A1639
Title: Learning the distribution map in reverse causal performative prediction
Authors: Daniele Bracale - University of Michigan (United States) [presenting]
Yuekai Sun - University of Michigan (United States)
Moulinath Banerjee - University of Michigan (United States)
Subha Maity - University of Pennsylvania (United States)
Abstract: In numerous predictive scenarios, the predictive model affects the sampling distribution; for example, job applicants often meticulously craft their resumes to navigate through screening systems. Such shifts in distribution are particularly prevalent in the realm of social computing, yet the strategies to learn these shifts from data remain remarkably limited. Inspired by a microeconomic model that adeptly characterizes agents' behavior within labor markets, a novel approach is introduced to learn the distribution shift. The method is predicated on a reverse causal model, wherein the predictive model instigates a distribution shift exclusively through a finite set of agents' actions. Within this framework, a microfoundation model is employed for the agents' actions, and a statistically justified methodology is developed to learn the distribution shift map, which is demonstrated to be effective in minimizing the performative prediction risk.
