Provides a formal framework for optimizing models whose decisions actively change the distribution of the data they encounter.
April 1, 2026
Original Paper
Performative Scenario Optimization
arXiv · 2603.29982
The Takeaway
Standard ML assumes static data distributions, but real-world systems like LLM guardrails or recommendation engines suffer from feedback loops. This work provides a model-free way to reach stable equilibria in these performative settings, which is essential for long-term AI safety and stability.
From the abstract
This paper introduces a performative scenario optimization framework for decision-dependent chance-constrained problems. Unlike classical stochastic optimization, we account for the feedback loop where decisions actively shape the underlying data-generating process. We define performative solutions as self-consistent equilibria and establish their existence using Kakutani's fixed-point theorem. To ensure computational tractability without requiring an explicit model of the environment, we propos