"Eye of the Tiger'': An LLM-Powered Approach to Personalized Academic Advising

Abstract

This thesis introduces Tiggy, an LLM-powered academic advising chatbot designed to assist Princeton University undergraduates with course selection. Tiggy serves as an always-available course advisor, capable of primarily handling three types of queries: similarity queries, which recommend courses conceptually related to a reference course; requirement queries, which filter by specific distribution or degree requirements; and subject-area queries, which surface courses within a particular academic department. Students complete a brief onboarding process to input their class year, major, and past coursework, which Tiggy uses to personalize every recommendation. A within-subject study comparing Tiggy to traditional manual course planning across fourteen Princeton undergraduates found that Tiggy significantly outperformed manual planning in five of six evaluated dimensions: efficiency and speed, frustration and cognitive load, outcome satisfaction, exploration and discovery, and trust and adoption. Most notably, more than half of Tiggy users discovered a course they would not have otherwise considered, demonstrating Tiggy’s power as a recommender rather than a mere filter. These findings suggest that personalized, institution-specific AI advising tools have the potential to meaningfully transform the course selection experience for university students. The code is available on GitHub (https://github.com/smichaelides/Tiggy) and the app is deployed on Vercel (https://tiggy-vercel.vercel.app/login).