What if…and “Predictive Policing”

As described here by Talking Points Memo, the police department of Santa Cruz, California is experimenting with an algorithm designed to predict when and where a crime will take place.

Originally designed to predict earthquake aftershocks, the algorithm works under the assumption that crimes are likely to be repeated in the same place and at the same time of day. For example, if a burglar knows a neighborhood has few dogs and nearly everyone is gone around mid-day, he is likely to rob several houses in that neighborhood around mid-day. The software has been surprisingly effective at stopping property crimes, as police officers can be more effectively positioned throughout the city, ready to question suspicious individuals who happen to be the wrong place at the wrong time (in Santa Cruz, the crime is often repeated four days later).

U.S. cities are increasingly turning to such software. New York experimented with a system, Compstat, in the mid-90s to some success. But unlike Compstat, the Santa Cruz system can be updated daily, and the more data it is fed, the more accurate the system becomes.

In “Minority Report,” Philip K. Dick wrote about the consequences of a law enforcement system built around mutants who could predict crime. How would people react to widespread use of software-based predictive crime? What if it went beyond property crimes, and it was designed to predict all crime? What if people encouraged its adoption, believing it was impartial? What if people worried it could be hacked? If the system becomes increasingly accurate as it is fed more data, what if a “law enforcement singularity” took place? What else might the software predict? Then how would people react to a system that was able to accurately predict much of human behavior, on a global scale?

How would the criminals react? How long after such publicity do the criminals begin to shift their behavior, and to what extent would they be successful? Criminals who knew how the algorithm worked would be able to break the law more effectively. What if they tried to kidnap the software designer? What if criminals could continually shift their behavior so as to take advantage of the gap time the system would need to determine the new behavioral patterns? What if, in response to the system’s increased accuracy, criminals begin to act more randomly and more violently?

What do you think of predictive crime? What do you imagine the consequences would be?

2 thoughts on “What if…and “Predictive Policing””

    1. Well, “prediction” is not quite correct here, but an interesting article nevertheless. It would be interesting to see the software jump to prediction, then correlation, then causation, so you could determine the causes of unrest and whether it could be prevented.