Decoding police algorithms in County Durham
By Sam Thomas
It’s 2054 and Tom Cruise captains a ‘Pre-Crime’ police unit using predictive technology to stop murderers before they act, reducing the murder rate to zero. As algorithms become embedded in daily life and policing, Steven Spielberg’s adaptation of Philip K Dick’s Novel The Minority Report seems more prescient than ever.
Now, in 2019, the next time someone steals a laptop from a kitchen table in the Viaduct or nicks a bike from outside Big Tesco, it could be their last time. Durham Constabulary is the first force in Britain to trial the use of algorithms to predict the likelihood of an individual reoffending, and then use that prediction to influence further police action.
Durham is at the national forefront of innovative policing
Durham is at the national forefront of innovative policing and they’ve had to be – local officer numbers are down 23% since 2010. A nationwide, austerity-driven, spending squeeze over the last six years has made police forces look to more efficient policing technologies including the use of predictive ‘AI’.
If you are arrested in Durham this year for whatever reason, then the judgement of whether you should be made part of Durham’s 4-month rehabilitation programme is not made by the officer you’ve been talking to but is influenced by the algorithm they have been feeding your data into. The result from the algorithm’s 34-variable analysis is supposedly legitimised through its use of Durham reoffender data from as far back as 2008. You are not being judged on who you are but what those similar to you have done in the past.
You are not being judged on who you are but what those similar to you have done in the past.
As a student living on the Bailey this year, I might therefore back myself to be released with a slap on the wrist simply because my data profile doesn’t fit the bill of a reoffender. However, for an unemployed 19-year-old from Bishop-Auckland (the most deprived ward in County Durham) the outcome of an arrest after a drunken night out is unlikely to be so generous – simply because of their similarities to other past reoffenders. You can see how ethical questions of using this technology start to arise.
Further still, one of the key arguments against similar predictive-policing technology has been the ethnic biases that have come out in trials across the USA. In one notable study it was found that African-Americans in custody were disproportionately given harsher punishment than any other ethnic group. America’s collaboration with private firms has also placed the algorithms used in the worryingly opaque territory of ‘corporate secrets’.
The above paints a bleak picture for algorithmic-policing. Some commentators might say that this is a classic neoliberal love-story: austerity meets public sector; public sector meets technology company; and the unconscious biases of detached data scientists end up dictating the lives of the masses. In contrast, I’m inclined to be more optimistic – with a caveat.
Police officers in both the UK and USA are inherently fallible to unconscious (and conscious) biases, especially around race. In the UK last year just 9% of new Police recruits identified as BME (Black and Ethnic Minority). The possibility for bias and impulsive human judgement in the context of a less diverse and overall smaller force remains high. Therefore, the use of adaptable, critique-able algorithms should be seen instead in terms of its possibility to iron out those biases and produce a more ethical, ‘blind’ judgement. But this takes time and mistakes will be made.
The possibility for bias and impulsive human judgement in the context of a less diverse and overall smaller force remains high.
The caveat then, at the crux of supporting algorithm-infused policing, is for the algorithms to be publicly accessible – only from this point can they be properly scrutinised. The Durham force’s algorithm trial uses publicly available data and is run in collaboration with University researchers who feedback on the ethics of the system. Durham sets a positive precedent for future algorithm initiatives in British policing where exposure to oversight and capacity for change should be top of the agenda.
The key revelation in Cruise’s Minority Report was that once people knew their future, they were able to change it. For algorithm-policing, people should know how they are being judged so they can have the chance to change it.
Photo: Jordi Cucurull via Flickr
