Establishing Responsibility Systems for AI Application in Clinical Decision-Making: A Relational Responsibility Perspective

Abstract

AI in clinical decision-making challenges liability frameworks grounded in subject–object dichotomies, linear causation, and atomised attribution. Drawing on relational responsibility, this article argues that the human–machine association, rather than the AI system or clinician alone, should be treated as the fundamental decision-making unit. Responsibility should therefore emphasise interaction quality, harm prevention, and relationships that sustain future cooperation. The article proposes dialogic responsibilities for technology providers, medical institutions, clinicians, and patients. For external compensation, medical institutions should act as the primary responsibility hub, reducing patients’ burden of proof while retaining rights of internal recourse based on specific human–machine interaction patterns. Internally, liability should reflect calibrated trust, documented dialogue, and the respective commitments of technology providers and clinical users. It thereby promotes accountable collaboration throughout the clinical AI lifecycle. By linking physician–patient trust with algorithmic trust, this relational framework seeks to reconcile technological innovation, professional autonomy, and the protection of patients’ rights.