Care to Explain? AI Explanation Types Differentially Impact Chest Radiograph Diagnostic Performance and Physician Trust in AI
PDF
Link
@article{prinster2024care,
title = {Care to explain? AI explanation types differentially impact chest radiograph diagnostic performance and physician trust in AI},
author = {Prinster, Drew and Mahmood, Amama and Saria, Suchi and Jeudy, Jean and Lin, Cheng Ting and Yi, Paul H and Huang, Chien-Ming},
journal = {Radiology},
volume = {313},
number = {2},
pages = {e233261},
year = {2024},
publisher = {Radiological Society of North America},
url = {https://pubs.rsna.org/doi/abs/10.1148/radiol.233261}
}
Citation
Prinster, D., Mahmood, A., Saria, S., Jeudy, J., Lin, C. T., Yi, P. H., & Huang, C.-M. (2024). Care to Explain? AI Explanation Types Differentially Impact Chest Radiograph Diagnostic Performance and Physician Trust in AI. Radiology, 313(2), e233261.
Informal Summary
A reader study with physicians examining how different types of AI explanation—local (feature-based) versus global (prototype-based)—affect chest radiograph diagnostic performance and trust in AI advice. Compared with global explanations, local explanations yielded better diagnostic accuracy when the AI advice was correct, and improved diagnostic efficiency overall by reducing time spent considering AI advice.