Generative AI Tools in an Undergraduate Computer Science Program

We surveyed all undergraduate students (N = 121) enrolled in a three-year Computer Science (CS) program in France to learn about their use of generative AI tools.Our findings reveal widespread use of these tools across all academic years, with a higher proportion of lateryear students reporting using these tools for a wider range of coding tasks compared to firstyear students.As the accessibility of generative Artificial Intelligence (AI) tools increases, the relevance of traditional Computer Science (CS) education is being reevaluated.Generative AI tools built on large language models like ChatGPT and Github Copilot can not only "read" and "write" code in multiple programming languages, but also provide detailed explanations for each piece of code based on natural language queries.Given these advances, the CS education research community is contemplating what CS education should look like now (e.g., Becker et al., 2023).Without concurrence on a clear path for integrating generative AI into CS education, studies highlight how generative AI tools might support learning to code-alleviating programmer's writer's block, explaining algorithmic concepts, clarifying error messages; studies showcase challenges including over-reliance, ethical concerns, harmful bias, sustainability, and hallucinations (Becker et al., 2023).There is emerging evidence on roles for these tools to support learning to code (e.g., Liu et al., 2024;Jacques et al., 2023;Wang et al., 2024).Few studies investigate how students integrate these tools in their learning practices.Insights into how students at various learning stages use these tools can inform novel, participatory design studies (Gomez et al., 2018) focusing on specific needs of students and teachers.Such insights could even contribute to envisioning a generative AI-supported CS undergraduate curriculum (Becker et al., 2023).This paper explores uses of these tools by CS students across three academic years.We surveyed first to third-year undergraduate CS students at a French university to address the following questions: (1) Which generative AI tools are favored by CS students in different academic years?(2) For what coding tasks are students using these tools across academic years?