Pre­dic­tive polic­ing and pre-crime’ algo­rithms: the new age of law enforcement

“When does prevention trickle so far up the chain that it slides into Minority Report territory and flip the presumption of innocence on its head?”

On Jan­u­ary 8, the aspir­ing box­er Jaden Mood­ie was knocked off a moped by a black Mer­cedes and stabbed to death in East Lon­don. He was 14. In March, the police were called to six stab­bings across the city in less than six hours. On May 2, Tashaun Jones, a fledg­ling music pro­duc­er and Arse­nal sup­port­er, was killed in Hack­ney. Three days lat­er, 18 year-old McCaulay Junior Urug­bezi-Edwards died in South­wark. Almost mid-way through 2019, and there have been 29 fatal stab­bings in Lon­don so far. In 2018, 135 peo­ple were unlaw­ful­ly killed in Lon­don, the high­est fig­ure in a decade. Of that num­ber, 76 were stabbed. 

Last Sep­tem­ber, Lon­don May­or Sadiq Khan announced his plan to try and com­bat the knife crime cri­sis: fun­nelling £500,000 toward cre­at­ing a Vio­lence Reduc­tion Unit (V.R.U). Adopt­ing the so-called pub­lic health approach – which piv­ots on the premise that vio­lence, like ill­ness, can be encour­aged or deterred by con­tex­tu­al fac­tors – the V.R.U. would twine the police’s work with hos­pi­tals, schools and local coun­cils to try and focus on ear­ly inter­ven­tion. The gov­ern­ment fol­lowed suit, estab­lish­ing a £200 mil­lion Youth Endow­ment fund to under­pin a push for pre­ven­tion over the next decade. 

While opin­ions about what, or who, is stok­ing London’s spike in vio­lence dif­fer (cuts to youth ser­vices; with­er­ing police bud­gets; orga­nized crime; drill music; gangs; too much focus on gangs), there has been a com­mon con­sen­sus that this approach, which weaves togeth­er dif­fer­ent parts of the state, is a sen­si­ble, sup­port­ive solu­tion. And why not? A ten­tac­u­lar focus on pre­ven­tion bur­rows beyond the bina­ry that peo­ple are good or bad, that crime is immutable; it implic­it­ly accepts that myr­i­ad caus­es under­pin some­thing as knot­ty as a climb in knife crime.

Besides, the pub­lic health approach comes test­ed. The V.R.U. is bor­rowed from Glas­gow, which has seen a 60% slide in homi­cides since a 2005 World Health Orga­ni­za­tion report named it the mur­der cap­i­tal” of Europe. Glas­gow, in turn, took the idea from Chica­go, where it was pio­neered by epi­demi­ol­o­gist Gary Slutkin in the mid-nineties. Return­ing home from Africa, where he had spent decades work­ing on infec­tious dis­eases, Slutkin found his home­town gripped by its own epi­dem­ic: vio­lent crime. Curi­ous, he start­ed dig­ging into the data and realised that, just as flu caus­es flu, the vio­lence was repli­cat­ing itself – passed from per­son to per­son. A depar­ture from main­stream think­ing, which focused on enforce­ment, Slutkin’s pub­lic health approach inte­grat­ed agen­cies beyond the police to try and stop vio­lence before it broke out, and cor­rall it once it had. In 2000, his pilot project launched in Chicago’s West Garfield neigh­bour­hood; with­in the first year, there was a 67% drop in homicides.

But, just because the arc of the pub­lic health approach comes proven, and feels pos­i­tive, doesn’t mean its appli­ca­tion in con­tem­po­rary Lon­don isn’t flecked with eth­i­cal pock­marks. Beyond its proac­tive topline, it’s not clear how it will work in prac­tice, or whether it will pro­tect those most vul­ner­a­ble to knife crime in Lon­don – young black males. The pub­lic health approach has become a bit of an emp­ty ves­sel into which politi­cians pour poli­cies they want to imple­ment,” says aca­d­e­m­ic Dr Adam Elliott-Cooper.

Pro­posed leg­is­la­tion could make doc­tors, teach­ers and nurs­es oblig­at­ed to report con­cerns about those deemed vul­ner­a­ble to becom­ing involved with knife crime.”

