Predictive Policing. Predpol is an algorithm that predicts crime occurrence by location in nearly real time, using historical data on crime type, crime location, and crime date/time as predictors. Police use the service to deploy patrols efficiently.

a. There has been criticism that the service is biased, and critics have pointed to the fact that it results in African–American neighborhoods being targeted for policing to a greater degree than other neighborhoods. Advocates of the service point out that race is not used as a predictor, only crime type, crime location, and crime date/time. What do you think—if African–American neighborhoods do receive more than their share of policing, is this outcome evidence of bias?

b. Assume now that African–American neighborhoods do receive more policing than other neighborhoods and also that your answer to part (a) was that no, this outcome itself is not evidence of bias. Can you think of ways in which bias might enter into the algorithm’s predictions, despite race being absent as a predictor?