8th Global Gastroenterology and Artificial Intelligence Summit | September 2026-2 -Roberts

Pdf Summary

The presentation argues that endoscopy AI should be designed as a <strong>Human-AI team</strong>, not a replacement for clinicians. Drawing on <strong>human factors research from transportation</strong>, especially aviation and automated driving, it explains that effective Human-AI systems depend on coordination, cooperation, interdependence, and resilience.

Key lessons from transportation include:
- <strong>Clarity and actionability</strong>: information must be easy to interpret and useful for decisions.
- <strong>Calibrated trust</strong>: users should understand when to rely on AI and when to question it.
- <strong>Context-specific modes</strong>: systems should make it obvious what mode they are in and what task they are supporting.
- <strong>Continuous human control</strong>: clinicians must retain the ability to override automation.
- <strong>Preventing skill degradation</strong>: training and performance testing are needed so specialists do not lose core skills as automation increases.

Applied to gastroenterology, these principles suggest design features such as:
- <strong>Hierarchical alerts</strong>
- <strong>Confidence tiers</strong>
- <strong>Task-based modes</strong>
- <strong>Manual override</strong>
- <strong>Sequenced training</strong>
- <strong>Regulatory benchmarks</strong>

The speaker emphasizes that technology development should involve practitioners from the beginning, not after the product is built. The message to industry is that great AI alone is not enough: if the interface is poor or the workflow is mismatched, the technology will not be used effectively.

Meta Tag

Concept Human-AI Teaming

Concept Human Factors

Concept Information Processing

Concept Out-of-the-Loop Risk

Concept Calibrated Trust

Keywords

endoscopy AI

human-AI team

human factors

aviation

automated driving

calibrated trust

manual override

hierarchical alerts

task-based modes

regulatory benchmarks

Human-AI Teaming

Human Factors

Information Processing

Out-of-the-Loop Risk

Calibrated Trust