When Personalization Harms Performance

Machine learning models often include group attributes like sex, age, and HIV status for the sake of personalization – i.e., to assign more accurate predictions to heterogeneous subpopulations. In this talk, I will describe how such practices inadvertently lead to worsenalization, by assigning unnecessarily inaccurate predictions to minority groups. I will discuss how these effects violate our basic expectations from personalization and describe how these violations arise due to standard practices in model development. I will end by highlighting recent work on how to address these issues in practice – first, by setting "personalization budgets" to test for worsenalization; second, by developing models where individuals can consent to personalization at prediction time.

Please report video problems to itsupport@slmath.org.

See more of our Streaming videos on our main VMath Videos page.