Shaghayegh (Sherry) Sahebi

Assistant Professor

University at Albany, SUNY, Department of Computer Science

Machine Learning for Human Learning: The Case of Academic Procrastination

Academic procrastination, i.e., postponing planned studies despite unfavorable consequences, has been associated with negative side effects on students' performance and psychological well-being. This behavior is common, especially in online education settings that rely on student self-regulation. To detect and manage this behavior, it is essential to model and understand the dynamics of procrastination. In this talk, I present my research on modeling student learning activities and their relation to procrastination.
Procrastination is often determined using self-reports or summarized static behavioral measures, such as averaged assignment starting time. I showcase the need for continuous-time modeling of student behaviors in detecting procrastination, and the first-of-its-kind point processes model proposed for this purpose. For predicting procrastination, however, traditional point processes are not directly applicable as they model each process independently and are unable to deal with sparse and missing data. I proceed to present our relaxed clustered Hawkes process model that addresses these challenges and discovers a personalized joint model of partially observed student activity timings. Furthermore, I introduce our stimuli-sensitive Hawkes modeling of student activities that estimates the effects of external triggers, such as deadlines, on their learning behaviors and predicts when a student comes back to study. Finally, I present my future directions on holistic modeling of student knowledge, performance, and procrastination.

RSVP: https://pitt.co1.qualtrics.com/jfe/form/SV_cvwbRGyPQh5jUWi