Responsible AI Governance in Organizational Decision-Making: A Conceptual Framework of Calibrated Trust, Ethical Acceptability, and Decision Legitimacy.

Abstract

By functioning in an advisory capacity, AI systems are playing an increasingly central role in organizational decision-making, yet their value cannot be assessed based solely on predictive performance. This article describes the governance-centered conceptual model of how technological governance capability and organizational governance capability jointly shape decision quality and decision legitimacy, both of which are distinct outcomes of AI-augmented decision-making. Our proposal not only considers calibration of trust in AI but also ethical acceptability (ecological validity) as parallel mechanisms rather than a fixed serial sequence. Calibrated trust is whether a decision-maker can count on AI at the level that it claims to perform, and ethical acceptability whether AI-supported decisions are justifiable in fairness, safety, privacy, human oversight, traceability & accountability. Learning culture is suggested as an organizational boundary condition by which the influence of organizational governance capability (with a positive sign) on calibrated trust and ethical acceptability is enhanced, while regulatory institutional pressure is proposed to strengthen (with a positive sign) the link between ethical acceptability and decision legitimacy. The framework also differentiates formative higher-order governance capabilities from evaluative mechanisms at the individual level and clarifies their levels of analysis. An example conceptual application to AI-enabled credit decisions is provided to demonstrate how the framework can be connected to verifiable governance evidence without claiming empirical validation. The model provides a unified explanation of the ways in which AI-governance capabilities can influence sound and institutionally justifiable organizational decisions across levels.