Flight rules for clinical AI: lessons from aviation for human-AI collaboration in medicine

Area of Science:

• Medical technology
• Human-computer interaction
• Patient safety

Background:

• The aviation industry has valuable lessons from integrating automation, which improved safety but introduced new risks.
• Early automation in aviation led to over-reliance and unforeseen vulnerabilities, necessitating robust safety frameworks.
• Artificial intelligence (AI) is increasingly being adopted in clinical settings, mirroring aviation's automation trajectory.

Purpose of the Study:

• To draw parallels between aviation automation lessons and the integration of AI in medicine.
• To propose a framework for safe and effective AI implementation in healthcare.
• To advocate for a collaborative human-AI model in clinical practice.

Main Methods:

• This perspective draws on the experiences of clinicians and aviation safety experts.
• It analyzes historical data and safety frameworks from the aviation industry.
• It proposes a conceptual shift in how AI is perceived and utilized in medicine.

Main Results:

• Viewing AI as a "digital copilot" rather than an "autopilot" is crucial for safe integration.
• Practical considerations like scenario-based training and clinician benchmarking are essential.
• Minimum unaided practice is necessary to maintain core clinical skills.

Conclusions:

• The medical community must proactively learn from aviation's automation experiences to ensure patient safety.
• Adopting a human-AI collaboration model, supported by specific training and practice guidelines, will optimize AI's role in healthcare.
• This approach aims to enhance patient care by leveraging AI effectively while mitigating its inherent risks.