The proprietary program used to estimate a convict's chance of recidivism proves to be a little less accurate in its forecasts than averaging people found through Mechanical Turk.

I wonder, are human experts any better at it?