När algoritmen får förtroende : En kvalitativ studie om rekryterares praktiker kring AI-verktyg och hur dessa kan påverka reproduktion och motverkning av bias

Sammanfattning

The use of generative AI has increased rapidly in society, particularly within organizations and recruitment, where AI-based recruitment tools are expected to enhance efficiency and objectivity. At the same time, concerns regarding both algorithmic and human bias have gained increasing attention. The aim of this study is to examine how recruiters perceive and use AI tools and how these perceptions and practices may contribute to the reproduction or mitigation of bias. Previous research highlights the presence of bias in both traditional and AI-based recruitment processes, the mechanisms underlying implicit and algorithmic bias, as well as issues of objectivity, trust, and mistrust in relation to AI tools. This study adopts a qualitative approach based on semi-structured interviews, which were analyzed thematically using Joan Acker’s theory of Inequality Regimes and Massimo Airoldi's theory of Machine Habitus as theoretical frameworks. The findings show that recruiters' experiences of AI tools, particularly experiences of trust or mistrust, influence how these AI-tools are used in practice. High levels of trust tend to reduce reflexivity, which may contribute to the reproduction of bias, whereas mistrust can enable practices that counteract bias. The study further demonstrates that AI tools are not inherently neutral, rather their outputs are shaped through the interaction between individual practices and organizational structures. The findings therefore highlight the importance of organizational responsibility, clear guidelines, and reflexive practices in reducing the risk of bias in AI-based recruitment.