Ethical challenges of AI-based psychotherapy. The case of explainability

Abstract

This paper examines the reemergence of some of the traditional ethical issues of psychotherapy in therapeutic interfaces resorting to AI-driven conversational agents. I will begin by 1) proposing the extension of the operatory concept of explainability to encompass ethical problems related to the “in between” of therapeutic communication. Then, 2) I attend to the evolution of therapeutic chatbots, and how, as self-thematization media, they optimize an algorithmic resonance with the problems of their person of reference. For this, I argue, instead of resisting the contingency of the users’ inputs, chatbots rely on it to create new generative distinctions. I conclude that a consistent explainability of AI-driven chatbots needs to move beyond the clarification of algorithmic mechanisms to address the potential effects of this technology of self-thematization.