Course Allocation with Friendships as an Asymmetric Distributed Constraint Optimization Problem Ilya Khakhiashvili ilk17895@gmail.com Ariel Cyber Innovation Center Ariel University Israel Tal Grinshpoun talgr@ariel.ac.il Ariel Cyber Innovation Center Ariel University Israel Lihi Dery lihid@ariel.ac.il Ariel Cyber Innovation Center Ariel University Israel Figure 1: Course allocation with friendships. On the left: the traditional setting for course allocation. On the right: our sugges- tion, course allocation with the addition of friendships ABSTRACT Course allocation, i.e., the problem of assigning students to courses, is a difcult problem. Students value being assigned to the same course as their friends. We propose a model that considers not only the students’ preferences over courses but also their prefer- ences over classmates. We formulate the problem as an asymmetric distributed constraint optimization problem. This solution has an additional interesting feature: it is solved in a distributed manner, thus removing the need to directly share private preferences with anyone. An extensive evaluation of our proposed model on real- world student preferences over courses shows that it obtains high utility for the students, while keeping the solution fair and observ- ing courses’ seat capacity limitations. Our model is general and can be adapted to solve a variety of multi-allocation problems where it is required to consider friendships. CCS CONCEPTS · Computing methodologies → Multi-agent systems; · Infor- mation systems → Expert systems; · Applied computing → Education; · Social and professional topics → Pricing and re- source allocation. Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for proft or commercial advantage and that copies bear this notice and the full citation on the frst page. Copyrights for components of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specifc permission and/or a fee. Request permissions from permissions@acm.org. WI-IAT ’21, December 14ś17, 2021, ESSENDON, VIC, Australia © 2021 Association for Computing Machinery. ACM ISBN 978-1-4503-9115-3/21/12. . . $15.00 https://doi.org/10.1145/3486622.3493990 KEYWORDS multi-unit allocation, course allocation, friendships, ADCOP ACM Reference Format: Ilya Khakhiashvili, Tal Grinshpoun, and Lihi Dery. 2021. Course Allocation with Friendships as an Asymmetric Distributed Constraint Optimization Problem. In IEEE/WIC/ACM International Conference on Web Intelligence (WI-IAT ’21), December 14ś17, 2021, ESSENDON, VIC, Australia. ACM, New York, NY, USA, 6 pages. https://doi.org/10.1145/3486622.3493990 1 INTRODUCTION Most universities face a course allocation problem as demand for popular courses exceeds supply. Often, there is a constraint on the number of students who can attend each course and a few courses are highly popular while in others demand is scarce. Money does not solve the problem since it is not possible for students to pay a higher fee for much wanted courses. Therefore, institutes have to decide how to allocate students to courses under these limitations. Course allocation is a multi-unit allocation problem, since each student is required to take several courses (units). While mecha- nisms exist for single-unit allocation under certain conditions (see works that stem from [8]), for multi-unit problems, the only solu- tions which are Pareto efcient and strategy-proof are dictatorships [4, 14]. Furthermore, course allocation incorporates a unique prop- erty among multi-unit allocation problems: friendships between the agents. The agents are students, and as such, being allocated to a course with their friends and study groups is a top priority for them. It may be that a student will prefer an assignment to a course with her friends, to an assignment to a course that is high on her preference list but will not be attended by her friends. In this paper, we model the allocation of courses to students, while considering not only the students’ preferences over courses 688