Laura Graf, Ramona Beinstingel, Stephan Kusche, Oleksandra Poquet (2026) — arXiv:2606.23237 (cs.HC; cs.CY) 📄 Full text (arXiv)
Graf et al. (2026) identify a new challenge in collaborative programming education: AI use is now an invisible yet consequential dimension of collaboration, and partners often misread ability and effort from code. In a three-wave longitudinal study of 103 student pairs in an introductory software engineering course, they found that greater misalignment between partners' beliefs about each other's AI use early in the project was associated with lower final project scores. This effect was strongest in teams with lower prior programming performance, suggesting low-performing students pay a higher cost of misaligned perceptions. Notably, perception misalignment did not consistently decrease through face-to-face pair-programming sessions, implying that transparency mechanisms (disclosures, shared logs) may be needed. This work connects student-experience research in cs-education with collaborative-ai-tutoring and raises important questions about over-reliance and academic-integrity in AI-augmented collaborative learning.
Related Pages
- student-perception-ai-use-collaboration -- This page
- cross-cultural-student-perceptions-genai-computing — Did Alice Do Wrong? Cross-Cultural Differences in Student Perceptions of Generat