AUTHOR=Suzuki Honoka , Hong Maxwell , Ober Teresa , Cheng Ying TITLE=Prediction of differential performance between advanced placement exam scores and class grades using machine learning JOURNAL=Frontiers in Education VOLUME=7 YEAR=2022 URL=https://www.frontiersin.org/journals/education/articles/10.3389/feduc.2022.1007779 DOI=10.3389/feduc.2022.1007779 ISSN=2504-284X ABSTRACT=Introduction

Past studies have found students to perform differently between class grades and standardized test scores – two essential and complementary measures of student achievement. This study examines predictors of the relative performance between these two measures in the context of the advanced placement (AP) program, namely, we compared students’ AP exam scores to the class grade they received in the corresponding AP course. For example, if a student received a high AP class grade but a low AP exam score, what characteristics about the student or their learning context might explain such discrepancy?

Methods

We used machine learning, specifically random forests, and model interpretation methods on data collected from 381 high school students enrolled in an AP Statistics course in the 2017–2018 academic year, and additionally replicated our analyses on a separate cohort of 422 AP Statistics students from the 2018–2019 academic year.

Results

Both analyses highlighted students’ school and behavioral engagement as predictors of differential performance between AP class grades and AP exam scores.

Discussion

Associations between behavioral engagement and differential performance suggest that the ways in which a student interacts with AP course material to obtain high class grades can differ from study habits that lead to optimal performance on the AP exam. Additionally, school-level differences in relative performance pose equity concerns towards the use of AP exam scores in high-stakes decisions, such as college admissions. Implications are discussed from a pedagogical and policy perspective.