Stakeholder perceptions of clinical AI applications
| Positive perceptions | Negative perceptions |
| Clinicians | |
| Improved diagnostic accuracy; fewer errors3 5 More efficient work flows4 5 17 18 Less time spent on administrative and other mundane tasks3 13 Synthesis of clinical information15 18 Updating of clinical records14 More time spent with patients5 Improved access to care3 | Liability for AI-mediated errors3 Insufficient training and continuing professional development in AI3 5 7 8 12 Reputational loss and reduced demand for specialist opinion9 18 Potential erosion of empathetic communication with patients13 18 Risk of privacy breaches and loss of confidentiality of patient information17 Lack of proof of efficacy of AI applications in clinical settings3 29 Lack of explainability16 |
| Consumers | |
| Second opinions to clinicians21 22 25 Improved access to care23 | Dehumanisation of the clinician–patient relationship18 19 Threat to shared decision-making involving patients22 Low trustworthiness of AI advice19 20 23 Insufficient clinician and regulatory oversight21 Uncertainty around fairness and equity in treatment allocation26 |
| Healthcare executives | |
| Improved operational efficiency, cybersecurity, analytic capacity, cost savings27 | Uncertainty around patient satisfaction, access to care, improved patient outcomes27 |
| Industry professionals | |
| Shared many of the positive attitudes listed above27–29 | Limited access to high quality data for model development29 Unresolved legal liability question29 Lack of explicit and robust regulatory frameworks29 Low levels of funding for independent, investigator-led research in AI29 |
AI, artificial intelligence.