Research Case Studies is a project-based module with students splitting into small groups of five to seven, and each group works quantitatively on a real research question for a full semester, from a feasibility pitch to a data-driven NLP pipeline to a poster presentation. Groups are self-organizing, with supervisors guiding rather than managing: setting expectations, checking in biweekly, and helping steer scope so the semester doesn’t dissolve into busywork on non-essential parts of the project.
This year I’m supervising a group on AI Psychometrics: Construct Validity and Biases in LLMs for Social Science. The aim is to investigate how well Large Language Models (LLMs) hold up as stand-ins for human respondents on established psychometric instruments, and to surface the systematic biases that show up when they’re pushed into that role. It’s a direct extension of the work I did throughout my Ph.D. on LLMs as human simulacra, and this time the goal is to get the students’ findings into shape for submission to a workshop, turning a semester project into an actual contribution to the field, not just a poster and a portfolio.