The Benefits and Harms of Generative AI for Novice Programmers [video]
This video explores how generative AI tools like ChatGPT can both help and hinder novice programmers, highlighting benefits such as accelerated learning and code assistance alongside risks like over-reliance, reduced problem-solving skills, and difficulties in verifying AI-generated code.
Background
- This is a talk by James Prather, a prominent computing education researcher who studies how novice programmers interact with AI coding tools like GitHub Copilot and ChatGPT.
- Generative AI for coding refers to large language models that can write, explain, or debug code in response to natural-language prompts — a major shift from traditional "write code from scratch" learning.
- Key prior context: Prather's research group published widely cited papers (e.g., "The Robot in the Classroom") showing that while AI tools help novices be more productive, they can also create over-reliance (students copy-paste without understanding) and "illusions of competence" (learners think they understand code the AI wrote for them, but don't).
- The tension: the same tools that lower barriers for beginners also risk short-circuiting the deliberate practice needed to build fundamental programming skills. This debate is central to how introductory computer science courses are being redesigned right now.