Performance of ChatGPT on the pharmacist licensing examination in Taiwan

Abstract

BACKGROUND: ChatGPT is an artificial intelligence model trained for conversations. ChatGPT has been widely applied in general medical education and cardiology, but its application in pharmacy has been lacking. This study examined the accuracy of ChatGPT on the Taiwanese Pharmacist Licensing Examination and investigated its potential role in pharmacy education. METHODS: ChatGPT was used on the first Taiwanese Pharmacist Licensing Examination in 2023 in Mandarin and English. The questions were entered manually one by one. Graphical questions, chemical formulae, and tables were excluded. Textual questions were scored according to the number of correct answers. Chart question scores were determined by multiplying the number and the correct rate of text questions. This study was conducted from March 5 to March 10, 2023, by using ChatGPT 3.5. RESULTS: The correct rate of ChatGPT in Chinese and English questions was 54.4% and 56.9% in the first stage, and 53.8% and 67.6% in the second stage. On the Chinese test, only pharmacology and pharmacochemistry sections received passing scores. The English test scores were higher than the Chinese test scores across all subjects and were significantly higher in dispensing pharmacy and clinical pharmacy as well as therapeutics. CONCLUSION: ChatGPT 3.5 failed the Taiwanese Pharmacist Licensing Examination. Although it is not able to pass the examination, it can be improved quickly through deep learning. It reminds us that we should not only use multiple-choice questions to assess a pharmacist's ability, but also use more variety of evaluations in the future. Pharmacy education should be changed in line with the examination, and students must be able to use AI technology for self-learning. More importantly, we need to help students develop humanistic qualities and strengthen their ability to interact with patients, so that they can become warm-hearted healthcare professionals.

Ähnliche Arbeiten