Human Versus Machine: How Do We Know Who Is Winning? ROC Analysis for Comparing Human and Machine Performance under Varying Cost-Prevalence Assumptions

Abstract

Background: Receiver operating characteristic (ROC) analysis is commonly used for comparing models and humans; however, the exact analytical techniques vary and some are flawed.

Objectives: The aim of the study is to identify common flaws in ROC analysis for human versus model performance, and address them.

Methods: We review current use and identify common errors. We also review the ROC analysis literature for more appropriate techniques.

Results: We identify concerns in three techniques: (1) using mean human sensitivity and specificity; (2) assuming humans can be approximated by ROCs; and (3) matching sensitivity and specificity. We identify a technique from Provost et al using dominance tables and cost-prevalence gradients that can be adapted to address these concerns.

Conclusion: Dominance tables and cost-prevalence gradients provide far greater detail when comparing performances of models and humans, and address common failings in other approaches. This should be the standard method for such analyses moving forward.

Fig. 1

A set of human performances ( purple dots ) with mean sensitivity and specificity ( purple cross ), the convex hull of optimal human performance ( black ), the estimated ROC of human performance ( green ) and a model's ROC ( red ). All ROCs/ROC estimates dominate the mean performance. The model ROC dominates in high-specificity conditions, and has a higher AUC, but is dominated in the range where humans primarily operate. AUC, area under the curve; ROC, receiver operating characteristic.