Evaluating of the readability of ChatGPT generated responses on travel health risks

Abstract

BACKGROUND: Impressive advances in artificial intelligence have given travellers a new source of health information. AI-powered tools, such as ChatGPT, allow users to obtain health information in a fast and accessible way. The aim of this study was to assess the readability of ChatGPT responses to questions about health risks when travelling. MATERIAL AND METHODS: Ten Questions about the health risks when travelling in the "Questions and Answers" section of the World Health Organisation's website have been asked 10 times to ChatGPT. A total of 100 answers was obtained and analyzed for readability. RESULTS: The mean ± SD of Flesch Reading Ease was 35.82 ± 6.46, Flesch-Kincaid grade level was 13.25 ± 1.45, Simple Measure of Gobbledygook was 12.34 ± 1.29, Gunning Fog Index was 13.77 ± 1.34, Coleman-Liau Index was 14.52 ± 1.09, Automated Readability Index was 14.93 ± 1.81. CONCLUSIONS: The readability of the answers produced by ChatGPT was 'difficult' and a college level education is required to understand the text. Lack of understanding of information can reduce the likelihood of travellers making good health decisions. To improve the understandability of ChatGPT responses, it may be useful to generate responses at a significantly lower reading level.

Ähnliche Arbeiten