Artificiell rättvisa? En experimentell studie på effekten av automatiserat beslutsfattande på proceduriell rättvisa

Abstract

We are currently facing a dramatic change in the public sector. Technological advances in recent years have made it possible for allowing algorithms and artificial intelligence to take over what used to be human tasks. For example an algorithm could soon be the one deciding whether or not someone is qualified for social assistance or financial aid. Needless to say this change from human to automated decision making is a change in procedures, and there is hence a compelling need to investigate how this in turn might affect trust and satisfaction among citizens. This study uses an experimental design with 187 respondents to examine the effect of automated decision making on procedural justice. The study was performed as a survey experiment and data was collected in Gothenburg, Sweden. Results of the study show a significant negative effect of automated decision making on the perceived ethicality and reliability, but a positive effect on the perceived efficiency and impartiality. Furthermore the study concluded that a decision made by an algorithm was perceived by the respondents as significantly less fair than a decision made by a social worker.