APPROPRIATE REJECTION OF INCORRECT AI ADVICE IN CLINICAL DECISION SUPPORT: A REVIEW AND SAFETY FRAMEWORK

Background: Explainable artificial intelligence is frequently presented as a route to transparent and safer clinical decision support. In high-risk decisions, increased trust is not equivalent to safer use. When artificial intelligence advice is incorrect, incomplete, or clinically under - specifi ed, a plausible explanation may convert error into an apparently acceptable recommendation. Materials and Methods: This narrative review synthesised DOI-indexed literature on clinical explainable artificial intelligence, clinical decision support, trust calibration, advice- taking, unsafe recommendat ions, false confirmation, overreliance, cognitive forcing, human-AI coll aboration, reporting standards, and health AI assurance. The analysis was organised around appropriate rejection: the clinicians capacity to reject incorrect AI advice while preserving appropriate reliance when AI advice is clinically valid.