Pursuing Open Source Development of Predictive Algorithms: The Case of Criminal Sentencing Algorithms

• In Event: Machine Learning and Predictive Models

Brief Overview

Issues plaguing sentencing algorithms are exacerbated by their proprietary nature. We replicate a major algorithm using real criminal data, then fit three machine learning algorithms, finding an increase in predictive accuracy of open source options.

Authors

• Philip Waggoner, University of Chicago

• Alec Macmillen, University of Chicago