Distributed Course Allocation with Asymmetric Friendships

Students’ decisions about course enrollment are affected by whether their friends plan to take the same class. In some cases, a student may prefer to enroll in a less desirable course to study with friends rather than take a more preferred course alone. Prior research has shown that taking classes with friends can improve academic performance. This paper studies course allocation mechanisms that explicitly account for students’ friendship relations in a distributed setting. We formulate the problem as an asymmetric distributed constraint optimization problem and introduce a dedicated algorithm tailored to this formulation. The evaluation includes both simulated data and a user study involving 177 students, based on reported preferences over courses and friendships. The results indicate that the proposed algorithm commonly finds feasible allocations with high social welfare, while maintaining fairness and respecting course seat capacity constraints.