Ansvar vid användning av AI-system : En kvalitativ studie om ansvarsgap mellan människor och organisationer vid användning av generativ AI

Sammanfattning

Generative artificial intelligence (AI) is increasingly used in organizational work processes such as information management, analysis, documentation, and decision support. At the same time, the use of generative AI creates new challenges related to responsibility, accountability, and the distribution of responsibility within organizations. Previous research shows that accountability gaps may emerge when multiple human and technological actors jointly influence work processes and decisions. The purpose of this study is to examine how accountability gaps are managed within organizations when generative AI is used in work processes. The study is based on a qualitative research approach in which eight semi-structured interviews were conducted with respondents from different organizations and professional roles. The empirical material was analyzed through a theory-driven thematic analysis based on sociotechnical theory and the concepts of responsibility, accountability, and accountability gaps. The findings show that organizations attempt to manage accountability gaps through guidelines, policies, training, and human oversight of AI-generated results. At the same time, the practical responsibility in the respondents' descriptions tends to be placed on the end user, despite the involvement of several organizational and technological actors. The study also shows that limited transparency and technical understanding make it more difficult to review and hold actors accountable for AI-generated outcomes. The study contributes with an empirical understanding of how accountability gaps emerge not only from technical complexity, but also from how organizations manage the use of generative AI in practice.