Innov Clin Neurosci. 2024;21(4–6):8–10.
Funding/financial disclosures. The authors have no conflicts of interest relevant to the contents of this letter. No funding was received for the preparation of this letter.
Dear Editor:
It was with great pleasure that we read the correspondence regarding our article “ChatGPT and Transcranial Direct Current Stimulation for Chronic Pain.”1 The primary aim of the article was to initiate a discussion on the topic of artificial intelligence (AI) and clinical neuromodulation decision-making based on responses provided by chatbots.1 The authors discussed the outcomes generated by AI, highlighting concerns that the responses might not accurately reflect the current state of the art or industry best practices; therein lies the danger and the need for extra caution when seeking responses to clinical questions from this type of AI.
We selected ChatGPT as a representation of AI chatbots due to its popularity. According to a single search, it is possible to find 2,956 results on PubMed for ChatGPT and 124 results for Copilot at the time of composing this response letter. We emphasize that the responses generated by ChatGPT lack depth and have not been evaluated by experts. One concern regarding ChatGPT responses is the method used to find the answer; presumably, the program searches texts on the internet and calculates the probability of the best response.2 We acknowledge that our study did not include expert opinions or a thorough analysis of the AI-generated responses. We believe that expert opinions, supported by the best available evidence, are adequately represented in existing guidelines. However, we recognize that a qualitative analysis could enhance the robustness of our results. Currently, we are of the opinion that the integration of AI into clinical and scientific settings should be approached with caution. Although the AI-generated responses aligned closely with existing guidelines for the use of transcranial direct current stimulation (tDCS) in chronic pain, we emphasize the importance of adhering to guidelines and consensus statements validated by expert peers in clinical practice. As such, AI-generated results lack scientific validation, and they should not be considered for clinical decision-making.
The integration of AI into clinical contexts and decisions is already a reality in modern medicine.2 Researchers and clinicians should be discerning regarding the use of these technologies in their practice routines. It is time to discuss and propose a framework for the future clinical decision support system, including the utilization of AI. However, in due time, it might become challenging to distinguish whether a technology is employing AI for decision-making. Let us remain cautious and ultimately in control of the decisions. We appreciate the authors for sparking a scientific discussion on this current, high-interest topic in the field of science and neuromodulation.
References
- Silva-Filho E, Pegado R. ChatGPT and transcranial direct current stimulation for chronic pain. Innov Clin Neurosci. 2024;
21(1–3):61–62. - Ferdush J, Begum M, Hossain ST. ChatGPT and clinical decision support: scope, application, and limitations. Ann Biomed Eng. 2024;52(5):1119–1124.
With regards,
Edson Silva-Filho, PhD, and Rodrigo Pegado, PhD
Both authors are with Graduate Program in Heath Science, Graduate Program in Physical Therapy, Federal University of Rio Grande do Norte in Rio Grande do Norte, Brazil. Dr. Pegado is additionally with Graduate Program in Rehabilitation Science, Federal University of Rio Grande do Norte in Rio Grande do Norte, Brazil.