|
KMID : 0361020240670030146
|
|
Korean Journal of Otolaryngology - Head and Neck Surgery
2024 Volume.67 No. 3 p.146 ~ p.151
|
|
|
Can We Trust the Information on Allergic Rhinitis From Chat Generated Pre-Trained Transformer?
|
|
Lee Sang-Jae
Choi Yoon-Seok
Bae Chang-Hoon
Song Si-Youn
Kim Yong-Dae
Na Hyung-Gyun
|
|
|
Abstract
|
|
|
Background and Objectives Chat Generated Pre-Trained Transformer (ChatGPT) is alarge language model, which allows consumers to get information with one simple questionfor free. There are few studies that reported the reliability and usefullness of ChatGPT in thefield of otorhinolarynoglogy, so we would like to investigate the reliability of informationabout allergic rhinitis generated by ChatGPT.
Materials and Method We asked ChatGPT 35 questions related to allergic rhinitis based onthe textbook from Korean Society of Otorhinolaryngology-Head and Neck Surgery, the clini-cal guidelines of American Academy of Otolaryngology-Head and Neck Surgery, and theguidelines of Allergic Rhinitis and Its impact on Asthma, and recorded the generated answers.
Five specialists from our department were shown the answers in a blind test to assess their re-liability. Each question was rated either ‘accurate’ or ‘inaccurate,’ and if the contents were notfound in the textbook or the guideline, it was rated as ‘unreliable.’
Results Of the 35 questions, 26 (74%) were identified as ‘accurate,’ 9 (26%) were ‘inaccu-rate,’ and 0 (0%) were ‘unreliable.’ Questions about epidemiology, causes, diagnosis, andprognosis were found to be more accurate than the average, whereas definitions and treat-ments were less accurate than the average.
Conclusion The information about allergic rhinitis generated by ChatGPT was quite reli-able, showing that ChatGPT can be helpful in understanding and treating the disease. It isnecessary to use the developing medical artificial intelligence wisely.
|
|
|
KEYWORD
|
|
|
Allergy and immunology, Artificial intelligence, ChatGPT, Data accuracy, Rhinitis, allergic
|
|
|
FullTexts / Linksout information
|
|
|
|
|
|
Listed journal information
|
|
|
|
|