LLM-assisted self-study yields higher post-intervention exam performance than traditional resources among early-career knowledge workers¶

Claim · All claims
Evidence · 1 study · causal r2 · q3
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q3 i? Participants randomized to LLM-assisted self-study scored significantly higher on the post-intervention exam than those using traditional resources, controlling for baseline score, GPA, gender, and study preferences. → Bharat Anand 2026

Evidence¶

Bharat Anand 2026¶

Bharat Anand, Lihi Idan. (2026). Generative AI and the Productivity Divide: Human–AI Complementarities in Education and Knowledge Work (Working Paper). https://arxiv.org/abs/2605.18143 doi:10.48550/arxiv.2605.18143

q3 · i? · causal · r2

Randomized controlled experiment at Texas A&M University with 179 participants, mostly engineering undergraduates, randomized to LLM or baseline self-study for three days. The OLS treatment coefficient remained positive and statistically significant (p<0.05) with controls; descriptive means show "M=.56,SD=.26, compared toM=.48,SD=.23".

"The average post-intervention exam score among LLM participants wasM=.56,SD=.26, compared toM=.48,SD=.23 in the baseline group."

Discussion¶

Related Claims¶

• AI Interaction Competence, not GPA or prior knowledge, is the decisive moderator of LLM-assisted learning gains — related
• GPA and prior topic knowledge do not moderate the performance effect of LLM-assisted self-study — related
• Denial of LLM access increases attrition and LLMs are the most preferred study resource, acting as revealed preference — related
• In one randomized working-paper experiment, conceptual-map scaffolding gave LLM-using novices marginally higher scores (p < .10) and lower score variance, with a marginal benefit concentrated among low-AIC novices — related
• 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
• Unguarded answer-giving AI harmed unaided exam performance while a guarded version of the same model erased the harm (Bastani et al., 2025, as reported) — reports the opposite
• LLM assistance can enhance employees' creativity, especially for those with skills to reflect on AI use — related
• Intervention students reported significantly higher exam confidence than controls (M=4.3 vs 3.4 on a 5-point scale, d=0.76) — related
• AI access raises immediate unaided test scores by 7.2 percentage points (0.28 SD) after a timed learning phase — related
• Persistence and performance costs concentrate among participants who used the AI to obtain direct solutions, while hint-users resemble controls — related