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¶

Claim · All claims
Evidence · 1 study · causal r3 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? Among treated participants, intensive AI assistance predicts strong main-task performance even when task engagement is low, but follow-up performance is substantially higher only when assistance is combined with sustained task engagement. → Guillermo Cruces 2026

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

q2 · i? · causal · r3

Exploratory analysis dividing treated participants into four groups by AI-assistance intensity and task engagement (above/below treatment-group medians on share of task components for which help was requested and time spent on task). The article reports that "follow-up performance is substantially higher when assistance is com- bined with sustained task engagement."

"Third, on when gains carry over, intensive AI assistance predicts strong main-task perfor- mance, but follow-up performance is substantially higher when assistance is com- bined with sustained task engagement."

Discussion¶

Learner Variables¶

• Motivation — moderator: an instructional effect differs with it

Related Claims¶

• AI assistance raises the probability of correctly identifying the task's root cause for both education groups, with near-equalization among treated participants — related
• Learners' task-specific self-efficacy beliefs predict whether they initiate, persist through, and recover from difficulty on academic tasks. — a broader claim this one bears on
• 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 — related
• AI assistance reduces persistence: persistence costs concentrate among learners who used AI for direct solutions, not hints — related
• Learners invest more effort in tasks they perceive as interesting, personally important, or useful to their future goals. — a broader claim this one bears on
• Persistence and performance costs concentrate among participants who used the AI to obtain direct solutions, while hint-users resemble controls — related
• Chat-log analysis indicates lower-education participants obtain substantial assistance from AI while higher-education participants use the tool somewhat more effectively across several margins, explaining why the gap narrows but does not disappear — related
• Immediate task success does not establish learning that persists without AI support — a broader claim this one bears on
• No AI participants spent roughly 1.5 times as long on the assessment as AI-Assisted participants, descriptively reported as a marker of offloading — related
• Hints that instruction is suboptimal trigger compensatory self-regulation, but self-management carries cognitive and motivational costs that deplete over time — related