The Future of Education

Peer Collaboration and AI-Interactions: Assessing Group Agency in

Project-Based Programming

Ayfer Alper¹, Gülcan Canbolat², Barbara Bruschi³

¹ Ankara University, Türkiye | [email protected]

² Ankara University, Türkiye | [email protected]

³ University of Turin, Italy | [email protected]

Large Language Models serve as conversational mentors and Generative AI acts as a versatile content creator, offering a transformative potential for education. However, the success of these tools depends on students moving beyond passive consumption to become 'active agents' who exercise responsibility and evaluate learning through a critical filter.

Notably, empirical evidence suggests that student agency is primarily examined individually within AI learning environments. Positioned at the heart of student-centered pedagogy, agency serves as the fundamental element fostering engagement and commitment to the learning process (Starkey, 2019). According to Roe and Perkins (2024), the potential for GenAI to foster student agency is at risk of being replaced by a restrictive automation that diminishes critical depth and individual exploration. In this context, it is crucial to determine whether the student remains a passive user or becomes an active subject directing the learning process. Within Generative AI-mediated environments, targeted measures are essential for monitoring and adjusting prompt strategies, questioning methods, and AI-generated content (Xia et al., 2025). When peer support and group work, such as Project-Based Learning (PBL), are introduced, "Student Agency" evolves from an individual effort into a dynamic balance between "Collective Agency" and "Individual Contribution" within the group.

Integrating agent-focused engagement theory with a PBL framework, this study investigates project development and problem-solving in a Programming course, focusing on the authentic experiences of students across 14 distinct groups. Through the analysis of GenAI interaction recordings and open-ended questions, the research captures students' underlying thought processes and decision-making habits. Accordingly, the study examines the multifaceted nature of student interactions with Generative AI, specifically focusing on engagement levels and the manifestation of agency. By analyzing these dialogues, the research explores how individual agency transitions into collective agency during collaborative problem-solving, revealing how students negotiate, co-create, and exercise responsibility within AI-enhanced environments.

Keywords: Generative AI, Large Language Models (LLMs), Project-Based Learning (PBL), Student Agency, Collective Agency, Programming Education, Agent-focused Engagement.

References

1. Roe, J., & Perkins, M. (2024). Generative AI and agency in education: A critical scoping review and thematic analysis. arXiv preprint arXiv:2411.00631. <-block _nghost-ng-c923084010="">https://doi.org/10.48550/arXiv.2411.00631
2. Starkey, L. (2019). Three Dimensions of student-centered education: A framework for policy and practice. Critical Studies in Education, 60(3), 375–390. <-block _nghost-ng-c923084010="">https://doi.org/10.1080/17508487.2017.1281829
3. Xia, L., et al. (2025). Developing and validating the student learning agency scale in generative AI-supported contexts. Education and Information Technology, 30, 13999–14021. <-block _nghost-ng-c923084010="">https://doi.org/10.1007/s10639-024-13137-5