Learning the Distribution Map in Reverse Causal Performative Prediction

2024-05-24

arXiv.org

AI Summary

Predictive models induce distributional shifts through strategic agent behavior—e.g., job applicants optimizing resumes in response to hiring algorithms—posing an inverse-causal challenge. Method: We propose the first learnable distribution mapping framework grounded in microeconomic principles, modeling agents’ boundedly rational responses as the sole driver of distribution shift. Unlike prior approaches, our framework integrates inverse-causal reasoning, micro-behavioral modeling, and statistical risk minimization to yield a falsifiable and estimable distribution mapping estimator. Contribution/Results: We establish theoretical guarantees of statistical consistency for the estimator. Empirical evaluation demonstrates substantial reduction in performative prediction risk, with strong efficacy and robustness across both synthetic benchmarks and real-world socio-computational tasks—including algorithmic hiring and credit scoring.

Technology Category

• Machine Learning: Causal Learning
• Game Theory and Economic Paradigms: Adversarial Learning
• Reasoning under Uncertainty: Sequential Decision Making

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

In numerous predictive scenarios, the predictive model affects the sampling distribution; for example, job applicants often meticulously craft their resumes to navigate through a 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, we introduce a novel approach to learn the distribution shift. Our 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, we employ a microfoundation model for the agents' actions and develop a statistically justified methodology to learn the distribution shift map, which we demonstrate to be effective in minimizing the performative prediction risk.

Problem

• Learning distribution shifts caused by predictive models
• Modeling agents' actions in reverse causal scenarios
• Minimizing performative prediction risk in social computing

Innovation

• Reverse causal model for distribution shift
• Microfoundation model for agent actions
• Statistically justified learning methodology