Addressing Polarization and Unfairness in Performative Prediction