(B-57) Ethical Challenges of Artificial Intelligence in Dentistry

With the recent widespread access to large language models like ChatGPT, interest in Artificial Intelligence (AI) technologies has exploded. In dentistry, applications of AI are a natural progression of the digitalization of procedures and workflows ongoing in the profession. It cannot be denied that AI will have a place in the dental office of the future, but the nature of that role is still being discovered. During this unprecedented period of rapid innovation, it is important to consider ethical challenges associated with these new technologies. The American Dental Association principles of ethics and code of conduct list five values centeral to the practice of dentistry: patient autonomy, non-maleficence, beneficence, justice, and veracity. Through the lens of these core values, we will examine the potential pitfalls of AI technology in dentistry to help guide ethical adoption.

Preferred sub-tract: Machine Learning for Biomedical Applications

This study is based on analysis of publications concerning digital dentistry, artificial intelligence, and ethics in the field of medical dentistry. The methodological basis for identifying ethical pitfalls is the application of the 5 principles of ethics from the American Dental Association (patient autonomy, non-maleficence, beneficence, justice, and veracity) with special consideration for the environmental impact. AI technologies include, for the purposes of this study, those that assist in diagnosis, decision-making, treatment planning, and determining prognosis. The applications of AI in dentistry are a rapidly evolving field and it is impossible to address every unique application so we will focus on more common implementations like radiographic interpretation, restoration design, and treatment planning.

Specific pitfalls were identified and categorized based on ethical principles. Concerns related to patient autonomy include the privacy of patient data, the patient’s ability to give informed consent, and self-diagnosing with the help of AI resources. For the principle of non-maleficence, potential pitfalls include the environmental impact of training AI models, the environmental impact of equipment needed to implement AI technologies, and overdiagnosis from more accurate imaging or a reluctance to deviate from the recommendations of AI. The principle of beneficence includes pitfalls related to peer reviewed evidence lagging implementation of new technologies, the inclusion of ethics in research, and the amortization cost trap. Potential issues identified relating to the principle of justice include diffusion of responsibility from complex systems, accuracy of AI models when working with minority populations, and unequal implementation of AI technologies due to cost. The final principle of veracity addressed pitfalls of overreliance on AI by inexperienced clinicians and a shift in standards from the assumption that AI is more accurate than human judgement and wider spread adoption. While there appear to be many pitfalls related to implementation of AI in dentistry, the is a substantial benefit to advancing these technologies. Being aware of the social implications related to technological advancements can help us guide implementation into a more equitable and ethical future.