As algorithms increasingly determine access to valued resources and opportunities, we want to investigate whether introducing structured randomization into decisions can increase fairness while maintaining efficiency. We aim to develop a theoretical framework to study tradeoffs between decisions under uncertainty. This involves defining what fairness means for different stakeholders and what utility function the decision-maker is using. We will validate our framework with an empirical case study using real-world data. For example, exam school policies admit students based on a composite of their exam score, GPA, and socioeconomic status. We would see how randomization affects the balance of admitted students, and whether this would lead to a more “fair” outcome overall.

Through this SuperUROP, I’m hoping to explore algorithmic fairness in depth. I’ve previously worked on conducting black-box audits of recommendation algorithms and synthesizing global AI regulation and policies, and I hope to learn more about the field via this experience and from other members of the lab. I’m really excited by the chance to connect our research to a real-world case study.