The New York City crime rate famously plummeted in the mid-1990s under the watch of police chief William Bratton, who introduced a computerized mapping system called CompStat to help cops track crime hot spots. He later took the system to Los Angeles, where once again crime plunged. CompStat is now used nationwide, reports Miller-McCune, and Bratton is a law-enforcement superhero.
Bratton’s latest innovation, called predictive policing, represents the next step in computerized law enforcement—moving beyond charting past felonies to forecasting future offenses. “The only way for us to continue to have crime reduction is to start anticipating where crime is going to occur,” says L.A. police lieutenant Sean Malinowski, with whom Bratton conceived the sophisticated data analysis program.
I guess I’d be really interested to know how knowing where crime tends to happen (CompStat) is different from knowing where crime will occur (New Unnamed Thing). Is it an inference thing, where we put up a new bank and predict that based on our experience bank robberies will go up in the neighborhood?
I actually believe this might work, I just want more detailed information on it. The finally paragraph suggests the counter-intuitive example of street lights *increasing* crime because muggers “need them to do their work”. But there’s a lot of confounding associated with streetlight placement — population density, socioeconomic status, level of foot traffic — is it accounted for?