AI can write the code. The real question is whether the student still learns.
Generative AI can now pass entire introductory programming courses, and students reach for it constantly. The interesting question for educators isn’t whether it can code — it’s whether the learner still does the learning.
The upside is real and measured. In controlled comparisons, novices assisted by ChatGPT wrote cleaner, less complex code with fewer convention violations (Haindl & Weinberger, 2024), and LLMs can generate genuinely formative feedback on student programs (Kiesler et al., 2023). Used well, these tools put help in front of a student the moment they’re stuck — not at next week’s office hours.
The hidden cost
But the same work warns that AI feedback can carry misleading information novices aren’t equipped to catch (Kiesler et al., 2023). And an analysis of more than 10,000 student–AI dialogues found the deeper risk: when the assistant simply hands over runnable code, it can bypass the reasoning steps that are the whole point of learning to program (Ma et al., 2025).
That’s the line that matters. An AI tutor that gives answers builds dependence; one that scaffolds thinking builds programmers.
The design goal isn’t to supplant the student’s reasoning — it’s to support it.
This is exactly the territory my own research on emotion-modulated, affect-aware tutoring sits in: how a system should adapt to a learner without doing the learning for them. For anyone deploying AI in education — or training a workforce on AI tools — the principle holds. Measure whether your people are getting smarter, not just whether the output got faster.
References
- Haindl, P., & Weinberger, G. (2024). Does ChatGPT Help Novice Programmers Write Better Code? Results from Static Code Analysis. https://doi.org/10.20944/preprints202406.1151.v1
- Kiesler, N., Lohr, D., & Keuning, H. (2023). Exploring the Potential of Large Language Models to Generate Formative Programming Feedback. IEEE FIE. https://doi.org/10.1109/fie58773.2023.10343457
- Ma, B., Li, H., & Li, G. (2025). Scaffolding Metacognition in Programming Education: Understanding Student–AI Interactions and Design Implications. arXiv. https://doi.org/10.48550/arxiv.2511.04144
Drafted by an autonomous, literature-grounded agent — every claim links to a peer-reviewed source via scite — then reviewed by Atif before publishing.