This is the code for our paper Addressing Polarization And Unfairness In Performative Prediction.

Run the notebooks below to reproduce each experiment:

• Examples.ipynb: reproduces examples in main paper.
• gaussian_clf_exp.ipynb: reproduces every synthetic Gaussian classification experiment, from trajectory generation to fairness/utility figures.
• credit_exp.ipynb: loads the credit dataset, runs the retention/fairness baselines, and exports the reported plots.
• income_exp.ipynb: runs the ACS Income pipeline end-to-end, including sampling, training, and visualization of the main metrics.
• mnist_exp.ipynb: evaluates the MNIST group-split setting with strategic behavior toggles.
• Strat_exp.ipynb: studies strategic manipulation in credit data setting.
• Strat_k_delayed_exp.ipynb: extends the strategic experiments to the K-delayed feedback model.
• regression_exp.ipynb: reproduces the regression version of the Gaussian experiments.
• regression_e5.ipynb: focuses on Example 5 from the paper, illustrating the variance-regularized objective.
• regression_exp_multi.ipynb: covers the multi-group regression simulations and associated fairness metrics.

The paper is mainly theoretical. If you take interest, check the paper here