A University of Southern Maine (USM) team presented their innovative research at the Consortium for Computing Sciences in Colleges Northeastern Conference (CCSCNE). The presentation, titled “Large Language Models and Introductory Lab Exercises: Susceptibility, Resistance, and Potential”, opened the conference and sparked significant discussion among attendees.
The project was led by Devin Chamberlain (Cybersecurity ’25), Abby Pitcairn (Computer Science ’25), Nicholas Snow (Information Technology ’25), Ben Sweeney (Computer Science ’26), and Professor David Levine from USM’s Department of Computer Science. Their research during the Spring 2024 semester explored how undergraduate students interact with AI tools—specifically, large language models—when completing programming assignments in introductory computer science courses.
Central to the study was an innovative approach: students completed lab exercises while role-playing one of three carefully designed personas, each representing a different attitude toward using AI tools in academic work. These personas ranged from those eager to use AI for shortcuts to those focused on learning and skill development. The team then analyzed how each type of assignment fared under these different approaches, identifying patterns in which tasks were more or less vulnerable to AI-assisted completion.
Among the key findings:
- Assignments involving spatial reasoning proved particularly challenging for AI tools and were less likely to be completed through shortcut-driven personas.
- Process-oriented assignments, which emphasized the steps of software development rather than just outcomes, were less susceptible to misuse and encouraged deeper engagement.
- The paper also explores constructive uses of AI tools, proposing ways they can be integrated to enhance student learning rather than undermine it.
During the presentation, Professor Levine led the bulk of the discussion, while Abby Pitcairn provided insight into the persona development and role-playing aspects, designed and enacted by the student team.
The paper was well-received and generated lively conversation among educators interested in the evolving role of AI in computer science education. It has since been published in Volume 40, Issue 8 of the Journal of Computing Sciences in Colleges.
