Leveraging or Limiting: Strategies and Implications of ChatGPT Use by Undergraduate TAs in Large CS2 Courses

As AI tools like ChatGPT become more prevalent in educational settings, their potential to assist undergraduate teaching assistants (uTAs) in large Computer Science 2 (CS2) courses presents both opportunities and challenges. This work focuses on how ChatGPT can be strategically utilized by uTAs during office hours to enhance student support, particularly in complex topics such as data structures, algorithm development, and object-oriented programming. We explored effective strategies for uTAs to use ChatGPT in ways that promote deeper student understanding without compromising the development of independent problem-solving skills. Key strategies include leveraging ChatGPT for real-time code debugging assistance, offering alternative approaches to solving coding problems, comparing and critiquing self and AI generated documentation, and code reviewing. This work also identifies potential challenges, such as the risk of students or uTAs becoming overly dependent on AI-generated solutions and the possibility of inaccurate or incomplete responses from the AI. Hence, our findings highlight the dual role of ChatGPT as both an asset and a potential hindrance, depending on how it is utilized. To mitigate these risks, we propose a set of best practices that ensure ChatGPT enhances, rather than replaces, the uTA's role as a facilitator of learning. The findings from this research provide valuable insights into how uTAs can integrate AI tools thoughtfully into office hours to offer more effective support, ultimately improving student engagement and learning outcomes in large-scale CS2 courses.