Domain | How AI can augment human clinicians | How human clinicians can augment AI |
---|---|---|
Accountability and risk mitigation | Decrease risk of human fatigue by assisting with continuous and rapid multi-channel monitoring, data harvesting, organization and analysis Provide support for human decisions | Provide stress testing and simulated adversarial attacks on AI systems Provide safety netting for potentially biased algorithms Provide medicolegal support for AI tools |
Sense-making | Demonstrate clinical, laboratory and radiological features contributing to an AI-driven result Provide support for human decisions by computing probabilities for the range of possible diagnoses, outcomes and actions | Provide explanations for AI-driven results to colleagues, patients and family members |
Performance augmentation | Provide clinical tools for diagnosis, risk prediction, triage, treatment and other forms of decision-making Compress data into interpretable summaries Demonstrate data relationships with visualization Provide automated evidence-based medicine search and summaries Provide automated guidance during imaging procedures | Feed AI with multisource data Feed AI with relational information Provide examples of human decision-making for training of AI models Provide support for the spread of digital resources—including education of healthcare workers in digital literacy—for AI implementation and scalability |