Bortom det mänskliga omdömet : En kvalitativ studie om arbetstagares uppfattningar när autonoma AI-agenter beslutar om prestationsbaserad ersättning

Sammanfattning

As autonomous AI-agents become increasingly integrated into organizational decision- making, questions arise about how employees perceive AI-assessments, particularly when these systems determine performance-based compensation. The purpose of this study is to describe how employees perceive the use of autonomous AI-agents in decisions about performance-based compensation. The study adopts a qualitative research design based on semi-structured interviews combined with a vignette method. Data were collected from 16 full-time employees and analyzed through a deductive approach with inductive elements, drawing on attribution theory and organizational justice theory alongside a structured literature review. The findings reveal that employees perceive autonomous AI-agents in assessments as rationally acceptable, but relationally and emotionally insufficient. AI is perceived to enhance objectivity and reduce managerial bias, reflected in higher perceived distributive justice. These advantages are consistently counterbalanced by the absence of dialogue, empathy and contextual understanding. Interactional justice emerges as the most critical deficiency. Attribution processes are activated toward AI-agents. Positive decisions are attributed internally while negative decisions are attributed externally to the system's design, a pattern that shapes responsibility and controllability. Employees' perceptions are further moderated by the measurability of their work tasks, the experience of surveillance, the degree of human involvement, managerial role and level of AI-knowledge. The study contributes theoretically by demonstrating how the dimensions of attribution theory and organizational justice operate when applied to autonomous AI-agents as decision- makers. Practically, the findings highlight that transparency, meaningful human involvement and task characteristics are central to employee acceptance of autonomous AI-agents in decisions in a Swedish organizational context.