Reproduce “Revisiting the Predictability of Performative, Social Events” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Juan C. Perdomo
Juan Carlos Perdomo's research at Harvard University develops an algorithmic framework for efficiently finding performatively multicalibrated predictions, notably demonstrating that these rigorously "valid" forecasts can paradoxically lead to maximal risk in dynamic social systems. The work provides a modern answer to a long-standing problem regarding the predictability of self-influencing social events.
Key insights
- Performatively multicalibrated (or outcome indistinguishable) predictions are algorithmically achievable with computational and statistical efficiency comparable to supervised learning, under minimal assumptions.
- While performative multicalibration implies performative stability, it does not guarantee performative optimality or risk minimization.
- A rigorously calibrated predictor can surprisingly maximize performative risk in certain scenarios, challenging the traditional desirability of calibration in dynamic social contexts.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.