crime map.png

PredPol maps this red box onto areas where heightened crime is predicted to occur. Also.. it's their logo. (via Predpol)


With budget cuts and shortages of officers, police departments are feeling the pressure to become smarter about preventing crime. One innovation born out of this pressure is PredPol, a newly developed “predictive policing” cloud-based software that uses past crime data and a fancy algorithm to predict where and when crimes are likely to occur.


Cofounders Jeff Brantingham and George Mohler were inspired to create PredPol after noticing that criminal activity follows surprisingly predictable patterns. They scrutinized the reported causes of archived crimes as well as their likelihood to trigger more crimes in the area. They then used these patterns to craft a mathematical model of criminal behavior, which became the PredPol algorithm. PredPol uses this algorithm along with adaptive computer learning to produce predictions they say are twice as accurate as predictions made using traditional practices.


PredPol uses three pieces of data—type, place, and time of crime—which it then feeds through its algorithm. Then, using 500 by 500 foot red boxes, it tells the officers which areas of the district are at the highest risk for criminal activity for that shift. PredPol suggests that officers patrol these red-boxed areas for about 5-15% of their shift, hopefully deterring crime with their presence. 


Departments that have tried PredPol, which now number 60, have had some success in reducing crime. For instance, Santa Cruz had an 11% reduction in burglaries and a 27% drop in robberies within the first year of employing the technology. Los Angeles and Atlanta, the largest police departments trying out PredPol, also saw significant reductions: Los Angeles’ Foothil Division saw a 13% decrease in crime after four months, and Atlanta saw a 19% decrease after integrating the software citywide. Although it’s impossible to know for sure that PredPol is responsible for these drops in crime, there certainly seems to be a correlation. To be more rigorous, co-founder Brantingham hopes to have PredPol’s six-month randomized controlled study in Los Angeles published in a peer-reviewed journal. This study found that PredPol was twice as effective in accurately predicting crimes than LAPD’s analysts.


Although its predictive powers are impressive, misgivings about the new technology are inevitable. The public, already suspicious of the police in the wake of recent displays of police brutality and racial profiling, will need to be fully informed about what exactly this technology is used for. Because the software does not take into account personal information about community members, says PredPol, personal freedom is not compromised and profiling is not a concern. Another worry, Louisiana State criminologist Peter Scharf points out, is that the red-box areas will be more feared by the officers, putting them on edge and making accidents more likely to occur.


Although harmful effects remain to be analyzed, the implications of PredPol software at this stage seem to be pretty good: Departments are able to use their resources more wisely; police officers can use their time on duty more effectively; new police officers may be trained more quickly; and communities can save the money they would have otherwise spent on catching, trying, and jailing criminals. Keep an eye on PredPol—it may be coming to your community soon.

It's interesting. I sometimes can predict gunshots in my neighborhood with high accuracy... Every weekend evening!


See more news at: