SD is not significant. → Guillermo Cruces 2026 q3 i? Low-education treated participants still lag high-education counterparts by 0.200 SD on the follow-up (significant at 5%); the difference in treatment effects (−0.100) is not distinguishable from zero (p-value=0.398). → Guillermo Cruces 2026 (2)

Evidence¶

Guillermo Cruces 2026¶

Guillermo Cruces, Diego Fernández Meijide, Sebastian Galiani, Ramiro H. Gálvez, María Lombardi. (2026). Does generative AI narrow education-based productivity gaps? Evidence from a randomized experiment. https://www.socialscienceregistry.org/trials/16607 doi:10.2139/ssrn.6246347

q3 · i? · causal · r3

Non-AI-assisted follow-up module administered immediately after the task to all participants in the randomized experiment. The article reports the low-education follow-up score "rises by 0.171 SD (significant at the 5% level)".

"their follow-up score rises by 0.171 SD (significant at the 5% level). For high-education individuals, the estimated treatment effect is smaller (0.071 SD) and not statistically significant."

Guillermo Cruces 2026 (2)¶

Guillermo Cruces, Diego Fernández Meijide, Sebastian Galiani, Ramiro H. Gálvez, María Lombardi. (2026). Does generative AI narrow education-based productivity gaps? Evidence from a randomized experiment. https://www.socialscienceregistry.org/trials/16607

q3 · i? · causal · r3

Follow-up results section: low-education treated participants "stilllagbehindtheirhigh-educationcounterpartsby0.200 SD", showing underlying human-capital differences persist once the tool is unavailable.

"despite the positive effect of access to AI on the follow-up scores of low- educationparticipants,theystilllagbehindtheirhigh-educationcounterpartsby0.200 SD (significant at the 5% level)"

Discussion¶

Related Claims¶

• Access to a generative-AI assistant raises task performance for both lower- and higher-education adults, with significantly larger gains for lower-education participants — related
• AI assistance closes about three-quarters of the baseline education-based task-performance gap, though a 0.139 SD gap remains among treated participants — related
• Heavy reliance on generative AI can undermine learning: students with GPT-4 access performed better on practice problems but worse on exams once the tool was removed — reports the opposite
• Falsifiable hypothesis: unrestricted AI use on deep-processing tasks will produce a product–process dissociation—higher-rated assignments but lower unaided delayed transfer — reports the opposite
• Students given AI assistance in peer feedback tended to rely on the AI-generated feedback and struggled to perform the task when the assistance was removed — reports the opposite
• Performance improvements from generative AI assistance diminished once the assistance was removed — reports the opposite
• Intensive AI assistance predicts strong task performance even with low engagement, but follow-up performance is substantially higher only when intensive AI use is combined with sustained task engagement — related
• High-school students given GPT-4 access solved more problems in the moment but performed worse than controls once the tool was removed (review attribution) — reports the opposite
• Immediate task success does not establish learning that persists without AI support — related
• Over four months and four sessions, LLM-group participants performed worse than Brain-only participants at neural, linguistic, and scoring levels — reports the opposite
• Meta-analysis beyond education found human-AI integration performed worse than the better performer alone in 58% of studies — reports the opposite