Individual Assessment created for International Baccalaureate Math Analysis and Approaches Standard Level class
Using ProPublica's data set of COMPAS's predicted risk of recidivism v. actual recidivism rates for thousands of individuals, this project analyzes bias in the COMPAS system using confusion matrices, graphs, and statistical non-discrimination criteria. The author concludes that depending on what criteria you use, the model is either biased towards or against BIPOC individuals, and that the overall problem lies not with the model but the system that it works within.