More clear is that a core ele­ment of this government’s pol­i­cy is data col­lec­tion and the umbrel­la shar­ing of that data. Indeed, pro­posed leg­is­la­tion could make front­line work­ers (doc­tors, teach­ers, nurs­es) oblig­at­ed to report con­cerns about those deemed vul­ner­a­ble to becom­ing involved with knife crime. In April, Home Sec­re­tary Sajid Javid – who has called for knife crime to be treat­ed like a dis­ease” – used a sem­i­nal speech on crime to advo­cate the pub­lic health approach, argu­ing the gov­ern­ment must use data more effec­tive­ly to trace the routes that lead into vio­lence. This is meant to help map the prob­lem, and involve agen­cies out­side the police. But, with­out ade­quate tech­nolo­gies and safe­guards, this gath­er­ing and dis­trib­ut­ing of data pos­es tan­gi­ble risks: rein­forc­ing bias and rup­tur­ing trust between author­i­ties and young people. 

For a pre­cau­tion­ary tale about the uneasy alliance between pre­ven­ta­tive polic­ing and data, look no fur­ther than the Gangs Matrix – wide­ly cast as a dis­crim­i­na­to­ry tool that oper­ates, hap­haz­ard­ly, in breach of data laws. Estab­lished in the wake of the Tot­ten­ham Riots, the Gangs Matrix con­tains the names and per­son­al details of rough­ly 3,000 peo­ple, and osten­si­bly helps the police track gang-affil­i­at­ed offend­ers in Lon­don, and exer­cise pow­ers like intel­li­gence-led stop and search. 

The process of being added to the data­base is fog­gy. Over­seen by the Tri­dent Gang Com­mand, it is man­aged local­ly in each of London’s 32 bor­oughs, who gath­er intel­li­gence thought to be sourced from vari­ables includ­ing social media posts, friend­ships, pre­vi­ous offences – even being a vic­tim of gang crime can count. This intel­li­gence is then entered into a for­mu­la that assigns so-called gang nom­i­nals’ a dan­ger code: red, amber and green, mark­ing the low­est like­li­hood for engag­ing with gang vio­lence. Just as there is no for­mal pro­ce­dure for being added to the Matrix, there is no offi­cial method for being noti­fied once you are on it, or for being tak­en off. 

More obvi­ous is that the make­up of the Matrix is biased. In 2017, an Amnesty Inter­na­tion­al report revealed that three-quar­ters of the indi­vid­u­als on it were black, although the Met’s fig­ures showed that only 27% of those respon­si­ble for seri­ous youth crime are black. The youngest mem­ber was 12-years-old. 5% of entrants had been cod­ed red. 64% were marked as green; many of that num­ber had no pre­vi­ous offences, and no known links to gangs.

The lack of due dili­gence in the devel­op­ment, imple­men­ta­tion and uti­liza­tion of the Matrix has meant that it is not an effec­tive, fair or account­able polic­ing asset,” says Kat­ri­na Ffrench, CEO of Stop­Watch. The Matrix is a racial­ized tool that lacks trans­paren­cy and clear pur­pose. Often the infor­ma­tion col­lect­ed and shared via this mul­ti-agency mech­a­nism is inac­cu­rate. There is a stereo­typ­ing of indi­vid­u­als by prac­ti­tion­ers and an over-polic­ing of black com­mu­ni­ties in Lon­don, which recent research has shown can push indi­vid­u­als to crime. This is very iron­ic as the Gangs Matrix was alleged­ly cre­at­ed to sup­port the police in reduc­ing gang crime.”

When does pre­ven­tion trick­le so far up the chain that it slides into Minor­i­ty Report ter­ri­to­ry and flip the pre­sump­tion of inno­cence on its head?”

The Gangs Matrix rais­es a ques­tion that inevitably crops up around the sub­ject of pre­ven­ta­tive polic­ing. When does pre­ven­tion trick­le so far up the chain that it slides into Minor­i­ty Report ter­ri­to­ry (Tom Cruise’s 2002 film that saw a Pre­Crime depart­ment arrest crim­i­nals based on pre­dic­tions) and flip the pre­sump­tion of inno­cence on its head? This is par­tic­u­lar­ly rel­e­vant in the con­text of some of the oth­er pre­ven­ta­tive tools rolled out by British police forces out­side Lon­don, which use algo­rithms so impen­e­tra­ble they are known as black boxes. 

