-
1
Human-AI divergence visibility documents how clinicians and AI interact when their outputs differ, rather than treating the event as a simple choice to accept or override.
-
2
Divergence can arise from different information available to clinicians and AI, genuine differences in interpreting shared evidence, or an error by either party.
-
3
Clinicians contribute bedside examination, communication, and procedural information, whereas AI systems process structured records and patterns from prior cases or databases.
-
4
Technical explanations of an AI output do not reveal why the clinician’s judgment and the AI’s conclusion diverged, and post hoc explanations may foster false confidence in erroneous outputs.
-
5
The article argues that dedicated visibility metrics are needed to examine the human-AI interaction and identify what each contributed when their outputs diverged.
