Skip to main content

Table 2 Dos and don’ts while implementing AI-based prognostication models

From: Ethical considerations about artificial intelligence for prognostication in intensive care

Do Don’t
Define individual goals Rely on a single model
Clarify responsibilities Use models without testing
Choose a suitable model Make decisions in paternalistic ways
Assure technical robustness Be cavalier with human oversight
Seek maximum transparency Ignore data privacy issues
Quantify uncertainty Underestimate the role of empiricism
Share decision-making
Monitor performance and update model
Contribute to research and education