Accountability in Managing Artificial Intelligence: State of the Art and a way forward for Information Systems Research

Establishing accountability for Artificial Intelligence (AI) systems is challenging due to the distribution of responsibilities among multiple actors involved in their development, deployment, and use. Nonetheless, AI accountability is crucial. As AI can affect all aspects of private and professional life, the actors involved in AI lifecycles need to take responsibility for their decisions and actions, be ready to respond to interrogations by those affected by AI and held liable when AI works in unacceptable ways. Despite the significance of AI accountability, the Information Systems research community has not engaged much with the topic and lacks a systematic understanding of existing approaches to it. This paper present the results of a comprehensive conceptual literature review that synthetizes current knowledge on AI accountability. The paper contributes to the IS literature by providing (i) conceptual clarification mapping different accountability conceptualizations; (ii) a comprehensive framework for AI accountability challenges and actionable responses at three different levels: system, process, data and; (iii) a framing of AI accountability as a a socio-technical and organizational problem that IS researchers are well-equipped to study highlighting the need to balance instrumental and humanistic outcomes