Be Intentional About Fairness - Fairness, Size, and Multiplicity in the Rashomon Set

Explore the complex relationship between algorithmic fairness and model multiplicity in this 15-minute conference talk from the Association for Computing Machinery's Algorithmic Fairness & Recourse session. Delve into how the size and characteristics of the Rashomon set—the collection of models that perform similarly well on a given task—impact fairness considerations in machine learning systems. Learn about the critical importance of being intentional when selecting models from this set, as different equally-performing models can exhibit vastly different fairness properties. Examine research findings on how model multiplicity affects fairness outcomes and discover practical approaches for navigating the trade-offs between predictive performance and equitable treatment across different demographic groups. Gain insights into why simply optimizing for accuracy is insufficient and understand the methodological considerations necessary for developing fair and responsible AI systems.