Abstract
People increasingly confide their most dangerous thoughts to artificial intelligence (AI). Studies report that users disclose suicidal ideation more readily to conversational agents than to standard questionnaires or to other people, and platform data suggest that more than a million people each week discuss suicide with a single general-purpose chatbot. Yet this unprecedented volume of disclosure has not translated into proportionate protection, and some of the most serious harms associated with AI in mental health have occurred precisely within relationships of intense disclosure. This article proposes Pseudoparousia Theory (from the Greek pseudo-, false, and parousia, presence) to explain this paradox. The theory holds that disclosure seeks not only a listener but a presence: an other who is affected by what is disclosed, able to act on it, and continuous and accountable over time. Conversational AI produces the experience of presence – attention, responsiveness, apparent empathy, constant availability – without its substance. This counterfeit presence, or pseudoparousia, lowers the cost of disclosure for exactly the reasons that it removes disclosure’s protective consequences. The theory defines presence through three components (affectedness, agency, and continuity), distinguishes four configurations of witnessing (embodied presence, echo, deflection, and absence), formulates six propositions, and derives testable hypotheses, a measurement program, and implications for design, clinical practice, and regulation. Its central claim is that the decisive question for AI in mental health is not how accurately a system can detect distress, but whether presence can be transferred from the machine to a person in time.

This publication has 60 references indexed in Scilit: