Related Experiment Video
Updated: Apr 16, 2026
Three-Dimensional Finger Motion Tracking during Needling: A Solution for the Kinematic Analysis of Acupuncture Manipulation
Published on: October 28, 2021
A Comparative Analysis of Human and AI-Augmented Feedback in Anesthesiology Education
Daniel J Rosenkrans1, Katherine B Owensby1, Justin C Magin1
1The authors are in the Department of Anesthesiology, University of North Carolina School of Medicine in Chapel Hill, NC: is an assistant professor of anesthesiology; is a research assistant; is a medical student; is a medical student; is a professor of anesthesiology; is an associate professor of anesthesiology; is an associate professor of anesthesiology.
Background:
High-quality feedback is essential for resident development, yet barriers to its provision persist. Artificial intelligence (AI) offers a promising tool to augment feedback, potentially addressing these challenges. This study explored whether exposing attending anesthesiologists to AI-generated feedback samples before crafting their own feedback improved its quality as rated by residents.
Methods:
Thirty attending anesthesiologists provided feedback on 2 vignettes addressing issues of preparedness and professionalism. Before delivering feedback on the second vignette, attendings reviewed an AI-generated feedback sample. Utilizing a validated rubric, 6 blinded residents randomly evaluated the quality of feedback across 2 conditions: attendings alone (human-only) and after attendings reviewed the AI sample (AI-augmented). Feedback ratings and vignette types were compared. Residents were assessed if they could correctly identify the origin of feedback. A qualitative analysis explored attendings' perceptions of using AI for feedback.
Results:
AI augmentation did not significantly improve feedback quality ratings compared with human-only feedback (p = .7). Preparedness feedback was rated higher than professionalism feedback (p = .02). Residents could not reliably distinguish the use of AI for feedback (χ 2 = .92, p = .63). Attendings reported that AI provided helpful structure and phrasing, particularly for professionalism issues, and would use it if readily available.
Conclusions:
Although AI augmentation did not significantly improve feedback quality, it showed promise as a tool for supporting feedback provision, particularly with crafting feedback on nontechnical, more subjective issues. Additional studies are needed to better understand AI as a tool for feedback enhancement.
