Preference for LLMs aligns only weakly with AI Interaction Competence, producing enthusiastic-but-unskilled and skilled-but-reluctant subgroups¶

Claim · All claims
Evidence · 1 study · causal r2 · q3 · i2 medium
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q3 i2 LLM-preferring participants exhibited higher AIC levels on average, but the preference–ability correlation was only small to moderate (r= 0.26). → 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 · i2 · causal · r2

Preference–ability comparison within the experiment, with a printed correlation of "r= 0.26" between preference and ability. The article identifies Group 1 (enthusiastic but unskilled) and Group 2 (skilled but reluctant, often attending optional lectures); a logistic regression with social-learning preference predicting LLM preference was positive but non-significant (p≈0.1).

"A Mann–Whitney U test revealed a statistically significant difference (p<.05), with a moderate effect size: participants who preferred LLMs generally exhibited higher AIC levels. However, the correlation between preference and ability was only small to moderate (r= 0.26)"

Discussion¶

Related Claims¶

• AI Interaction Competence, not GPA or prior knowledge, is the decisive moderator of LLM-assisted learning gains — related
• Denial of LLM access increases attrition and LLMs are the most preferred study resource, acting as revealed preference — related
• Most participants expressed reluctance to adopt an eco-feedback LLM for regular use, with nearly two-thirds reporting some degree of unwillingness — related