Exploring the potential of AI assistance for teamwork in programming education

Abstract

This study investigated the impact of using generative AI (GenAI) for programming in a collaborative setting in higher education. The study involved 44 Computer Science students, working on their final bachelors projects, to explore how these tools can be beneficial, what challenges they may present, and how they impact the collaboration dynamics. The results of the quantitative data, obtained by Likert scale questions, showed that students who made extensive use of GenAI spend significantly less time on explaining their code to their peers, had more ease with producing code, were more confident about their capabilities to produce code, and believed that the workload was distributed more equally. A more in-depth and contextual view of using GenAI in collaborative programming is extracted from the qualitative data, gathered through observational sessions, semi-structured interviews, and open-ended questions. This data showed improvements in the efficiency and autonomy of the project by reducing students' dependency on team members for simple questions. Which allowed students to have more time to discuss complex questions. However, this efficiency is limited by a legitimate lack of trust in the output of GenAI tools. Moreover, several challenges need to be addressed to successfully incorporate GenAI into a collaborative programming setting, such as over-reliance on the GenAI tools and the risk of individualization.