Alley Heaps internship allows computer science students to tackle big research questions

Can experienced programmers really tell when code was written by AI? Does sounding confident after a hockey game actually mean a team played well? And how can machine learning help in medical research? These are questions three StFX computer science students spent their summer exploring, thanks to the Alley Heaps Undergraduate Research Internship.

Avery Doiron, Neel Gopaul, and Keshav Lakhan each received the $9,250 research award, which allows computer science students to conduct research under the guidance of faculty mentors. The internships are part of the Dr. H. Stanley & Doreen Alley Heaps Chair, which provides for the support, exploration, and advancement of computing science at StFX.

DISTINGUISHING AI GENERATED CODE

"The experience gave me an opportunity to go beyond what I had learned in the classroom and experience the research process in a much more independent way. It challenged me to think critically, work through uncertainty, and become more confident in approaching problems where the answer is not necessarily known beforehand," says Keshav Lakhan, a third year computer science student from Trinidad and Tobago.

"It has also changed the way I think about AI. Instead of only thinking about what these systems can or cannot do, I have become more interested in the questions surrounding how people interact with them, how we evaluate their outputs, and how institutions respond to them. I also gained a much better appreciation for the amount of thought, patience, and refinement that goes into meaningful research."

In his research, Mr. Lakhan looked at whether humans can reliably distinguish AI-generated source code from code written by humans.

"With generative AI tools becoming increasingly common in programming and education, there is often an assumption that experienced programmers can recognize AI-generated code based on things like its structure, naming conventions, or commenting style. I wanted to examine whether that assumption actually holds up when tested."

He hopes the research can contribute evidence to discussions around AI, computer science education, and academic integrity rather than relying solely on assumptions about what AI-generated work looks like.

Mr. Lakhan says what interested him most was how quickly generative AI has become part of programming and education.

"As a computer science student, I have seen how much discussion there is around AI-generated code and whether someone can simply 'tell' when code was produced using AI. I became interested in questioning that assumption. If people are making decisions about academic integrity or authorship based on their intuition, I think it is important to understand how reliable that intuition actually is. I also found the human side of the problem particularly interesting. It is not just about what AI can generate, but about how we interpret what we see, what clues we rely on, and whether our confidence in those judgments matches our actual accuracy."

His research supervisors were Dr. Milton King, Dr. Taylor Smith, and Dr. Jean-Alexis Delamer.

EXPLORING SENTIMENT BEHIND WORDS

Avery Doiron, a fourth year honours computer science student from Boylston, NS, says opportunities like this are valuable to students as "they let us pursue a project we're genuinely interested in while learning from experienced faculty and gaining skills that can be applied in the future, whether it's coursework, more research, postgraduate studies, or industry work."

He explored how NHL players and coaches speak in interviews, specifically the sentiment behind their words (positive, neutral, or negative) to see whether that relates to how they perform in games.

"My interest came from wanting to apply what I learned in Dr. Milton King's 'Natural Language Processing' class to a domain I'm very passionate about, hockey and the NHL," says Mr. Doiron, who was supervised by Dr. King and Dr. Kyran Cupido.

"The most interesting thing I discovered was that coaches tend to speak more neutrally than players. I also found that a player sounding confident in a pre-game interview doesn't make a win more likely, but players are much more likely to sound confident in a post-game interview after they've actually won."

Mr. Doiron says this experience gave him hands-on experience with a full research project and taught him how to navigate the uncertainty of an exploratory research question, from data collection to problem-solving, and to collaborate with professors to improve his approach.

MACHINE LEARNING IN MEDICAL RESEARCH

"I'm very grateful for the opportunity to take part in this research. It was an enriching experience that allowed me to apply what I've learned in my courses to a real-world problem while developing my technical and problem-solving skills. Working with Dr. Jacob Levman was also very valuable, as he was very helpful throughout the project and gave me the opportunity to learn more about the applications of machine learning in medical research," says Neel Gopaul, a fourth year computer science student from Flic en Flac, Mauritius.

He focused on using structural Magnetic Resonance Imaging (MRI) and machine learning to investigate potential biomarkers of schizophrenia.

He worked with MRI data from several different research datasets and extracted measurements of the brain structure such as volume, thickness, and surface area of different brain regions.

"Using df-analyze, a machine-learning platform that automates different stages of the data analysis process, I investigated whether combinations of these measurements could help distinguish patients with schizophrenia from healthy controls. The goal was to determine how accurately these measurements could distinguish the two groups, while also understanding which structural brain features were most informative and whether the findings remained reliable across different datasets and sources."

He was interested in medical imaging, machine learning, and artificial intelligence. The idea of using computational methods to search through thousands of MRI-derived measurements was something he found particularly interesting, since these techniques could help understand patterns hard to manually identify. He was also interested in the potential for these methods to provide more objective, measurable information about schizophrenia and complement traditional approaches to understanding and diagnosing the disorder.

His supervisor was Dr. Jacob Levman, whose research is based on using computational technology and imaging to investigate neurodevelopmental disorders and healthy brain development.

Undergraduate research awards like these give students hands-on experience, help them build important career and academic skills, and bring a sense of personal achievement, he says.