Could the pandemic have been predicted?

When the coro­n­avirus pan­dem­ic hit UK busi­ness­es in the spring, forc­ing organ­i­sa­tions to lock down, it required open minds to grasp tech­nol­o­gy and reimag­ine ways of work­ing. Gov­ern­ment and the pub­lic sec­tor sought to solve chal­lenges old and new, includ­ing rush­ing through essen­tial finan­cial sup­port to com­pa­nies and their fur­loughed staff, and improve ser­vice deliv­ery and data-dri­ven deci­sion-mak­ing by dialling up invest­ment in tech, espe­cial­ly arti­fi­cial intel­li­gence (AI).

After all, with pre­dic­tive ana­lyt­ics, gov­ern­ments can con­ceiv­ably pre­vent, rather than cure, issues or respond to cit­i­zens’ needs before they arise. But how far off are we from gov­ern­ing in advance? And what are the eth­i­cal impli­ca­tions of such a sys­tem?

Around the world, there are numer­ous nar­row-scope use cas­es of author­i­ties using pre­dic­tive ana­lyt­ics to life-sav­ing and life-enhanc­ing effect. In Durham, North Car­oli­na, the police depart­ment report­ed a 39 per cent drop in vio­lent crime from 2007 to 2014 after using AI to observe pat­terns and inter­re­la­tions in crim­i­nal activ­i­ties and to iden­ti­fy hotspots, thus enabling quick­er inter­ven­tions.

Also in the Unit­ed States, AI has helped reduce human traf­fick­ing by locat­ing and res­cu­ing thou­sands of vic­tims. Know­ing that approx­i­mate­ly 75 per cent of child traf­fick­ing involves online adver­tise­ments, the Defense Advanced Research Projects Agency devel­oped a plat­form using soft­ware that mon­i­tors sus­pi­cious online ads, detects code words, and infers con­nec­tions between them and traf­fick­ing rings.

Fur­ther afield, the Indone­sian gov­ern­ment has part­nered with a local tech start­up to bet­ter pre­dict nat­ur­al dis­as­ters. By analysing his­tor­i­cal flood data, col­lect­ed from sen­sors, and access­ing cit­i­zen-com­plaint data, prone areas can now be quick­ly iden­ti­fied, speed­ing up the emer­gency response and improv­ing man­age­ment.

Actionable intelligence and data scientists needed

In the UK, the pub­lic sec­tor has much work to do, and requires peo­ple to do it, if gov­ern­ing in advance is to become a real­i­ty, says David Shri­er, advis­er to the Euro­pean Par­lia­ment in the Cen­tre for AI. “More invest­ment in pre­dic­tive ana­lyt­ics will help with risk mit­i­ga­tion, although this exac­er­bates the already extant short­age of data sci­en­tists who can devel­op and man­age these mod­els.”

Pre­dict­ing trends through data analy­sis is vital for gov­ern­ments and has been for some time. “Fore­cast­ing approach­es using his­tor­i­cal data to build math­e­mat­i­cal pre­dic­tive mod­els have been core to gov­ern­ment eco­nom­ic pol­i­cy for decades,” says Andrew Hood, chief exec­u­tive of Edin­burgh-head­quar­tered ana­lyt­ics con­sul­tan­cy Lynch­pin. “Whether those mod­els allow gov­ern­ments to gov­ern in advance effec­tive­ly depends on to what extent they have enough polit­i­cal moti­va­tion and cap­i­tal to apply the mod­el out­puts direct­ly.

“Arguably, there has been no short­age of pre­dic­tive mod­els kick­ing around as the pan­dem­ic took hold. How­ev­er, the pan­dem­ic also points to the real­i­ty of a lot of pre­dic­tion and fore­cast­ing: it is not about hav­ing one crys­tal ball to rely on, rather a set of pre­dic­tions based on the best data to hand that need to be reviewed con­stant­ly, updat­ed and crit­i­cal­ly applied.”

