Response to: Investigating the impact of innovative <scp>AI</scp> chatbot on post‐pandemic medical education and clinical assistance: a comprehensive analysis

Abstract

We are grateful for the thoughtful commentary provided by Kleebayoon and Wiwanitkit on our recent article published in the ANZ Journal of Surgery.1 While we welcome this scholarly engagement, we find it imperative to address their concerns in order to elucidate the robustness of our study's design, methodology, and findings. First, the concern regarding the sample size and scope appears to overlook our study's qualitative nature, aimed at understanding the foundational capabilities of large language models (LLMs) in a controlled medical environment. It is important to underscore that our study serves as an exploratory assessment, where a large sample size is not the primary focus. Second, our study deliberately restricts its application to controlled settings as a foundational step, with future work intending to address real-world efficacy, as explicitly stated in our conclusions. Third, although our evaluation criteria focus on readability, reliability, and consistency with clinical guidelines, these are integral components that inherently contribute to clinical accuracy and patient safety, with ethical considerations forming an underlying theme. Fourth, the critique regarding potential bias in the evaluation process seems to underestimate the diversity of expertise among our panel of evaluators, which included three plastic surgeons and two junior doctors, thereby bringing multiple perspectives to the assessment. Fifth, while the study does not compare LLM performance to human expertise, it is not designed to propose LLMs as substitutes for human clinicians but rather as supplementary tools. Lastly, we acknowledge the ethical dimensions surrounding AI deployment in clinical settings and emphasize in our study the need for ongoing human supervision and algorithmic auditing to mitigate risks and biases. While LLM's have existed in the academic and scientific space for some time, it was the introduction of ChatGPT that captured the public's attention and imagination.2, 3 Since then, multiple papers have been published exploring the role of ChatGPT and other LLMs in the academic and clinical setting. Initial inquiries into the utility of LLMs have included topics such as medical education, clinical management, and scientific research.2-4 As interest in LLMs continued to grow, various other models have arisen, each with their advantages and drawbacks when compared with ChatGPT. Current literature demonstrates that LLMs show significant deficits in referencing, a tangible information inaccuracy rate, and are susceptible to bias. We agree with our readers that our study only reinforces these concerns, while the true extent to which LLMs can be developed into safe, ethical and clinically viable tools remains to be seen. However, only by asking the right questions and raising the relevant issue can we further drive research and lead to a greater understanding of this field. Consequently, we believe that our study offers valuable preliminary insights into the role of LLMs in medical education and clinical assistance. We advocate for a responsible integration of AI into clinical practice that adheres to stringent ethical and safety standards, and our paper mentions the need for future studies that scrutinize these ethical concerns in greater detail. Moreover, our study underscores the importance of interdisciplinary collaborations involving clinicians, data scientists, and ethicists to ensure that AI systems are both effective and ethical. Yi Xie: Conceptualization; project administration; writing – original draft; writing – review and editing. Ishith Seth: Conceptualization; investigation; writing – original draft; writing – review and editing. David J. Hunter-Smith: Supervision, writing - original draft; writing - review and editing. Marc A. Seifman: Supervision; writing – original draft; writing – review and editing. Warren M. Rozen: Supervision; writing – original draft; writing – review and editing.

Ähnliche Arbeiten