Abstract

Generative AI (GenAI) is rapidly reshaping learning, raising questions about how students’ collaboration with GenAI affects their self-efficacy and self-perceptions of expertise and performance. This study distinguishes between two collaboration patterns: collaborating with GenAI as a sounding board to refine one’s own work, or relying on it as a ghostwriter to produce content. Drawing on survey evidence from university students, the findings show that sounding board collaborations are positively associated with self-efficacy, as students remain active in shaping their ideas and attribute progress to their own abilities. Ghostwriter ones, on the other hand, foster an overestimation of academic self-perceptions, giving students a false sense of expertise and higher performance, consistent with the Dunning-Kruger effect. Mixed-use patterns suggest that combining collaboration modalities can partly offset but not eliminate risks associated with ghostwriting. The study highlights the importance of student-AI collaboration patterns in learning and offers insights for responsible GenAI integration that sustains effective collaborations.

Download

Share

COinS