Hood stress­es that skilled humans must remain in the dri­ving seat and warns of the dan­gers of sole­ly rely­ing on tech­nol­o­gy to steer choic­es. “As with any appli­ca­tion of pre­dic­tive ana­lyt­ics,” he says, “it is the inte­gra­tion of those mod­els with­in the con­text of gov­ern­ing and the process­es of human deci­sion-mak­ing that is the crit­i­cal suc­cess fac­tor.”

Public trust in AI must be won

Futur­ist Tom Cheesewright, whose job is to pre­dict trends, posits that pre­dic­tive ana­lyt­ics is “one sub­set of a wider array of fore­sight tools for scan­ning near and far hori­zons”. Should gov­ern­ments be mak­ing bet­ter use of such tools? “Absolute­ly,” he answers. “But I think it’s too tempt­ing with pre­dic­tive ana­lyt­ics to see this as a mag­i­cal answer, a black box that can solve all our chal­lenges. It’s not like Minor­i­ty Report-style pre­dic­tive jus­tice. It’s about pulling pol­i­cy levers in time to dodge obsta­cles or max­imise oppor­tu­ni­ties.”

Echo­ing Hood’s advice, Cheesewright adds: “Fore­sight needs time and invest­ment of cash and polit­i­cal cap­i­tal, both of which are in short sup­ply in our volatile, post-aus­ter­i­ty era.”

Nick McQuire, of spe­cial­ist tech­nol­o­gy mar­ket intel­li­gence and advi­so­ry firm CCS Insight, says: “His­tor­i­cal­ly, the pub­lic sec­tor has been behind most sec­tors in terms of matu­ri­ty in deploy­ing and invest­ing in AI,” but sens­es the purse strings are being loos­ened. “We are start­ing to see more AI appli­ca­tions in the pub­lic sec­tor: chat­bots, con­tact cen­tre assis­tance and demand fore­cast­ing,” says the senior vice pres­i­dent and head of enter­prise research.

AI has been exco­ri­at­ed in the UK media this year, though, mak­ing cit­i­zens and politi­cians wary of the tech and by exten­sion pre­dic­tive ana­lyt­ics. “Pub­lic con­fi­dence in AI is not high,” McQuire con­cedes. “To build trust in AI, organ­i­sa­tions are now hav­ing to dou­ble-down on areas like data gov­er­nance and secu­ri­ty, pri­va­cy, explain­abil­i­ty and ethics.”

It didn’t help that prime min­is­ter Boris John­son, the most pow­er­ful politi­cian in the UK, blamed the Ofqual exam-mark­ing fias­co in August on “a mutant algo­rithm”, says Dr Jeni Ten­ni­son, vice pres­i­dent and chief strat­e­gy advis­er at the Open Data Insti­tute. “We have to recog­nise peo­ple are at the heart of design­ing algo­rithms; it’s not that algo­rithms go off and mutate on their own and we have no con­trol over them,” she says. “We need to ensure there is a good end-to-end process that recog­nis­es the AI isn’t always going to get things right.”

Ten­ni­son, a fer­vent sup­port­er of open data, believes those in the pub­lic sec­tor must take care of how they deploy the tech­nol­o­gy. And, as such, pre­dic­tive ana­lyt­ics, if applied, should be close­ly man­aged. “Algo­rithms that are used by the pub­lic sec­tor have a much big­ger impact on people’s lives. Gov­ern­ment has a par­tic­u­lar respon­si­bil­i­ty to make sure it uses AI and data well,” she says.

“Right now we’re oper­at­ing from a posi­tion where peo­ple dis­trust the use of algo­rithms. The pub­lic sec­tor has to be very proac­tive and win that trust.”

Giv­en the pub­lic scep­ti­cism around AI, and the pauci­ty of data sci­en­tists to make best use of pre­dic­tive ana­lyt­ics, it seems we are some way off the UK gov­ern­ing in advance. Eth­i­cal­ly, per­haps that is no bad thing.