RE: Communicative competence of generative artificial intelligence in responding to patient queries about colorectal cancer surgery
Hinpetch Daungsupawong, Viroj Wiwanitkit
Abstract
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsArtificial Intelligence in Healthcare and Education
Dear Editor,
We would like to discuss material arising from this published article “Communicative competence of generative artificial intelligence in responding to patient queries about colorectal cancer surgery [1].” In summary, the study compared the responses from a colorectal cancer (CRC) information book to the response capabilities of generative artificial intelligence (GAI) tools when it came to answering questions concerning colorectal cancer surgery in Korean. The results showed that, for a number of evaluation criteria, GPT-4, Google Bard, and CLOVA X performed comparably to the CRC information book. The conclusions’ generalizability is called into question by a few constraints, including the CRC book’s perhaps out-of-date content and the small sample size of evaluators. Furthermore, the lack of a direct assessment of hallucinations in GAI responses in the study may have affected the veracity and correctness of the data presented.
Using a CRC information book from 2020, which might have featured out-of-date material, was one of the method’s shortcomings. This might have inflated the accuracy of GAI by distorting the comparison between the book and GAI responses. Furthermore, bias may have been introduced into the results due to the small number of assessors, particularly when they were divided into MDT and patient groups. Furthermore, the study did not specifically examine whether or not GAI responses included hallucinations, which is important to consider when assessing the accuracy of the data presented.
One possible query for the author would be whether they took into account the effect that out-of-date information in the CRC book would have on the comparison with GAI replies. It would also be helpful to know how the study took into consideration any biases that might have been caused by the limited sample size of assessors and the literary style differences between the GAI responses and the CRC book. Furthermore, the study skipped over what might be a crucial area for additional research: the possible effects of errors or hallucinations in GAI responses on patient decision-making and clinical results.
Uncovered topics in the global literature that require greater investigation include a thorough examination of how out-of-date data affects the comparison of GAI replies and the CRC book. Furthermore, it could be insightful to look into the long-term implications of using GAI for medical information in patient decision-making and clinical outcomes. Prospective avenues for investigation may comprise evaluating the efficacy of therapies aimed at mitigating hallucinations in GAI reactions and enhancing the precision and dependability of the data furnished to patients. Additionally, looking at how GAI functions in shared decision-making and patient education in healthcare settings may provide important insights into how it may be used to enhance patient outcomes.