21 September 2012

Predictive Policing

Okay, these two may be the most annoying people I've seen on YouTube, but they do raise an interesting topic:

Computerized predictive policing. It's a wonderfully cool thing if you are a police force trying to justify your budget. However, does it really do any good in the real world?

That's an interesting question. It probably doesn't accomplish much for static crime areas. If there are big shopping outlets in an area near transient traffic or economically less well endowed areas there will always be more shoplifting there. The Last Chance trailer park down at the end of the county or the slums over on 5th street, where people who can't get anyone else to rent to them are eking out an existence will always be a place where there will be problems on Friday and Saturday nights. The one chain of convenient stores which still takes checks will be the one where people go to write bad checks. I can point to patterns of behavior in my jurisdiction consistent with all of these. I don't need a computer to do it. I guarantee that the deputies and police officers who work the streets could predict these things even better than I can, without using computers.

Where predictive policing might be more useful is in marginal areas or when crime will transfer location for some reason. A marginal area is a place where the crime rate could be significantly curtailed by a police presence. It's not Snobby Mount Gated Community or Last Chance trailer park. It's the shopping center outside a lower middle class suburb which has a mid-level crime rate that could drop quite a bit if two officers were constantly walking a beat back and forth through from store to store. A predictive model might be helpful in determining which ares could benefit most from this.

Predicting crime transfer could also be extremely useful. If the feds sweep in and clean up 5th Street so that there are no dealers left there, where will new dealers pop up? More importantly, how could police act so that fewer new dealers pop up? After all, if the predictive model says that shutting down 5th Street means that dealing will shift to 7th does it do much good to station three police cars constantly on 7th? Probably not. That will just cause it to shift to 9th instead. However, if the perdictive model states that shutting down the dealers on 5th and keeping a one car patrol on 7th, 8th, and 9th will lead to a 70% reduction in drug dealing, that would be extremely useful information.

Of course, the problem with a predictive model is that it does not prove that any individual has or will commit a crime. It can only show trends and therefore, for the predictive model to be useful officers must be tasked so that an area is constantly covered by a law enforcement presence. A city police force with 25 people on shift, or a sheriff's department with a 6 deputy shift covering an entire county, probably won't have the ability to do this. As much as we would like our law enforcement to be proactive, reactive policing will always take priority. After all, there's no police chief in the country who wants to stand in front of his city coucil and say "Well, yes, Officer Smith could have intercepted the guy who robbed the pharmacy. That would have taken him three blocks outside of his mandated area." Consequently, even if the model is correct there will be few local law enforcement agencies that will be able to use it effectively.

No comments: