Distributionally Robust Performative Prediction

2024-12-05

Neural Information Processing Systems

AI Summary

In performative prediction, misspecification of the distribution map causes the conventional performative optimal (PO) solution to deviate from the true optimum, introducing deployment risk. To address this, we propose Distributionally Robust Performative Optimality (DRPO), the first framework to integrate distributionally robust optimization into performative prediction. DRPO ensures robust approximation of the true performative optimum under both microscopic and macroscopic errors in the distribution map. We formally define DRPO, establish its theoretical guarantees, and prove its equivalence to an augmented performative prediction problem—enabling direct adoption of existing solution frameworks. Experiments across diverse misspecification scenarios demonstrate that DRPO significantly outperforms PO, enhancing post-deployment stability and generalization. Our approach provides a principled, reliability-oriented paradigm for deploying predictive models in real-world systems.

Technology Category

• Intelligent Robots: Learning & Optimization for ROB
• Reasoning under Uncertainty: Stochastic Optimization
• Search and Optimization: Learning to Search

Application Category

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

Abstract

Performative prediction aims to model scenarios where predictive outcomes subsequently influence the very systems they target. The pursuit of a performative optimum (PO) -- minimizing performative risk -- is generally reliant on modeling of the distribution map, which characterizes how a deployed ML model alters the data distribution. Unfortunately, inevitable misspecification of the distribution map can lead to a poor approximation of the true PO. To address this issue, we introduce a novel framework of distributionally robust performative prediction and study a new solution concept termed as distributionally robust performative optimum (DRPO). We show provable guarantees for DRPO as a robust approximation to the true PO when the nominal distribution map is different from the actual one. Moreover, distributionally robust performative prediction can be reformulated as an augmented performative prediction problem, enabling efficient optimization. The experimental results demonstrate that DRPO offers potential advantages over traditional PO approach when the distribution map is misspecified at either micro- or macro-level.

Problem

• Modeling predictive outcomes influencing systems
• Addressing misspecification of distribution map
• Introducing robust approximation for performative optimum

Innovation

• Introduces distributionally robust performative prediction
• Defines distributionally robust performative optimum
• Reformulates as augmented performative prediction