A new study by Cornell researchers found that eliminating race as a factor in college admissions significantly reduces diversity without significantly improving the college credentials of admitted students. The findings challenge claims that affirmative action unfairly disadvantages more qualified applicants.

Using data from an anonymous university, researchers simulated the impact of the Supreme Court’s 2023 decision in Students for Fair Admissions (SFFA) v. Harvardwhich prohibits race-conscious admissions policies. The study showed a 62% drop in the number of top-ranked applicants identifying as underrepresented minorities (URM) when race was excluded from the school’s applicant ranking algorithm. Meanwhile, admitted students’ test scores increased only marginally, equivalent to an SAT score increase of just 10 points.

“We don’t see any evidence that would support the idea that black and Hispanic applicants are admitted even though there are more qualified applicants in the pool,” said René Kizilcec, associate professor of information sciences at Cornell and lead author of the study.

The research team, including doctoral students Jinsook Lee and Emma Harvey, presented their findings at the ACM Conference on Fairness and Access to Algorithms, Mechanisms, and Optimization in October. The study used AI-based ranking algorithms trained on past admissions data to predict the likelihood of acceptance. By retraining the algorithm to exclude race-related characteristics, researchers were able to assess changes in the demographics and academic metrics of top-ranked applicants.

According to the original algorithm, URM students made up 53% of the top-ranked group, which matches the composition of the admitted class before the SFFA decision. Removing race from the algorithm reduced this to 20%. Rankings also became more arbitrary without considering race, with greater variability depending on the subsets of data used to train the algorithm.

The study also found that test scores, often presented as objective measures of merit, contributed only marginally to applicant rankings due to the abundance of highly qualified applicants. “There is a huge drop in the number of URM students when you look at the top-ranked applicant pool,” Lee said, emphasizing the algorithm’s reliance on race to maintain diversity.

The team’s predictions were validated when the university announced its fall 2024 class demographics, which mirrored the study’s predictions. “Our approach did a pretty good job predicting the decline in URM student numbers,” Harvey said.

The findings highlight the challenges institutions face in maintaining diversity under race-neutral admissions policies. Future research, the authors note, could explore why some schools experienced different outcomes following the SFFA decision, including cases where diversity increased.

Researchers have highlighted the importance of using AI responsibly in admissions, given its growing role in the application review process. “This work is critical to ensuring that AI is used responsibly in admissions,” Kizilcec said.

The study was supported by the National Science Foundation, the Amazon Research Award, the Cornell Center for Social Sciences and other organizations.

FingerLakes1.com is the region’s leading all-digital news publication. The company was founded in 1998 and has been keeping residents informed for over two decades. Do you have a lead? Send it to (protected email).