Large Language Models in Otolaryngology Residency Admissions: A Random Sampling Analysis Demonstrating Inherent Bias

Introduction: The otolaryngology residency selection process has become increasingly challenging and time-consuming. The widespread use of large language models (LLMs) raises the opportunity for otolaryngology residency selection process optimization. This study investigates bias in an artificial intelligence (AI)-based simulation of otolaryngology residency selection committee (RSC) members.

Methods: This study employed random sampling of simulated RSC members using a novel Application Programming Interface (API) to virtually connect to OpenAI’s Generative Pre-Trained Transformer (GPT-4). Simulated RSC members with diverse demographics were tasked with ranking to match 1 applicant among 10 with varied racial, gender, and sexual orientations. All applicants had identical resumes; only demographics of the applicants and RSC members were varied for each simulation. Each RSC simulation ran 1000 times. Chi-square tests analyzed differences across categorical variables.

Results: Simulated RSCs consistently showed racial, gender, and sexual orientation bias. Most applicant pairwise comparisons showed statistical significance (p < 0.05). White and Black RSCs exhibited greatest preference for applicants sharing their own demographic characteristics, favoring White and Black female applicants respectively over others (all pairwise p < 0.001). Asian male applicants were consistently selected least by all RSCs (all pairwise p < 0.05). Male RSCs favored White male and female applicants, while female RSCs preferred LGBTQIA+, White and Black female applicants (all pairwise p < 0.05). High socioeconomic status (SES) RSCs favored White female and LGBTQIA+ applicants, while low SES RSCs favored Black female and LGBTQIA+ applicants over others (all pairwise p < 0.001).

Conclusions: Utilizing LLMs to aid in otolaryngology residency selection may introduce significant racial, gender, and sexual orientation bias. While AI-based tools may alleviate selection burden, the potential for significant bias should be appreciated and minimized to ensure an equitable and diverse field of future otolaryngologists.