GSE Colloquium Series in Education Data Science: Conrad Borchers

Toward Effort-Sensitive AI: Modeling and Supporting Student Persistence Across Contexts

Over four decades of AI in education research, intelligent tutoring systems have produced strong evidence of effectiveness. Yet, long-term student persistence in these systems remains a grand challenge. My research demonstrates that learning analytics can help students better regulate their effort while creating insights into how, why, and when learners persist, opening new opportunities for adaptive systems grounded in behavioral data.

This talk highlights two case studies. The first shows how student effort in K-12 classrooms can be sensed using system log data combined with classroom context to model persistence over time. My recent research shows that these models help students set and adjust weekly goals, sustainably enhancing effort, improving math proficiency by 40% compared to practice without weekly goals, and making goal achievement up to 80% more likely than with teacher-set goals. The second case study examines higher education, showing how enrollment and clickstream data from learning management systems can model student workload with greater accuracy than traditional credit hours, enabling large-scale analyses of course success and how stable learner traits shape persistence.

Looking ahead, my research will advance process models of student persistence and effort regulation (i.e., how learners start, stop, and take breaks) and the factors driving sustained effort. I will demonstrate how foundation models, such as large language models, can help assess persistence and motivation from multimodal data beyond traditional log-based analysis. Through partnerships with educators, platform providers, and academic advisors, this work aims to strengthen students' self-regulation and persistence, skills essential for success in school and beyond.

Conrad Borchers is a PhD candidate in the Human-Computer Interaction Institute at Carnegie Mellon University’s School of Computer Science, co-advised by Vincent Aleven and Ken Koedinger. His research advances intelligent systems that promote learner persistence, assessment, and educational outcomes through human-centered design and learning analytics. His adaptive goal-support systems are deployed in partnership with teachers and schools, reaching nearly 1,500 middle school students. He also develops novel methods for analyzing learner regulation through conversational interactions and trait modeling. Conrad holds an MSc in Social Data Science from the University of Oxford and a BSc in Psychology from the University of Tübingen.