Wasseem El Sarraj:

The contrast between them encapsulates a problem that therapy chatbots face. While they can imitate the language of therapy and reliably deliver a pastiche of validation, they cannot innovate an embodied response to you – a response born from being changed by what they witness. Consider the role of silence; while therapists sense what it holds, LLMs are not changed by silence. For them, silence isn’t training data.

During the therapeutic process, there is a kind of knowledge gained through unconscious mirroring of posture, experiencing shared emotions, imagining oneself in ‘another person’s shoes’. Feelings aren’t peripheral to this process; they’re central to it. This gets at one of psychotherapy’s oldest conundrums: is success the outcome of the quality of the relationship, or the inventory of the therapist’s skills? If therapy is effective through the transmission of information – asking the right questions, reframing the right issues – then machines will suffice. But if the relationship itself is therapeutic, if something emerges in the encounter that cannot be reduced to technique, then we are in a different realm.

it was in that gap between silence and presence where therapy nurtures. It does not offer information, but anchorage. The sense that you exist for someone else. That your suffering has been registered in another human being who will carry it forward and might be moved to act differently because of it. The witness doesn’t just hear your story; they become part of it.

Our societies have now been optimised for depersonalisation, yet they masquerade under the illusion of its opposite: personalisation.

Recognition is not the same as being witnessed.

In the therapeutic encounter, as in life, there are moments before we have linguistically categorised anything, anyone. Humans are landscapes to awe us; LLMs, however, are never speechless.