Therapeutic Chatbots and the Alliance: Relational Agency in Algorithmically Mediated Care

The therapeutic alliance – the collaborative bond between patient and clinician – is the most reliable predictor of positive treatment outcomes (Lambert & Barley, 2001) and a paradigm case of relational agency: a capacity exercised between persons through the agreement of goals, the sharing of tasks, and the development of bonds (Bordin, 1979). This paper asks what becomes of that alliance when therapeutic chatbots and predictive systems enter mental healthcare not as instruments a clinician wields, but as participants in the therapeutic relationship itself. Drawing on work on predictive AI in psychiatry, the paper argues that these systems reconfigure relational agency along three connected axes. Through technological mediation (Verbeek, 2005), they co-shape how patients interpret their own mental states; through performative predictivity (Khosrowi et al., 2025), their forecasts – a predicted panic attack, a relapse-risk score – act back on the very behaviour they describe; and through epistemic injustice (Fricker, 2007), an algorithmic output can be granted a credibility that displaces the patient's own testimony. The deeper disruption is a shift from a relationship aimed at helping a person in the context of their life as a whole toward one organised around classification and prediction. The paper closes by asking under what conditions such systems might instead scaffold the alliance – bridging gaps in access and strengthening, rather than supplanting, the human bond.

Part of panel Disruptive Technology and Relational Agency: Therapy, Intimacy, Memory, Trust, and Individualization - part 1