预测大学新生GPA：传统与公平意识机器学习模型的比较研究

October 2025 Predicting College Freshman GPA: AComparative Study of Traditional and Fairness-Aware Machine Learning Models EDGAR I. SANCHEZ Conclusions This study concludes that traditional logistic regression models, particularly those using ACTComposite scores, tend to demonstrate better fairness metrics across subgroups compared to afairness-aware machine learning gradient-boosted machine model. The exclusion ofrace/ethnicity from predictive models does not introduce notable bias and may even enhancefairness, providing a lawful and effective way to evaluate students’ potential success in college.The findings suggest that postsecondary institutions should adopt a combined approach usingboth high school GPA and ACT scores to strike a balance between fairness and predictiveaccuracy, while being cautious with fairness-aware machine learning models due to theircomplexity and potential biases. So What? The practical importance of this study lies in its implications for postsecondary institutions,especially in light of the 2023 U.S. Supreme Court decision that ended affirmative action incollege admissions. By comparing traditional logistic regression models with fairness-awaremachine learning models, the study provides insights into how institutions can developpredictive models that balance fairness and accuracy without relying on race/ethnicity. This iscrucial for complying with legal mandates while promoting equitable educational outcomes. Thefindings suggest that using a combined approach of high school GPA and ACT scores can helppromote fairness and improve the predictive accuracy of student success, allowing institutionsto more effectively allocate resources and supports to students who need them most. Now What? First, postsecondary institutions should consider adopting a combined approach using both highschool GPA and ACT scores to develop predictive models that balance fairness and accuracy.This approach helps mitigate potential biases that arise when a model relies solely on onemetric, particularly for African American and low-income students. Additionally, institutionsshould explore the use of fairness-aware machine learning models, but with caution, as thesemodels may require further optimization to address potential biases and may be difficult tojustify to parents and lawmakers. Postsecondary institutions should also focus on transparencyand accountability in their decision-making processes, ensuring that the selection of specificmodels and criteria can be easily understood by and justified to students, parents, and legalauthorities. Finally, institutions should incorporate nontraditional factors such as personalessays, socioeconomic background, and school context into their admissions processes topromote a more holistic evaluation of students. About the Author Acknowledgments Dr. Edgar I. Sanchez is a lead researchscientist at ACT, where he studiespostsecondary admissions, national testingprograms, test preparation efficacy, andintervention effectiveness. Throughout hiscareer, Dr. Sanchez has focused on studyingthe transition between high school andcollege and supporting the decision-makingcapacity of students, their families, andcollege administrators. His research hasbeen widely cited in academic literature andby the media, includingThe Wall StreetJournal,The Washington Post,USA Today,and the education trade press. The author would like to thank JoannMoore and Jill McVey for theircomments on previous drafts of thisreport. Introduction Predicting first-year college GPA (FYGPA) is foundational to assessing students’ educationalsuccess, empowering postsecondary institutions to identify student characteristics that influenceacademic performance, and facilitating the design of targeted interventions. Studies haveconsistently shown that the predictive power of standardized tests such as the ACT and SATand high school metrics like high school GPA (HSGPA) and coursework patterns are importantfor predicting freshman grades (Beard & Marini, 2018; Curabay, 2016; Friedman et al., 2024;Marini et al., 2019; McNeish et al., 2015; Sanchez, 2024; Warren & Goins, 2019; Westrick et al.,2015). Traditional metrics, like HSGPA and college entrance exam scores, offer invaluableinsight into students’ readiness for higher education and serve as important indicators ofacademic success and as tools for future planning. When institutions better understand howthese factors contribute to academic success, they can allocate resources and support tostudents who might be in more need, especially in the critical transition from high school tocollege. In 2023, the U.S. Supreme Court issued a landmark decision inStudents for Fair Admissions v.Harvard(2023) andStudents for Fair Admissions v. University of North Carolina(2023),effectively ending affirmative action in U.S. college admissions. The decision stated that racecould no longer be considered in admissions decisions, emphasizing that such practices violatethe Equ