In Feb­ru­ary, the human rights orga­ni­za­tion Lib­er­ty pub­lished a report show­ing that 14 forces are using, or have used, pre­dic­tive polic­ing” tech­niques, deployed with­out fixed legal guide­lines, or proof that they actu­al­ly work. The report broke these tech­niques into two cat­e­gories. First: pre­dic­tive map­ping. Kent Police used a machine-learn­ing algo­rithm devel­oped by pri­vate com­pa­ny Pred­Pol, which absorbed his­tor­i­cal crime data to antic­i­pate where crimes are like­ly to occur. Cit­ing a review of the system’s cost and effec­tive­ness, the force con­clud­ed its con­tract last March. The sec­ond cat­e­go­ry is indi­vid­ual risk assess­ment”, which pre­dicts how like­ly an indi­vid­ual is to com­mit a crime. For exam­ple, Durham police force use the machine learn­ing Harm Assess­ment Risk Tool (HART), which pre­dicts how like­ly an ex-offend­er is to com­mit a crime. HART has proved to be 62.8% accu­rate in its forecasts.

Pre­dic­tive polic­ing tools draw upon data which is intrin­si­cal­ly biased and feed it through com­plex algo­rithms to risk assess com­mu­ni­ties and indi­vid­ual peo­ple – allow­ing dis­crim­i­na­to­ry prac­tices to dri­ve future polic­ing,” says Han­nah Couch­man, author of the report. Mean­while, data­bas­es like the Gangs Matrix see police offi­cers use high­ly racial­ized cri­te­ria to iden­ti­fy peo­ple as gang mem­bers’. Each approach is designed to label peo­ple as pre-crim­i­nals’ based on crude pro­fil­ing, lead­ing to increased sur­veil­lance of peo­ple who will often already expe­ri­ence an unrea­son­able and intru­sive lev­el of polic­ing as they go about their day-to-day lives.” 

Do algo­rithms mag­ni­fy prej­u­dices, whilst scrub­bing them with the veneer of dig­i­tal neutrality?”

The impact of this type of pro­fil­ing is detailed in a Stop­Watch report writ­ten by Dr Patrick Williams, a senior lec­tur­er in Crim­i­nol­o­gy at Man­ches­ter Met­ro­pol­i­tan Uni­ver­si­ty, who inter­viewed 15 mem­bers of the Matrix to gauge its effect on their lives. “[S]eriously, I was get­ting stopped three times a week. There were times I got stopped three times a day,” says Andrew, one of the par­tic­i­pants (whose names have been changed). There was not a week I can real­ly remem­ber where I didn’t get stopped and searched. To the point where I realised it was not a thing any­more, it was just a normal.”

Bias in the jus­tice sys­tem is not new. But the con­cept of machines digest­ing exist­ing bias rais­es new con­cerns. Do algo­rithms mag­ni­fy prej­u­dices, whilst scrub­bing them with the veneer of dig­i­tal neu­tral­i­ty? And what hap­pens when their inscrutable pre­dic­tions are shared, and bounced between a web of agen­cies: edu­ca­tion, immi­gra­tion, hous­ing, health­care, the police? Whilst the Gangs Matrix is not a black box, it show­cas­es the spi­ralling impact that shar­ing dis­pro­por­tion­ate data can have. For those stuck in the Matrix, the impli­ca­tions reach far beyond addi­tion­al tar­get­ing by the police. It can affect their access to edu­ca­tion, hous­ing, jobs and can result in over-enforce­ment of stop and search and immi­gra­tion action,” says FFrench. For an exam­ple of the effect the label gang nom­i­nal’ can have, one need only look to Tot­ten­ham where those labelled gang nom­i­nals were sent a let­ter from the DVLA request­ing they take part in a drug test oth­er­wise their licens­es would be revoked.” Anoth­er, par­tic­u­lar­ly extreme exam­ple is described in the Amnesty report. One fam­i­ly was sent a let­ter threat­en­ing to evict them from their home unless their son stopped his involve­ment with gangs; he had been dead for over a year. 

