Beyond Average Effects: A Data-Informed Framework for Targeting Peer Mentorship in Higher Education

Abstract Book of the 11th International Conference on New Trends in Teaching and Education

Year: 2026

[PDF]

Beyond Average Effects: A Data-Informed Framework for Targeting Peer Mentorship in Higher Education

Dr. Thomas Mgonja, Jon Anderson

ABSTRACT:

Peer mentorship is widely used in higher education on the assumption that it helps struggling students most. This study challenges that assumption and proposes a practical framework to help institutions allocate student support more effectively. Using institutional data from over 6,000 remedial mathematics students at an open-enrollment university, the study examines whether mentorship benefits are distributed equally across students. The analytical approach combines machine learning treatment effect estimation, propensity score matching, and explainable AI techniques to capture both overall trends and variation across individual student profiles. The findings confirm modest positive average associations between mentorship and academic outcomes, but reveal substantial variation beneath that average. Students with stronger academic profiles, including higher college GPA and better credit completion rates, show greater estimated responsiveness to mentorship, while students with prior course failures demonstrate more limited gains under current program conditions. Older, non-traditional students also appear more responsive than their younger peers. These patterns raise an important equity concern: uniform mentorship delivery may be concentrating benefits among students who are already relatively advantaged, while those with the greatest academic need gain the least. In response, the study proposes a practical four-step framework for targeting student support based on predicted responsiveness, using data that most institutions already collect. No new infrastructure is required; key indicators such as GPA, prior course failure, and credit completion rate are routinely available in institutional records. By shifting the institutional question from whether mentorship works to which students benefit most and how resources should be allocated accordingly, this study contributes a data-informed approach to improving both the effectiveness and equity of student support systems in higher education.

Keywords: Higher Education; Peer Mentorship; Student Support; Teaching And Learning; Educational Equity; Data-Informed Practice; Learning Analytics; Educational Measurement; Personalized Learning; Educational Quality