On June 29, 2023, the Supreme Court of the United States restricted the ability of colleges and universities to use race-conscious admission policies. The decision is rooted in the premise that using race in admissions processes unfairly bars more academically qualified students from college and university enrollment.
However, a new study from Cornell University has challenged this theory, providing insight into how the ban on affirmative action impacts student diversity in higher education. According to the study’s findings, removing race data from universities’ previously developed applicant-ranking algorithms causes a significantly less diverse top-ranked pool of applicants without meaningfully increasing the academic merit of that group.
The study authors created an AI admissions algorithm for an unnamed university and trained it on previous admissions decisions. They then retrained the algorithm without race-related data. Over half of the top-ranking applicant pool pulled by the first algorithm consisted of students from underrepresented racial backgrounds, but the second algorithm produced a top-ranking applicant pool where only one fifth of students were from underrepresented racial groups.
Notably, removing race from the AI ranking system had little to no effect on the average standardized test score of the algorithms’ selections. Furthermore, the authors found these algorithms were somewhat arbitrary, as the majority of all student applicants were high-achieving. When race data was removed and the AI system was trained with other random sets of data, the algorithm increased arbitrariness in outcomes for most student applicants.
