A computational approach to creativity: Fostering success and equity in college admissions

Poster Session B - Sunday, April 14, 2024, 8:00 – 10:00 am EDT, Sheraton Hall ABC

Kibum Moon¹ (km1735@georgetown.edu), Kostadin Kushlev¹, John D. Patterson², Roger Beaty², Adam Green¹; ¹Georgetown University, ²Pennsylvania State University

Creativity, defined as the ability to generate novel and useful ideas or works, plays a pivotal role in fostering innovation and facilitating problem-solving. Existing literature reports that creativity can predict the future academic success of students, while less being biased by sociodemographic factors. However, the broader application of creativity, particularly in college admissions, has been limited due to challenges surrounding the inefficiency and subjectivity inherent in human creativity assessments. In this study, we explore the potential of computational creativity metrics—extracted from college applications—to better predict students' success and to foster equity in college admissions. To overcome assessment challenges associated with human ratings, we leverage state-of-the-art Large Language Models (BERT, GPT, and Llama2) to compute creativity metrics from college application essays. We then test how computational creativity metrics are associated with college admission results and future academic performance (i.e., GPAs). Using a pilot sample of 1,000 applicants, we found that computational creativity metrics positively predicted future academic performance in college, but not admission results—even after controlling for standardized test scores and socio-demographic factors. This suggests that creativity is an important factor in college success that may not be adequately considered in the admissions process. Moreover, unlike standardized test scores, creativity was not related to socio-demographic factors, suggesting creativity may be a less biased feature. We conclude with limitations and future directions, including a forthcoming multi-university, big-data replication with over 300,000 admission essays.

Topic Area: LANGUAGE: Semantic