Outline

This document contains the current main and appendix figures. A few points to note:

Some MTurk data is still outstanding, and I have indicated where plots will be updated
The titles of the figures will be added in the overleaf
The heading of each figure has (1) the number from the google doc, (2) the figure-name as in the google doc and sheet, and (3) the title of the figure

Main text figures

Figure 2 - fig:CombinedROC - Algorithm’s predictive accuracy for judge decisions

Figure 3 - fig:MLFaceDetentionEffect - “Relative Magnitude of Face Effect on Detention”

Figure 4 - fig:MLFaceRiskCorrelation - “Relationship between predicted judge detain decision and judge decision by defendant race”

Figure 6 - fig:GANMorphingCombined - “Illustration of morphed faces along age and detention gradients”

NOTE The final image is combined outside of R, so this only shows the trajectories which are pasted below the image morphs.

Figure 7 - fig:MTurkLearning

NOTE These are old plots, I am waiting for the new MTurk data (1-2 weeks according to James) to add the third line and update these.

Figure 9 - fig:WordCloudAfter - What do poeple see?

NOTE These are old plots, I am waiting for the new MTurk data (1-2 weeks according to James) to add the third line and update these.

Appendix Figures:

Appendix Figure 3 - fig:MTurkSkinToneDistribution - “Descriptive Statistics About Skin-Tones”

Appendix Figure 4 - fig:MturkReliabilityMetrics - “Reliability measures for human-rated psychological features”

Appendix Figure 5 - fig:HumanRatingDistribution - “Distribution of human ratings of psychological features based on face images”

Appendix Figure 6 - fig:HumanRatingsError - “Signal vs. noise in human ratings by number of ratings provided.”

Appendix Figure 10 - fig:RiskReArrestCorrelation - “Predicted Risk vs. Re-Arrest Rates”