LLMs and Computing Education

Project description

The rapid adoption of large language models in computing education has outpaced the research evidence. Students routinely use tools like Copilot and ChatGPT to write, debug, and explain code, yet we have limited understanding of how this shapes learning and self-regulation. Projects in this space investigate how LLM tools can be designed to support genuine learning rather than bypass it, and suit students with interests in HCI or CS education research. Possible areas for exploration include:

• Reflection prompting alongside AI hints: students dislike being asked to reflect but it improves outcomes; how do you design prompts that preserve both?
• Self-regulation scaffolding around LLM use: AI feedback tends to skip metacognitive components; build and evaluate a tool that prompts students to check their own understanding before and after receiving help.
• Instructor-moderated vs. fully automated LLM feedback: direct LLM answers risk undermining learning, but instructor-in-the-loop systems are currently understudied.
• LLM feedback calibration across ability levels: evidence from writing suggests LLM feedback helps lower-ability students most - does this same observation hold for programming?

Note: You need to register interest in projects from different supervisors (not a number of projects with the one supervisor).
You must also contact each supervisor directly to discuss both the project details and your suitability to undertake the project.