Performative Prediction

By Juan C. Perdomo, Tijana Zrnic, Celestine Mendler-Dünner, Moritz Hardt. International Conference on Machine Learning, 2020

478 citations recorded by the source corpus.

When predictions support decisions they may influence the outcome they aim to predict. We call such predictions performative; the prediction influences the target. Performativity is a well-studied phenomenon in policy-making that has so far been neglected in supervised learning. When ignored, performativity surfaces as undesirable distribution shift, routinely addressed with retraining. We develop a risk minimization framework for performative prediction bringing together concepts from statistics, game theory, and causality. A conceptual novelty is an equilibrium notion we call performative stability. Performative stability implies that the predictions are calibrated not against past outcomes, but against the future outcomes that manifest from acting on the prediction. Our main results are necessary and sufficient conditions for the convergence of retraining to a performatively stable point of nearly minimal loss. In full generality, performative prediction strictly subsumes the setting known as strategic classification. We thus also give the first sufficient conditions for retraining to overcome strategic feedback effects.

Subfields: Economics and Econometrics, Management Science and Operations Research, Safety Research

View Paper: https://www.semanticscholar.org/paper/cfa46317f14cc32bcbd21a8e02e913616fabdb68

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Economics and Econometrics

In Management Science and Operations Research

In Safety Research

Citations in the corpus

Data Feedback Loops: Model-driven Amplification of Dataset Biases (2022, background): “The general problem of modelinduced feedback in machine learning has been previously studied as performative prediction and strategic classification (Perdomo et al., 2020; Hardt et al., 2016), where future data distributions can change arbitrarily in response to the deployed model.”
The Values Encoded in Machine Learning Research (2021, background): “Furthermore, to the extent that predictions are performative [55], especially predictions that are enacted, those ML models which are deployed to the world will contribute to shaping social patterns.”
Fair Decisions Despite Imperfect Predictions (2019, background): “Such a simplistic approach, which we refer to as learning to predict, ignores that once decisions are based on these predictions, they may interact with the data collection or have direct impact on the underlying distribution relevant during deployment (Perdomo et al., 2020).”