Anthony P. Weiss, Matcheri Keshvan, John Torous
The rapid uptake of AI for therapeutic uses has outpaced our ability to carefully understand risks and benefits, to assess and regulate its use. This is particularly true with AI-based mental health chatbots, with large percentages of the population globally now using these tools as an adjunct to, or replacement for, formal psychiatric care. While there is great potential in the use of this new tool, there are also clear risks. We use the Platonic concept of pharmakon to describe this, and go on to hypothesize about the three mechanisms which may be driving AI Chatbot use despite consequences. We believe this frame allows for testable research endeavors, and also provides a set of precedents related to the incorporation of previous technologies in the care of patients with mental illness.