My Fault, Not AI’s Fault. Self-serving Bias Impacts Employees’ Attribution of AI Accountability

The widespread diffusion of Generative Artificial Intelligence (GAI)-based systems offers many opportunities, but it is also accompanied by public scandals causing harm to individuals, markets, and society. Practice and research are calling for ensuring AI accountability. However, accountability attributions are challenging because multiple actors are involved in the development, operation, and usage. It remains unclear who do employees hold accountable when using a GAI-based system. This study examined how employees attribute accountability based on their tendencies to enhance and protect their self-concept. We ran an experiment with 466 participants and compared their successful outcomes and failures when using GAI-based systems. Our experiment revealed that under success and failure scenarios, employees attribute accountability predominantly to themselves but also share attribution with other actors. Our study enriches understanding of employees' perceptions of accountability from a multi-actor’s perspective and aids in the establishment of accountability frameworks