REU: RI Small: Effective Preference Reasoning Over Combinatorial Domains: Principles, Problems, Algorithms, and Implementations

Preferences are fundamental attributes of human reasoning and decision making. Understanding and automation of preference reasoning is of major concern to artificial intelligence. In order to suc- cessfully integrate preference reasoning into AI systems, major advances in automating preference reasoning are needed. The project to which an REU supplement is sought aims (1) to identify a space of problems, and es- tablish principles and properties of reasoning about qualitative preferences over combinatorial domains, a setting typical to preference reasoning in practice, (2) to design appropriate reasoning algorithms, and (3) to implement them and demonstrate experimentally their practical effectiveness. The undergraduate mentee that will be selected for funding is expected to focus on and contribute to the objective (3). The student will be involved in researching and developing computational solutions for tasks specific to preference learning and optimization, and in building, collecting, and maintaining data sets to be used in experimental studies. The student will work in the research facilities of the Department of Computer Science. The mentee will work for 10 hours/week for 30 weeks during Fall and Spring semesters and for 20 hrs/week for 10 weeks during summer. The mentee will meet with the PI twice a week to report on their activities, discuss problems and develop work plans. The request is for funds for one student for one year; the total amount of the request is $7,606.