A validated framework for responsible AI in healthcare autonomous systems

Area of Science:

• Healthcare technology
• Artificial Intelligence (AI) in medicine
• Clinical informatics

Background:

• AI-powered autonomous systems face adoption barriers in healthcare due to reliability and safety concerns.
• Existing conceptual work requires empirical validation for practical application in clinical and regulatory settings.

Purpose of the Study:

• To introduce and empirically validate a refined framework for the safe and responsible integration of AI in healthcare.
• To provide actionable guidance for clinicians, developers, regulators, and procurement bodies.

Main Methods:

• Developed an initial framework through semi-structured interviews with 15 experts from diverse domains (clinical, technical, ethical, regulatory).
• Validated the framework with 10 new participants using quantitative ratings and qualitative feedback.
• Ensured alignment with international standards like ISO 21448 and NIST AI Risk Management Framework.

Main Results:

• The refined framework achieved high scores for relevance, clarity, and usability during validation.
• Participants strongly endorsed the framework's practical utility for responsible AI adoption.
• The framework comprises ten dimensions across technical, ethical, and operational categories.

Conclusions:

• The validated framework offers a structured tool for evaluating and monitoring autonomous AI systems in healthcare.
• It addresses critical issues such as data quality, explainability, fairness, and human-AI collaboration.
• The framework facilitates the responsible integration of AI, overcoming adoption barriers in clinical and regulatory contexts.