Table A1. Forms of resistance in... | Emerald Publishing

Table A1.

Forms of resistance in health-care AI adoption

Dimension	Diagnostic question	Empirical indicators in meetings/interviews	Typical governance response
Conceptual proximity	Are actors framing the same clinical problem using comparable categories and objectives?	Disagreement over outcome definitions; competing interpretations of diagnostic goals; divergent understandings of what the model is optimizing	Shared glossaries; cross-specialty clarification sessions; alignment workshops
Representational proximity	Are clinical phenomena encoded and structured in comparable ways across datasets and models?	Disputes over variable inclusion/exclusion; inconsistencies in coding practices; concerns over missing values or cross-specialty comparability	Standardized data dictionaries; harmonization protocols; unified test datasets
Evidential proximity	Do actors converge on what counts as sufficient and legitimate evidence for action?	Skepticism toward statistical metrics (e.g. AUC) without clinical plausibility; demands for interpretability; requests for validation pathways	Clinical explanation rounds; agreed performance thresholds; retraining triggers and documentation routines

Dimension	Diagnostic question	Empirical indicators in meetings/interviews	Typical governance response
Conceptual proximity	Are actors framing the same clinical problem using comparable categories and objectives?	Disagreement over outcome definitions; competing interpretations of diagnostic goals; divergent understandings of what the model is optimizing	Shared glossaries; cross-specialty clarification sessions; alignment workshops
Representational proximity	Are clinical phenomena encoded and structured in comparable ways across datasets and models?	Disputes over variable inclusion/exclusion; inconsistencies in coding practices; concerns over missing values or cross-specialty comparability	Standardized data dictionaries; harmonization protocols; unified test datasets
Evidential proximity	Do actors converge on what counts as sufficient and legitimate evidence for action?	Skepticism toward statistical metrics (e.g. AUC) without clinical plausibility; demands for interpretability; requests for validation pathways	Clinical explanation rounds; agreed performance thresholds; retraining triggers and documentation routines

Source(s): Authors’ own work