The article also discusses the issue of over-reliance on AI tools versus self-regulation. The study found instances of both, with some students frequently copying task descriptions to generate code without prior manual attempts, while others divided tasks into subgoals and asked the AI to generate only the first subgoal. The authors suggest that future tool designers and educators should promote opportunities for self-regulated use of AI code generators while discouraging unregulated usage.
Key takeaways:
- AI code generators like Codex can significantly improve the learning process for novice programmers, helping them complete tasks faster and with higher correctness scores.
- Despite concerns about over-reliance on AI tools, students using Codex performed similarly or better than those without on code-modifying tasks and retention tests.
- Analysis of AI tool usage revealed signs of both over-reliance and self-regulation among students, indicating the need for careful implementation and guidance in educational settings.
- Future tool designers and educators should encourage self-regulated use of AI code generators while discouraging unregulated usages to maximize learning benefits.