Markus Langer

Increasingly, physicians must oversee AI systems to detect system malfunctions. We studied 48 physicians in a randomized controlled study using an increasingly malfunctioning AI for cardiac auscultation. Diagnostic performance declined alongside AI reliability, even when physicians were cognitively forced to make an initial classification, and when human performance exceeded AI reliability. Physicians’ confidence declined faster than their trust in AI. Results call for specification of oversight goals and processes ensuring patient safety.

AI tools including generative AI (genAI) and agentic AI promise individual productivity gains, and users experience them as a resource that makes them more capable. However, research documents a robust social evaluation penalty for individual genAI use: people who use genAI, or are suspected of doing so, are judged as less competent, less diligent, and less caring. We argue that these judgments, followed through to their behavioral consequences, outline a socially corrosive path of collective genAI use. Along this path, when people substitute human contact with genAI and devalue what colleagues produce, relational interdependence, metaknowledge, and resulting opportunities for scalable collective intelligence decline, whereas trust erodes through a cycle of concealment, suspicion, policing, and blame. Current technology design and productivity narratives amplify these dynamics. We caution against waiting for genAI use to normalize and propose countermeasures: designing genAI for collective functioning, resisting the productivity trap, and making the social norms of genAI use explicit.

As public administrations increasingly explore the use of artificial intelligence (AI), the emergence of AI agents poses novel challenges to regulation. AI agents are capable of autonomously (and often adaptively) planning and executing complex, mid-to long-term tasks involving sequences of actions. This paper examines the governance of such systems under the EU Artificial Intelligence Act (AI Act). Initially, we focus on the requirement of human oversight in Article 14, analysing possible tensions when applying it to AI agents. Their high level of autonomy raises the question of whether effective human oversight can be achieved in practice without eroding the efficiency gains that justify their use in the first place. Drawing on the principle of "supervisory control" from human factors engineering, we recommend considering human oversight in relation to the entire lifecycle of AI systems. This would enable Article 14 to be incorporated into the broader governance regime of the AI Act, comprising anticipatory, runtime, and evaluative functions. In doing so, we provide interpretive guidance on the overarching governance concepts embedded in the AI Act, extending beyond the conditions in Article 14, and outline a potential approach to applying human oversight to autonomous AI agents.

Human oversight is widely held to mitigate the risks of AI systems. Even for systems that produce discrete outputs at identifiable decision points, the realisation of human oversight as a reactive measure is empirically fragile, yet increasingly well understood. However, for agentic AI-systems that plan, decompose goals, and execute multi-step actions over extended horizons-reactive oversight reaches its structural limits: intervention on individual actions defeats the autonomy that motivates the deployment, while intervention on aggregate patterns is too coarse for harms whose cumulative consequences only become legible after the fact. This paper argues that reactive oversight must be complemented by an anticipatory mode: oversight exercised before the agent acts, by specifying the normative agenda that structures the space of permissible action and refining it iteratively through specification, runtime, and inspection. The two are complements-the agenda's escalation conditions specify when reactive intervention is invoked. Drawing on Meaningful Human Control, we read anticipatory oversight as the operationalisation of distal-reason tracking. In addition, we argue that the proposed framework yields a specific responsibility architecture by design: occupying the anticipatory mode is the discharge of a role-grounded prospective obligation, 1 and backward-looking responsibility takes the form of strict moral answerabilityrationalistic, relational, and holding regardless of fault, in virtue of the principal's prior opportunity for precaution. We develop bridging failure modes, address objections including moral luck and the illusion of control, and close with regulatory, architectural, and empirical implications

The use of Artificial Intelligence (AI) in high-risk, decision-making scenarios presents technical, safety, and normative challenges; problems that may only be ameliorated by human oversight. However, notions of human oversight lack a common foundational understanding: oversight architectures are not well defined, the roles involved remain unclear, and implementation steps are opaque. Hence, researchers and practitioners struggle to determine how to design, implement, and evaluate systems that enable effective human oversight. This paper advances a practical framework for effective human oversight of AI systems, based on a cross-disciplinary perspective that draws on insights from computer science, human-computer interaction, psychology, philosophy, and law. The core contributions are: (1) a foundational framework, with a working definition, architecture and processes for effective human oversight of AI systems; (2) an initial template for documenting oversight architectures and processes, applied to diverse domains; and (3) a synthesis of open research challenges that need to be considered in the emerging field of effective human oversight of AI systems.

Robinson et al. propose embedding AI proficiency into psychology curricula, mapping it onto established educational guidelines. This comment argues that the proposition could reach further and be more ambitious. I contend that psychology should prepare graduates not only as competent users of AI, but as active contributors to its design, evaluation, and governance. I emphasize three domains where psychology graduates who combine psychology proficiency and AI proficiency can make distinctive, interdisciplinary contributions: evaluating AI systems, designing human–AI interaction, and shaping AI governance and policy. Taking such an ambitious perspective will strengthen psychology's role in realizing human-centered AI.