Dermatology in the wake of an AI revolution: Who gets a say?

To the Editor: We thank Ferreira and Lipoff for their response, which expands on salient ethical implications of artificial intelligence (AI) generated diagnoses in dermatology. Their insightful discussion raises additional ethical questions regarding patient autonomy, informed consent, and legal accountability as we look to a future in which AI and medicine coalesce. Since its release, ChatGPT ignited immense public and scholarly discourse around the potential integration of AI into medicine.1Dunn C. Hunter J. Steffes W. et al.Artificial intelligence-derived dermatology case reports are indistinguishable from those written by humans: a single-blinded observer study.J Am Acad Dermatol. 2023; Abstract Full Text Full Text PDF Scopus (3) Google Scholar,2Hirani R. Farabi B. Marmon S. Experimenting with ChatGPT: concerns for academic medicine.J Am Acad Dermatol. 2023; 89: e127-e129Abstract Full Text Full Text PDF PubMed Scopus (1) Google Scholar Despite varying physician and patient opinions, transformative AI advances are likely to become more numerous and clinically oriented in the coming years. Regarding patient autonomy, if patients object to the use of AI in the diagnostic process, a physician may simply agree to not utilize AI. However, if AI becomes a standard component of the clinical workflow, will patients truly be able to opt out of its use? The widespread transition to electronic health records, which store and may analyze personal health information, will potentially occur regardless of patient choice. Similarly, AI systems may be integrated without patient choice, leaving patients with the decision to engage with AI technology or not obtain care – a dilemma physicians ought to be wary of for the sake of patient autonomy. Training data sets for language models, such as ChatGPT, and pattern recognition-based artificial intelligence systems are vast and largely unknown to their users.3Murphree D.H. Puri P. Shamim H. et al.Deep learning for dermatologists: part I. Fundamental concepts.J Am Acad Dermatol. 2022; 87: 1343-1351Abstract Full Text Full Text PDF PubMed Scopus (17) Google Scholar With thousands to millions of images incorporated into training data, it is conceivable that many images, particularly if derived from the public domain, will be of individuals who did not consent to their inclusion.3Murphree D.H. Puri P. Shamim H. et al.Deep learning for dermatologists: part I. Fundamental concepts.J Am Acad Dermatol. 2022; 87: 1343-1351Abstract Full Text Full Text PDF PubMed Scopus (17) Google Scholar Considering the potential advantages, including monetary, of AI in dermatology, streamlining its development by circumventing the informed consent process may be attractive to developers. Therefore, the onus of advocating for the ethical development of training sets and protection of informed consent processes may fall on physicians, particularly those involved voluntarily or inadvertently in providing patient images for AI training. Who is at fault when AI technologies fail our patients? Today, if a patient suffers a poor outcome due to a dermatologist being misled by ChatGPT, it will likely be considered the fault of the dermatologist. However, the line of accountability may be blurred if AI integration into healthcare systems becomes standard rather than supplementary. AI pattern recognition technologies have demonstrated enhanced diagnostic accuracy relative to and when utilized by dermatologists.4Pham T.C. Luong C.M. Hoang V.D. Doucet A. AI outperformed every dermatologist in dermoscopic melanoma diagnosis, using an optimized deep-CNN architecture with custom mini-batch logic and loss function.Sci Rep. 2021; 1117485Crossref Scopus (25) Google Scholar,5Marchetti M.A. Liopyris K. Dusza S.W. et al.International Skin Imaging CollaborationComputer algorithms show potential for improving dermatologists' accuracy to diagnose cutaneous melanoma: results of the International Skin Imaging Collaboration 2017.J Am Acad Dermatol. 2020; 82: 622-627Abstract Full Text Full Text PDF PubMed Scopus (52) Google Scholar If AI utilization becomes the standard of care, is it the sole responsibility of the physician using the technology if they are misguided by the AI outputs? The role of AI developers and those encouraging use of AI, particularly for efficiency and financial gain, is not negligible. Clear legal guidelines for the use of AI in clinical practice will help prevent and better navigate malpractice litigation, particularly if the use of AI becomes less of a choice and more of an expectation in dermatology. In the wake of rapid AI expansion, physicians must consider the broader impacts of incorporating AI into clinical practice, ultimately remaining grounded in our ethical responsibilities to our patients and society at large. Jane Grant-Kels: DermaSensor: advisory board member. This company's specific technology is not discussed in this submission. Eric Beltrami has no conflicts of interest to declare. The complex ethics of applying ChatGPT and language model artificial intelligence in dermatologyJournal of the American Academy of DermatologyVol. 89Issue 4PreviewTo the Editor: We read with interest Beltrami and Grant-Kels’ article, “Consulting ChatGPT: Ethical dilemmas in language model artificial intelligence (AI),” a timely and evolving topic.1 This generative language technology may support precise and timely dermatologic diagnoses, but we must weigh ethical considerations. In addition to the ethical principles listed in this article, conversations about AI-generated diagnosis must address bias, informed consent, privacy and data security, and accountability. Full-Text PDF