[2002.06673] Performative Prediction

Abstract: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.

Performative Prediction Juan C. Perdomo* Tijana Zrnic* Celestine Mendler-Dünner Moritz Hardt {jcperdomo, tijana.zrnic, mendler, hardt}@berkeley.edu University of California, Berkeley arXiv:2002.06673v4 [cs.LG] 26 Feb 2021 March 2, 2021…

related reading

• Performative Predictionarxiv.org
• SFT, RL, and On-Policy Distillation Through a Distributional Lens | whnrehiew.github.io
• RL's Razor: Why Online Reinforcement Learning Forgets Lessarxiv.org
• aiesi_post-training_public.pdfkawine.github.io
• 1011.0686arxiv.org
• Fundamentals of predictionmlstory.org
• A Course in Machine Learningciml.info
• Optimisation over non-stationary distributions creates weirder minds — LessWronglesswrong.com
• arxiv.org/pdf/2511.08544arxiv.org
• Deep Anti-Determinismlydianottingham.substack.com
• Foundation Models for Oversight | Transluce AItransluce.org
• Is Capability a Liability? More Capable Language Models Make Worse Forecasts When It Matters Mostarxiv.org