From Computer Vision to the Vision of Law: Human Overs[a]ight 392

This paper reports on a translating practice between computer vision operations and the law, through the development of an art installation and research project, HUMAN OVERS[A]IGHT: THE OPS ROOM, presented at the Ars Electronica Festival 2025 in Linz. The installation is situated at the intersection of law and computer vision, questioning the legal understanding of the concepts underlying the development and deployment of high-risk AI systems. Under Article 14 of the European Artificial Intelligence Act (EU AI Act), the law requires adequate ‘human oversight’ in safety protocols for high-risk AI operations. However, the legislator seems to have merely addressed the potential complexities of interaction and human-machine operations. Article 14, suggesting a human-machine interface, a ‘stop button’, has opened for us a series of arguments and standpoints for investigating the scope of human agency, responsibility, and means of communication in decision-making processes, both machinic [automated] and human [cognitive]. In our research and in the artwork, we explore the complexities of interaction mechanisms that remain vague and underdeveloped in the law. We aim to expose potential frictions in translation between operational processes and their visual representation, as well as the limitations of social negotiation and contestation that the legislator appears to have overlooked. This research draws on the legal definitions of computation, reintroducing them into computer vision processes in the artistic representation of science fiction narratives, as invoked by the legislator through a trope such as a ‘stop button’ in a system of computational operations.