ShotSpotter Flex vs. ShotSpotter Connect
ShotSpotter Flex is a gunshot detection system that notifies the police department when and where gunfire occurs in the city. Currently, six square miles of Worcester is covered by ShotSpotter Flex.
ShotSpotter Connect uses crime data analysis to predict where crime will occur. Whereas ShotSpotter Flex reacts to events (like a gunshot, fireworks, or engine backfire) ShotSpotter Connect uses artificial intelligence to create what they claim are “high probability crime forecasts."1
The Cost of Predictive Policing (and at Whose Expense)
At the 1/19/2021 City Council meeting, Councilor Toomey and City Manager Augustus claimed they would have to spend “an additional $80,000” for an additional square mile of ShotSpotter coverage if the city rejected subscribing to the Connect program. This is misleading. The “free” mile of coverage valued at $80,000 would be given to Worcester in exchange for using ShotSpotter Connect, which costs $63,750 this first year. It is more accurate to say that rejecting Connect while still adopting the extra mile of Flex would only cost us an additional $16,250, not $80,000.1
ShotSpotter Flex and Connect services combined will cost nearly $200,000 per year.1
The money to pay for ShotSpotter Connect will come out of the Contingency Fund, which City Manager Ed Augustus set aside to save city jobs impacted by the ongoing pandemic and give pay raises to union and nonunion city employees.2
Geographic Data is Racially Biased
Several organizations and publications including the ACLU3, MIT’s Technology Review4, and various other independent researchers5 6 have criticized the safety and effectiveness of predictive policing, citing skewed policing data collected from the over-policed poorer communities and communities of color.7
City Councilors assure us that ShotSpotter Connect only targets locations and not individuals. However, many areas of Worcester are racially segregated, and ShotSpotter can target Black and brown folks simply by targeting the neighborhoods where they live. In 2020 the Fourth Circuit Court of Appeals found that “‘hot spot policing,’ which utilizes historical crime data to predict future crime hot spots, can be infected with years of racial bias…Predictive policing will inherently reinforce and perpetrate this structural racism.” Another judge on the same court criticized predictive policing technologies as “little more than racial profiling writ large."7
Racially Biased Data and the WPD
Past discrimination contaminates the historical crime data that ShotSpotter Connect would use to make predictions. There are many documented cases of racial discrimination within and by the WPD8 9 10.
Several councilors and Chief Deputy Saucier took issue with the suggestion that ShotSpotter Connect has the potential to lead to racially biased policing. City leadership’s continued reluctance to identify acts of racism and racial bias within Worcester, particularly within policing, does not inspire confidence that ShotSpotter Connect will be any different, and in fact causes grave concern that it will only exacerbate the issue.11
Predictive Policing’s Shaky Track Record
In Chief Sargent’s report to the City Manager, he listed 5 U.S. cities currently using ShotSpotter Connect. Not one of those cities has publicly stated they have seen a reduction in crime that can be directly attributed to ShotSpotter Connect, and one city even saw a 120% increase in homicides the year after Connect was implemented.13
Due to algorithms such as ShotSpotter Connect being owned by private companies – whose first and foremost interest is their own success and profitability – those who develop these programs often resist or refuse disclosure of their algorithmic policing tools, including information about the underlying algorithmic models, variables, data sets used, and testing that speaks to the tool’s level of reliability and accuracy.14
Currently, ShotSpotter offers next to no insight on how their Connect algorithm calculates “crime forecasts.” The company claims to incorporate “historical crime data,” “temporal cycles,” “proximity to known gathering locations,” and “upcoming events,” but does not specify how much weight is put on each of these data points when determining “hot spots” for officers to enter.14
According to the FBI’s crime data for murders in 2019, by far the most common relationship between offender and victim is acquaintance and by far the most common circumstance is an argument.15 When an argument between acquaintances will occur it is not something that can be predicted using Connect data. City Council stokes fear of gun violence to justify using ShotSpotter’s products, but the most common gun violence events are unpredictable.