The goal of predictive policing is to do away with racial bias. A computer is used to predict where crimes will take place. The algorithm can tell officers where to go in advance if there’s a high probability of this activity, and they can hopefully respond quickly. This seems like a wise move since a computer, in theory, can’t be biased.
The influence of skewed data
The problem is that you have to “train” the system by submitting data about crimes that take place. An officer who makes a drug arrest on the same street corner every other week knows that this is a hotspot for drug sales, and putting that data into the system means the computer is going to suggest that police go to that location more often.
The issue is that you can have biased training data that skews the results. The computer isn’t biased, but the officer may be.
For instance, say there are two neighborhoods, one with mostly white citizens and the other with mostly African American citizens. The same amount of drug sales happen in both neighborhoods, but the officer is biased against the African American community. He or she goes to that neighborhood far more often, makes far more arrests, and then puts that data into the system.
The computer then predicts that more crime will take place in the African American neighborhood, which sends more officers there. They make more arrests, and the whole thing becomes a sort of feedback loop that enforces that biased initial data.
After an arrest
Bias and prejudice can lead to arrests, and this is just one way it happens. If you get arrested, you must understand your legal defense options.