Algorithmic Accountability and the Use of Real-Time AI Tools: An Ethnography of Police Officers’ Daily Use of AI Violence Detection in Public Spaces

Algorithmic Accountability and the Use of Real-Time AI Tools: An Ethnography of Police Officers’ Daily Use of AI Violence Detection in Public Spaces Police organizations are increasingly deploying artificial intelligence (AI) tools to analyse data streams (e.g. audio, text and video images) in real time to improve their detection and response capabilities. This article explores how these AI tools introduce new challenges regarding algorithmic accountability at the level of individual police officers. It examines how digital surveillance officers from a Dutch regional police unit perceive their accountability regarding the use of an AI violence detection system. This ethnographic study, consisting of approximately 145 hours of observational research and eighteen semi-structured interviews, identifies three distinct AI working styles among digital surveillance officers, each reflecting different modes of interaction with the AI system. The findings indicate how surveillance officers feel only limited accountability to promptly respond to AI notifications and continue to prioritize requests and enquiries from other police colleagues and departments with whom they collaborate. This research argues that algorithmic accountability at the level of police officers should be explicitly integrated within existing internal police accountability frameworks to be effective.