Artificiell intelligens och tjänstepersoners handlingsutrymme inom Skatteverket: En systematisk litteraturöversikt

Abstract

This thesis examines how artificial intelligence may affect the discretionary power of public officials at the Swedish Tax Agency. The study is a literature-based systematic review, using PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) as the reporting framework and thematic analysis as the analytical method. The review began with 104 records retrieved from Google Scholar using Boolean/SQL (Structured Query Language)-style search logic and was narrowed to a final corpus of 15 scholarly papers with verifiable DOIs (Digital Object Identifiers). The analysis shows that AI may improve administrative efficiency, consistency, and analytical capacity, but it may also shift discretion from caseworkers to data models, system designers, organizational rules, and oversight arrangements. The central argument is therefore not that discretion disappears when AI is introduced, but that it becomes redistributed, less visible, and more dependent on organizational governance. In the context of the Swedish Tax Agency, AI should consequently be implemented as accountable decision support rather than as an uncontrolled substitute for public decision-making.