Working alliance

In plain words

The working alliance is whether you and your therapist are working on the same thing, in a way you both accept, and whether you trust each other enough to do it. It is not friendship and not liking each other. Some alliances are warm and some are plain, and both can work.

Where the idea comes from

Freud treated the patient as a "collaborator" and linked the trusting element of transference to the basis for collaboration. Ego psychologists, including Sterba, Bibring, Zetzel (1956) and Greenson, shifted attention to the real relationship between analyst and patient. Edward Bordin's 1979 paper in Psychotherapy: Theory, Research & Practice (16(3), 252–260), "The generalizability of the psychoanalytic concept of the working alliance", recast it as transtheoretical, with three interdependent parts: agreement on goals, agreement on tasks, and an emotional bond. Bordin argued that it applies beyond psychotherapy to any helping relationship.

The Working Alliance Inventory, the Penn Helping Alliance Questionnaire and the California Psychotherapy Alliance Scales are among the measures used. A rupture in the alliance and its repair is treated by some authors as a central change process.

How it shows up

A strong alliance looks like a client who can say "this is not helping" and be heard, and a therapist who can adjust. A weak one looks like polite agreement with nothing changing. Researchers usually measure the alliance around the third or fourth session, and Wampold and Flückiger (2023) report that it predicts outcome even after controlling for early symptom change.

Evidence is substantial but modest in size. A meta-analysis by Flückiger and colleagues (2018) of 295 studies and more than 30,000 patients found a moderate association between alliance and outcome, around r = .28, and Wampold and Flückiger (2023) report that most of the benefit appears to come from the therapist's contribution. They caution that the evidence is largely correlational.

On TwentyThird, there is no therapist in the conversation with Oneli, the AI agent, so Bordin's concept applies fully only to a member's relationship with their psychologist, where one exists. Whether a member's psychologist is a good fit is the member's call, and they can leave if it is not.

What it is not

It is not the same as transference, though they interact. It is not a guarantee: a good alliance goes with better outcomes on average and does not show that the alliance caused them. It does not describe a relationship with an AI, and the research above concerns human clinicians. Studies of alliance with chatbots exist but were not reviewed for this entry.