EVALUATION OF READABILITY INDICES OF CHATGPT-4 AND GOOGLE GEMINI IN PATIENT EDUCATION ABOUT INTRACRANIAL HEMORRHAGES

Abstract

Aim. Subarachnoid haemorrhage (SAH) and intracranial aneurysms are critical neurological conditions with significant implications for patient morbidity and mortality. The intersection of readability indices and artificial intelligence (AI) is an emerging field that aims to improve the accessibility and understanding of written material in different areas. Readability indices, such as the Automated Readability Index (ARI) and the Flesch–Kincaid grade level, provide quantitative measures of text complexity that are crucial for tailoring content to specific audiences . Therefore, in this study, we aimed to examine the answers given to questions asked by patients with intracranial haemorrhage using readability indices. Materials and Methods. In this study, questions directly posed by patients and their relatives concerning subarachnoid haemorrhage and intracranial aneurysms were compiled. The collated questions were then divided into subcategories, including definition, diagnosis, treatment options, surgical procedures, complications, and impact on daily life. Flesch Reading Ease (FRE) Formula, Fog Scale (Gunning FOG Formula), SMOG Index, Automated Readability Index (ARI), Coleman-Liau Index, Linsear Write Formula, Dale-Chall Readability Score, Spache Readability Formula. AI technologies were compared across groups. Results. The results indicate that for most readability indices, there is no statistically significant difference between the two models, with one notable exception; Coleman-Liau Readability Index: Gemini: 10.59 ± 0.98 vs. ChatGPT: 11.80 ± 1.64; p-Value: 0.014 The only exception is the Coleman-Liau Readability Index, where a statistically significant difference was found, with ChatGPT showing a slightly higher score, implying potentially greater complexity according to that specific measure. Conclusion. Our article provides valuable quantitative data on the readability of texts from ChatGPT and Gemini, its scope is narrow. A more comprehensive study would ideally include qualitative assessments, a broader range of text types, and detailed information on the methodology and model versions to provide a more holistic understanding of the models' performance.