Evaluating the efficiency of AI-generated transfer plans

• In Event: AI and Transfer Student Success: Learning From the Field

Abstract

Creating credit mobility pathways is only useful for students if they know about them. In this work, we utilize artificial intelligence to generate cross-campus course schedules for transfer students and then conduct a study to compare the accuracy and time-on-task of advisors when creating a plan for a student from scratch vs if they are given an AI-generated plan as a first draft. Using a constraint satisfaction approach, course schedules for students who have completed their first year at a two-year college can be generated, mapping out their remaining semesters at the 2-year as well as subsequent courses at the 4-year to meet all graduation requirements for the Criminology major. To evaluate the benefits of such an AI system, advisors from the State University of New York (SUNY) system participated in an experiment to measure how utilization of AI-generated schedules affects time spent and schedule quality. Results will inform if trade-offs exist between efficiency and quality when adopting AI-in-the-loop transfer advising.

Authors

• Ritik Sinha, University of California - Berkeley

• Zachary A. Pardos, University of California - Berkeley