Title: ML-based Preference Elicitation for Course Allocation
Abstract: In some of the most prominent universities, not all students can get a seat in all their preferred courses. As a response to this, some universities have used course allocation mechanisms such as the Bidding Points Auction. However, most of these mechanisms are both unfair and manipulable. To tackle this issue, Eric Budish introduced the Approximate Competitive Equilibrium from Equal Incomes (A-CEEI) mechanism, which satisfies strong fairness and efficiency criteria. This mechanism was implemented at Wharton as Course Match. Even though students could report their preferences accurately enough to realize the theoretical benefits of A-CEEI, preference reporting errors significantly impacted the mechanism's performance. We aim to improve the preference elicitation part of the mechanism using Machine Learning to realize the efficiency and the fairness guarantees of A-CEEI to its potential.
