Work

When do students seek AI help? Exploring within-person and between-person differences

Year: 2026
Type: article
Abstract: Research on student AI use typically asks who adopts these tools, treating use as a stable trait. Yet students face a fresh decision at every problem. We examined how much AI use reflects stable individual differences and how much varies within a student across tasks. Undergraduates (N = 398) at two institutions completed 30 quiz trials, choosing on each trial to answer independently, consult an AI chatbot, or rely on an AI-generated answer. Students rated their topic expertise before each decision; cognitive abilities, AI attitudes, and demographics were also measured. Choices were analysed with Bayesian multilevel multinomial logistic regression. With moderately reliable AI, about half the variance in AI use reflected stable individual differences (intraclass correlations of 41% for chatbot use, 48% for full reliance); the remainder varied within individuals across trials. Self-assessed expertise accounted for this within-person variation: students sought help primarily when they felt unknowledgeable. Fluid intelligence predicted more chatbot consultation but less full delegation. Male students and those with more favourable AI attitudes delegated more readily. AI use is better described as adaptive help-seeking than as a fixed trait. Interventions targeting metacognitive calibration may prove more effective than blanket policies assuming uniform use.
Language: English
Subfield: Education
Open Access status: diamond
OpenAlex ID: W7121333742
FWCI:
Cites: 51
Cited by:
Related to: 10