The Vanderbilt AI Cognition Lab uncovers and studies the cognitive phenomena that AI models, most notably LLMs, manifest.  These phenomena center around categorization, inference, decision making, and problem solving. Because LLMs are trained on human generated documents and other materials, it is not surprising that some behaviors such as typicality and fan effects align with analog effects in humans, but the lab is interested in identifying whatever manifestations…

While I have listed themes of categorization, inference, problem solving, decision making, learning, and creativity separately, they overlap very substantially. Indeed, learning will pervade the material we cover on the other themes. Because learning will be pervasive, I won’t call it out separately on the course Schedule — it will be part of many weeks’ material and discussions. Ideally, our discussions will synthesize across the topics listed as the course progresses, and I’ve set aside class time at the end of the semester to treat synthesis explicitly.

This is a seminar course of readings and discussion, not a lecture-based course, with a final student project being the focus of the last third of the semester. The coverage is far from exhaustive. I have assigned many readings that are “classics” or otherwise exemplars, but you will augment these articles by finding, reading, and reporting on more recent and more diverse works that reference and perhaps expand on these earlier papers. These additional readings and final project will allow you to somewhat customize the course to your particular interests, while staying within themes that we are addressing.