Tool or Tutor? Experimental evidence from AI deployment in cancer diagnosis
Project Overview
The document explores the application of generative AI in education, specifically within medical training for lung cancer diagnosis. It underscores AI's dual function as a tutor and an operational tool, positing that its integration into educational frameworks can significantly enhance human capabilities while offering real-time support during practical tasks. Findings from a field experiment involving medical students reveal that the synergistic use of AI in both training and diagnostic applications leads to improved accuracy in diagnostic performance. The study stresses the necessity of retaining human involvement in professional activities, advocating for a model where AI serves to augment, rather than supplant, human expertise. Overall, the document presents generative AI as a valuable asset in educational settings, particularly in complex fields such as medicine, where it can foster both learning and practical skills development.
Key Applications
AI-assisted diagnosis system for lung cancer
Context: Training of second-year medical students at West China Hospital
Implementation: Participants received AI-integrated training and real-time AI assistance during diagnostic tests
Outcomes: Significantly improved diagnostic accuracy, with the highest performance in groups using both AI as a tutor and a tool
Challenges: Potential overreliance on AI and skill decay in tasks delegated to AI
Implementation Barriers
Skill decay
Human professionals may experience skill decay in tasks delegated to AI due to selective delegation.
Proposed Solutions: Integrating AI in both training and practice to maintain and enhance human skills.
Project Team
Vivianna Fang He
Researcher
Sihan Li
Researcher
Phanish Puranam
Researcher
Feng Lin
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Vivianna Fang He, Sihan Li, Phanish Puranam, Feng Lin
Source Publication: View Original PaperLink opens in a new window
Project Contact: Dr. Jianhua Yang
LLM Model Version: gpt-4o-mini-2024-07-18
Analysis Provider: Openai