Aca­demics query soft­ware use for risk as­sess­ment

Otago Daily Times - - Dunedin - JOHN GIBB

UNIVER­SITY of Otago re­searchers are warn­ing of po­ten­tial prob­lems in gov­ern­ment de­part­ments us­ing com­puter­based risk­pre­dic­tion mod­els, in­clud­ing a con­tro­ver­sial new tool used by ACC to pro­file clients.

As­so­ci­ate Prof James Ma­clau­rin, spokesman for the univer­sity’s Ar­ti­fi­cial In­tel­li­gence and Law in New Zealand Project, said the ACC used a com­puter model to as­sist staff man­ag­ing claims.

More details about how the sys­tem op­er­ated should be pro­vided, to en­sure it was used in a ‘‘fair and even­handed way’’, and was pro­vid­ing ‘‘good out­comes’’, he said.

ACC had re­cently in­di­cated the tool was be­ing used to make sev­eral kinds of pre­dic­tions, in­clud­ing how long ACC would ex­pect a claim to be man­aged.

This de­scrip­tion left open the pos­si­bil­ity that ACC used th­ese pre­dic­tions to min­imise treat­ment times, ei­ther by in­ter­ven­ing in pa­tients’ treat­ment, or ‘‘more se­ri­ously’’ by de­clin­ing ap­pli­cants with long pre­dicted treat­ment times, he said.

The tool makes pre­dic­tions about fu­ture ACC cases us­ing a data­base of in­for­ma­tion about 364,000 past claims lodged be­tween 2007 and 2013.

Fel­low project re­searcher Prof Alis­tair Knott, of the Otago com­puter sci­ence depart­ment, said this kind of pre­dic­tive sys­tem guided a cor­po­ra­tion whose de­ci­sions had ‘‘se­ri­ous con­se­quences for peo­ple’s lives’’.

But it was es­sen­tial that gov­ern­ments and com­pa­nies re­ly­ing on them could an­swer sev­eral ques­tions, in­clud­ing how ac­cu­rate the tool was and how it had been eval­u­ated.

The re­searchers also asked that if by ‘‘pass­ing the buck’’ to the ma­chine, ACC was also duck­ing its re­spon­si­bil­ity to make ‘‘fair and hu­mane de­ci­sions’’ about treat­ing peo­ple in need. They also asked if the tool im­plic­itly dis­crim­i­nated against in­di­vid­u­als on grounds such as age, eth­nic­ity or gen­der.

ACC’s use of ‘‘data and an­a­lyt­ics’’ told staff about how peo­ple were in­jured, their likely re­cov­ery time and how to help.

Since ACC started us­ing its pre­dic­tive tool, client sat­is­fac­tion rose from 68% in June 2013 to 78% to­day, and the av­er­age time to set up weekly pay­ments had dropped from 11 days in June 2014 to seven days, ACC said.

john.gibb@odt.co.nz

Newspapers in English

Newspapers from New Zealand

© PressReader. All rights reserved.