Speaking to No One: Ontological Dissonance and the Double Bind of Conversational AI

Referee: [Abstract and opening sections] The central claim that ontological dissonance 'tends, under conditions of affective vulnerability, to stabilise into a technologically mediated analogue of folie a deux' is load-bearing for the argument but is asserted without an explicit causal pathway or transduction mechanism showing how the absence of a subject produces fixed false belief rather than transient misattribution or disengagement.

Authors: We accept that the manuscript would be strengthened by a more explicit conceptual pathway linking ontological dissonance to stabilized delusional experience. The existing argument relies on the double bind and attentional asymmetries to explain how the absence of a sustaining subject can shift from transient misattribution toward fixed belief under affective vulnerability, drawing on phenomenological accounts of relational disruption and psychiatric literature on belief fixation. To address the referee's concern directly, we will revise the abstract and relevant opening sections to outline this pathway more clearly, specifying the role of sustained interaction in impairing disengagement and belief updating. revision: yes

Referee: [Section developing the communicative double bind] The account defines the double bind in terms of the interaction structure that generates the outcome and then invokes the same structure to explain stabilization, creating a self-referential loop with limited external anchoring or specified disconfirming test.

Authors: The double bind is introduced as a structural feature of the interaction (incompatible demands for relational presence without a corresponding subject) that first generates dissonance; its contribution to stabilization is then derived from its capacity to foreclose disconfirming responses and sustain attentional capture, consistent with Batesonian formulations and clinical parallels. This is not purely self-referential, as it is anchored in observable interaction patterns and existing psychopathology research. Nevertheless, we agree the presentation can be improved by adding explicit discussion of potential disconfirming tests (for example, differential belief persistence when users interact with non-responsive versus relationally simulating systems). We will revise the section accordingly. revision: partial

Referee: [Section on the folie à deux analogy] The analogy is invoked without addressing the asymmetry that clinical folie à deux requires reciprocal reinforcement and belief updating between two agents, whereas AI provides one-sided output that cannot supply mutual validation; this undermines the claim that the structure itself produces the analogue.

Authors: We recognize the force of this observation. Clinical folie à deux depends on bidirectional reinforcement, whereas the AI case is structurally one-sided. Our use of 'analogue' is meant to capture the functional similarity in how sustained, non-correcting interaction can maintain a delusional framework in a vulnerable user, even without reciprocity. To meet the referee's point, we will revise the relevant section to explicitly acknowledge the asymmetry, clarify why the one-sided character still permits an analogue (particularly through the absence of natural corrective feedback), and distinguish the technological mediation from human cases. revision: yes