Decoding Legitimacy: Public Perceptions of AI-Assisted Decision-Making in the Norwegian Labour and Welfare Administration

Abstract

This master thesis addresses the emergence of artificial intelligence (AI) and automated decision-making (ADM) systems in public decision-making. The study’s main objective is to investigate to what extent the use of AI as decision support affects perceptions of legitimacy from a citizen’s perspective. Focusing on decisions made by the Norwegian Labour and Welfare Administration (NAV), this study explores the integration of AI as a decision support tool for NAV advisors in sick leave follow-up. While the project is currently on hold, the NAV case provides a credible example to investigate. To better understand perceptions of legitimacy, I conduct a Norwegian pre-registered, 3x2 between-subject survey experiment. The findings suggest that Norwegians perceive the decision-making process as less fair and are less willing to accept decisions made with AI assistance. Interestingly, there is a surprising interaction effect between the two key experimental treatments: when AI is used in the decision process, and the respondents receive a favorable outcome, this leads to a statistically significant drop in perceived legitimacy compared to when the respondents receive an unfavorable outcome. Nevertheless, much research still needs to be done before concluding how legitimacy can be maintained in the transition to increased use of AI in the public administration.