CPHAI 2026: Converging Perspectives on Health AI: Bridging Technical, Clinical, Sociolegal, and Ethical Research AIES Co-located Workshop Malmö, Sweden, October 15, 2026

Over the last decade, artificial intelligence (AI) systems have been increasingly deployed in the healthcare domain and have become commonplace within conversations around health. Our focus is on AI deployed in clinical and care settings - systems that enter diagnosis, treatment, and the work of clinicians, while remaining attentive to the wider conditions of health in which these systems operate. As AI systems rapidly reshape the clinical environment, meaningful impact, not merely good performance, becomes the critical concern. This has led to a growth in research pertaining to, among others, explainable and trustworthy AI. Nevertheless, many existing AI systems still fail to deliver on their promises in clinical workflows and other real-world use cases, often because legal, ethical, and epistemic challenges have not been integrated into their development and deployment.

Furthermore, AIES provides a groundwork for sharing research across computer and social sciences, law and policy, ethics, and philosophy. Within each of these domains, AI systems for health are developed and evaluated under different definitions of accountability, ethics, trustworthiness, safety, and alignment. The result is that each domain remains siloed from the others, with little articulation across the health AI field. Yet this articulation is essential for the development and evaluation of real-world AI systems. We count ourselves within this condition rather than outside it: as organizers drawn from computer science, social science, law, ethics, (and philosophy) we recognize that the meanings we attach to terms such as accountability, trust, safety, and alignment are themselves shaped by the disciplines that trained us. These are, in a precise sense, essentially contested concepts (Gallie, 1956), and the divergence among our usages is a structural feature of cross-disciplinary work rather than a failure of rigour. We do not assume that a single shared definition can or should be imposed. Instead, we offer this workshop as a trading zone (Galison & Panofsky, 1997) in which differing disciplinary understandings can be surfaced, held side by side, articulated. What we seek is the convergence of perspectives, understood as an achievement to be worked toward, not a premise to be assumed.

To this end, the Converging Perspectives on Health AI Workshop aims to provide a space for community building centered on the question: How can we develop clinical AI systems that are technically robust, clinically useful, socially acceptable, and responsibly deployed? Or, put simply, how should we be using AI in real-world clinical settings?

We especially encourage submissions that:

• are grounded in real clinical deployments, health-system settings, or concrete use cases

• connect normative or ethical claims to specific sociotechnical, organizational, legal, or regulatory contexts in health

• offer cross-national or comparative perspectives across health systems

• provide reflexive accounts of ethics-in-practice, from the clinic to the data pipeline

• speak clearly and substantively across disciplinary boundaries

Themes of Interest

The list is illustrative, not exhaustive. The groupings name shared problems, not disciplines; most contributions will speak to more than one, and we particularly welcome work that does.

Trust, Evidence, and Evaluation

• Defining, measuring, and contesting trust and trustworthiness across the fields that build and study clinical AI

• Evaluation frameworks beyond accuracy, i.e., safety, usability, robustness, and downstream clinical outcomes

• Communication of uncertainty, model limitations, and failure modes to clinicians, patients, and the public

• Standards of evidence for clinical AI, and what counts as proof of benefit across disciplines

Deployment, Translation, and Practice

• Translational gaps between research performance and real-world clinical use

• Clinician-centered design, human-AI interaction, and integration into clinical workflows

• Organizational, institutional, and cultural conditions shaping adoption, non-adoption, and resistance

• Lessons from other high-stakes domains (e.g., aviation, nuclear, finance) on safety and oversight in practice

Governance, Regulation, and Accountability

• Regulatory and governance frameworks for clinical AI — statutory regimes, soft law, and technical standards (e.g., GDPR, EU AI Act, EHDS, MDR, HIPAA, FAIR), including comparative and international perspectives

• Data governance: access, sharing, and reuse of health data across development and deployment

• Accountability, liability, and meaningful human oversight across the clinical AI value chain - developers, deployers, and clinicians

• Post-deployment harm, redress, and mechanisms for contesting automated decisions in care

Justice, Rights, and the Politics of Health AI

• Fundamental rights, non-discrimination, and the fair distribution of benefits and burdens

• Equity and access in clinical AI - fair performance across patient populations and the distribution of benefits and burdens in care

• Participatory and community-informed approaches to building and evaluating clinical AI

• Critique, refusal, and the limits of AI in health - where it should not be deployed at all

References

Gallie, W. B. (1955, January). Essentially contested concepts. In Proceedings of the Aristotelian society (Vol. 56, pp. 167-198). Aristotelian Society, Wiley.

Galison, P., & Panofsky, W. K. H. (1997). Image and logic: A material culture of microphysics.