after two studies onna accuracy of prediction algorithms for th'risk of relapse for criminals, their advantages and weaknesses are discussed.
the occasion tis current investigation inna trade magazine sci advances, which attested structural defects to the older of the two studies. the test conditions were not realistic cause the influence of too few factors had been examined. the 1st study was in 2018 inna same trade magazine ∴ that laypeople ‘d predict th'risk offa crime returning as accurately as an algorithm used inna ∪d states for this purpose.
speshists johannes schmees and stephan dreyer from the leibniz institute for media research na hans bredow institute atta university of hamburg point out that both cases are experimental studies, the aim of which was to test a certain hypothesis. this means'dat they aint transferable to decision-making situations by judges in germany. christian djefall, professor of law, sci and tek atta teknical university of munich, also does not ponder the results to be relevant for criminal law practice. the reason for this s'dat the performance of teknical systems is compared with that of laypersons and not with that of trained judges.
original content at: www.deutschlandfunk.de/authors…: