Answers are grounded in this paper and cite their sources.
The increasing integration of generative artificial intelligence (AI) into human decision-making creates opportunities to extend analytical capabilities, explore alternatives, anticipate consequences, and improve the consistency of complex judgments. However, the availability of AI-generated recommendations does not necessarily produce better decisions. Effective collaboration requires an appropriate distribution of cognitive functions, calibrated reliance on AI outputs, independent human judgment, and explicit mechanisms for resolving disagreements. This conceptual paper introduces the Human–AI Joint Decision Architecture (HA-JDA), a proposed framework for organizing collaborative decision-making between humans and generative AI systems. The framework distinguishes eight interconnected stages: problem formulation, independent human assessment, evidence acquisition, alternative generation, scenario analysis, evaluative deliberation, decision authorization, and outcome monitoring. Each stage assigns different responsibilities to human and artificial intelligence according to their capabilities, limitations, and the normative requirements of the decision. The paper conceptualizes joint intelligence not as the replacement of human judgment or the mechanical aggregation of human and machine recommendations, but as a structured process of cognitive complementarity. Particular attention is given to epistemic independence, metacognitive regulation, uncertainty management, critical evaluation, and the preservation of human agency. An operational decision protocol, a functional responsibility matrix, and a preliminary research agenda are presented. The framework also addresses automation bias, algorithm aversion, erroneous forecasting, accountability diffusion, and the potential erosion of independent decision-making competence. The principal contribution is a testable architecture for integrating human judgment and AI-assisted analysis while maintaining human responsibility for value-sensitive decisions. The proposed approach aims to support more informed, transparent, adaptable, and accountable decisions without assuming that human–AI collaboration is inherently superior to independent human or machine performance
No takes yet. Share an insight, caveat, or question.
Roman Shaban (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: