Investigating Human + Machine Complementarity for Recidivism Predictions

TL;DR

This study compares human and algorithmic risk assessments for recidivism, analyzing their differences and potential for combined use, but finds limited improvements through hybrid models on the dataset.

Contribution

It introduces a detailed analysis of human versus machine decision making in recidivism prediction and proposes best practices for leveraging their complementary strengths.

Findings

Human and COMPAS decisions differ but do not significantly improve predictions when combined.

Hybrid Human+Machine models do not outperform individual models on this dataset.

Recommendations for data collection to better utilize human and machine strengths in fairness assessments.

Abstract

When might human input help (or not) when assessing risk in fairness domains? Dressel and Farid (2018) asked Mechanical Turk workers to evaluate a subset of defendants in the ProPublica COMPAS data for risk of recidivism, and concluded that COMPAS predictions were no more accurate or fair than predictions made by humans. We delve deeper into this claim to explore differences in human and algorithmic decision making. We construct a Human Risk Score based on the predictions made by multiple Turk workers, characterize the features that determine agreement and disagreement between COMPAS and Human Scores, and construct hybrid Human+Machine models to predict recidivism. Our key finding is that on this data set, Human and COMPAS decision making differed, but not in ways that could be leveraged to significantly improve ground-truth prediction. We present the results of our analyses and suggestions for data collection best practices to leverage complementary strengths of human and machines in the fairness domain.