6th Global Gastroenterology and Artificial Intelligence Summit (Live) | September 2024-Balancing Human Expertise and AI in Gastroenterology

Pdf Summary

The lecture by Dr. Dennis L. Shung at Yale School of Medicine discusses the integration of Artificial Intelligence (AI) with human expertise in gastroenterology, highlighting both the potentials and challenges. One key point is the disparity between the efficacy of AI in controlled studies versus its real-world effectiveness. Randomized controlled trials (RCTs) show a strong relative risk (RR) but real-world implementation shows more variable results.<br /><br />A highlighted innovation is Computer-Aided Diagnosis (CADx) for automated optical biopsies. In trials, autonomous AI showed slightly higher diagnostic accuracy and better alignment with pathology-based surveillance recommendations compared to AI-assisted humans.<br /><br />The lecture also explores the "Paradox of Automation," cautioning against over-reliance on AI due to potential skill degradation and unfamiliar failure modes, exemplified by the Air France Flight 447 incident. This highlights the need for balanced human-computer interaction.<br /><br />Gastroenterologists' perceptions of AI-assisted systems show an increase in trust with experience, although they remain skeptical about AI predicting clinical outcomes. There is division about AI tools' impact on expert assessment. Interestingly, studies in other fields show "Bad AI" led to higher accuracy than "Good AI" or no AI, suggesting that experts can provide valuable input when they deviate from AI recommendations.<br /><br />Additionally, CADe systems in endoscopy could potentially reduce attentiveness among higher-performing endoscopists, who may relax on mucosal exposure with a sense of security that nothing is missed.<br /><br />Lastly, the concept of Generative AI for business problem-solving is discussed. AI was used to ideate and prioritize solutions in organizational settings, promoting AI from a tool to an active team member. This approach emphasizes tailored deployment, understanding team dynamics, and systematic usability, aiming for AI-enhanced human presence in clinical practice.<br /><br />The overall message encourages a future where AI augments human expertise rather than replacing it, urging mindfulness of automation's limitations and the need for robust human-AI collaboration.

Asset Subtitle

Dennis Shung, MD, MS, PhD

Keywords

Artificial Intelligence

Gastroenterology

Computer-Aided Diagnosis

Randomized Controlled Trials

Paradox of Automation

Human-Computer Interaction

Endoscopy

Generative AI

Clinical Practice

Human-AI Collaboration