As the Matrix’s flaws con­tin­ue to unfurl (in Decem­ber the Mayor’s Office gave the police a year to rad­i­cal­ly reform it) the Met­ro­pol­i­tan Police are devel­op­ing a new data­base, hailed as a cor­ner­stone of the pub­lic health approach. The Con­cern Hub aims to doc­u­ment indi­vid­u­als who are sus­cep­ti­ble to becom­ing involved with gangs. Offi­cial­ly pegged to launch in South East Lon­don last month, it was pilot­ed in Lewisham. Indi­vid­u­als iden­ti­fied as being at risk will be pro­vid­ed sup­port and path­ways away from vio­lence through part­ner­ship work­ing with local author­i­ties and a range of ini­tia­tives,” reads an emailed state­ment from the Met. 

Some are skep­ti­cal that local author­i­ties, hit by years worth of cuts, can help etch path­ways out of vio­lence – if not, the pub­lic health approach could be dom­i­nat­ed by the police (suf­fer­ing cuts of their own) and their troves of data rather than sub­stan­tial invest­ments in youth cen­tres and men­tal health ser­vices. We have seen tens of mil­lions of pounds of cuts to youth ser­vices across Eng­land and Wales since aus­ter­i­ty began,” says Elliott-Coop­er. All roads appear to lead to the police in one direc­tion rather than youth, health and com­mu­ni­ty ser­vices being joined up and work­ing togeth­er to address the problem.”

The thread that links data col­lec­tion, infor­ma­tion shar­ing, the loom­ing use of pre­dic­tive polic­ing tools and the impact of aus­ter­i­ty is trust. It is not news that the nar­ra­tive around knife crime is trained on young black men. As Matthew Ryder, for­mer deputy may­or for social inte­gra­tion, writes in a piece for Tor­toise, pub­lic health polic­ing, then, requires author­i­ties to build trust with their communities. 

There’s a real risk that pupils won’t want to con­fide in teach­ers or nurs­es if they know their dis­clo­sure could end up on a database.”

Cen­tral to this idea of trust – or the erod­ing of trust – are these plans to have teach­ers and doc­tors report data on knife crime; data that, in human terms, could com­prise dif­fi­cult, com­plex, inti­mate con­ver­sa­tions, or con­fes­sions. There’s a real risk that pupils won’t want to con­fide in teach­ers or nurs­es if they know their dis­clo­sure could end up on a data­base. Passed between agen­cies, pos­si­bly fed into an algo­rithm, it could be used to shape deci­sions about oth­er aspects of their, or their fam­i­lies, lives. If any­thing it actu­al­ly alien­ates young peo­ple from access­ing the kind of ser­vices that can help pre­vent them from being involved in vio­lent crime,” says Elliott-Cooper.

Teacher Rob Kazand­jian, who spent the first half of his career work­ing in a unit for teenagers expelled from main­stream edu­ca­tion, agrees. Pro­posed leg­is­la­tion will destroy trust between teach­ers and vul­ner­a­ble young peo­ple. Fac­tors that put chil­dren at-risk from per­pe­trat­ing vio­lence (pover­ty, expo­sure to domes­tic vio­lence, expul­sion from school) over­lap with those that put chil­dren at-risk from being vic­tims of vio­lence. A nice, neat line that sep­a­rates per­pe­tra­tors and vic­tims doesn’t exist,” he says.

Tak­ing steps to crim­i­nalise at-risk young peo­ple is immoral, in my opin­ion. Con­sid­er the stag­ger­ing sta­tis­tics that indi­cate prison is an almost inevitable out­come for young peo­ple who are expelled from school. Then con­sid­er that black boys are sig­nif­i­cant­ly more like­ly to be expelled from school than any oth­er group (the major­i­ty of young peo­ple in the unit where I taught were black boys, which says a lot about atti­tudes held towards black boys by the edu­ca­tion sys­tem). Giv­en that seri­ous youth vio­lence has been false­ly pre­sent­ed as an issue pre­dom­i­nant­ly affect­ing black boys, it’s quite easy to imag­ine what this pro­posed leg­is­la­tion will look like.”

00:00 / 00:00