The Chatting Cure?: Ethical Considerations of Unsupervised Psychotherapy Provided by AI-Powered Chatbots

Abstract: In recent years, there has been remarkable development of conversational agents powered by large language models and artificial intelligence (AI). There is immense interest in harnessing this development within mental healthcare, especially given the shortage of mental health professionals and stigma around mental healthcare. While entrepreneurs and health systems hope that AI-powered chatbots can promote justice and beneficence by enhancing access to mental healthcare and promoting patient well-being, ethical questions abound regarding the use of chatbots to provide psychotherapy without clinician supervision.

The bedrock of psychotherapeutic work is a strong therapeutic alliance rooted in trust, protected by ethical boundaries, and utilized to serve a shared therapeutic goal between the clinician and individual. In this presentation, we argue that chatbots cannot, by design, forge a therapeutic alliance like a human. First, chatbots rely on continuous active inputs to generate an output for the user, such that the space for silence—a therapeutic tool used to convey empathy, facilitate reflection, and communicate presence—is lost. Second, the on-demand availability of chatbots encourages indefinite engagement regardless of clinical outcomes; yet, limitless engagement contradicts the ultimate goal of psychotherapy: for the individual to no longer need it.

Given the limitations of chatbots in forming the therapeutic alliance important for positive therapeutic outcomes, the expanding utilization of these technologies warrants more ethical scrutiny to ensure that patients who are already disadvantaged are not further subjected to suboptimal management modalities. Healthcare systems must refrain from assuming that chatbots can replace investment in the mental health workforce.