The mayor and city council in the beachside city of Santa Cruz, California last week considered an ordinance that would put a stop to any predictive policing within its borders. With a rather modest-sized population of 65,000 and measuring only about 16 square miles in area (41 km-squared), the scale of this new push would not be especially significant but the move could trigger other cities throughout the USA to do likewise as California frequently acts as a bellwether for trends in governance.
To be clear, the city is not currently undertaking any predictive policing and had suspended such efforts in 2017 when there was a sense that the practice was no longer particularly effective and seemed to stoke adverse consequences such as continuing racial inequalities.
The difference now, in this latest consideration, is that the change would prohibit predictive policing entirely, removing its use from the police department’s options.
It is somewhat ironic that Santa Cruz was one of the first cities in the USA to adopt predictive policing. It did so in 2011,to great fanfare
, when it became an early adopter of automation provided by a local company called PredPol.
Predictive policing tends to spark controversy since it uses algorithms to analyse data and then proffers predictions about where crimes might be committed. Some liken this to the chilling pre-crime policing efforts fictionally depicted in the popular Tom Cruise movie Minority Report. The algorithms are often kept secret, there is little or no transparency in how the predictions are rendered and murkiness about how the police will end-up using those predictions.
PredPol emphasizes that their system uses only three pieces of data about crimes to make its predictions, consisting of a database of prior crimes and straightforward indicators of the type of crime, the location of the crime, and the date/time that the crime was committed.
Unlike other predictive policing automation, PredPol claims that they do not use arrest data
or criminal records databases, and prevent the algorithm from making predictions involving crimes that could “have the possibility of officer-initiated bias”.
Surprisingly, the proposed regulation is supported by PredPol, which points out that the ordinance would still leave open the possibility of a future resurrection of predictive policing and lays out strict guidelines if any such future efforts are launched. In short, PredPol indicated that they welcome setting a higher bar for exemptions and that they are ready to meet those conditions whenever a resumption is so considered.
A final vote by the city council will take place on August 11, 2020, including possible amendments before a vote for approval, and the ordinance could be purposely delayed or dropped altogether, rather than getting the go-ahead.
Reporting by Dr Lance Eliot, our US Associate Editor for Computers & Law