AI and therapy
“If you knew what I knew...” said a tech entrepreneur (not a client) to me recently “...You’d know that in five years time, AI will take your profession, too.” He was talking about artificial intelligence, a space he knew very well. I joked back, “Better get my skates on then”. I was dismissive and defensive at the same time, and then quite thoughtful. I have embraced technology all my life. As a therapist, for example, from being a sceptic on video vs in-person, video is now my preferred medium. It’s been a tech-fuelled journey. But, with AI, something didn’t feel right at all.
I put the case for AI to my teenage daughters. I thought they got to the nub of it. “If you know, as a client, you’re talking to AI, and not a person, isn’t that the key flaw?”. Nothing I’ve researched about AI and therapy since has changed my mind. Until AI can perceive or feel things, and until we know that AI is sentient, then it is just a therapy chatbot, that might be able to offer strategies and tactics but can’t ever hope to emulate our success with clients. To echo the US election messaging mantra “It’s the economy, stupid”, in therapy, it’s all about the therapeutic relationship.
I’m clear with clients, that, how effective our therapeutic relationship is, can be the key performance indicator for how effective the therapy will be. “If it’s not right bailout” is the stronger advice I give to friends and acquaintances when they ask for guidance in seeking a therapist. And, while it hurts my ego, when clients don’t persist with therapy with me, I accept, that it can be, that break-up cliche “it’s not you, it’s me”.
Having considered the importance of the therapeutic relationship between client and therapist I thought it might be useful to refer back to counselling theory. And see how AI might fare against theory. Carl Rogers triumphed with person-centred therapy (PCT) in developing a modality that now underpins most therapeutic approaches. I had a look at his “Six sufficient and necessary conditions” for effective therapy and assessed them in an AI context. If we do this, AI has a problem.
- Psychological contact between therapist and client. The problem for AI is there is no psychological contact with an AI tool, it does not have a mind.
- The client is incongruent. The client's state is unaffected. The client is incongruent.
- The counsellor is congruent. Can AI be congruent? It does not have a sense of self that could ever be in harmony. Or is AI fully functional in congruence terms? I think the former, congruence is linked directly to a consciousness of who we are. A state reached by a process, not a manufactured, out-of-the-box ready state.
- The counsellor shows unconditional positive regard (UPR) towards the client. This too is a problem for AI. This is one of the already heralded benefits of AI counselling. “No judgement, 24/7, and only £9.99 a month”. I think the power of receiving UPR is that it comes from a human. Humans do judge. A machine that we know can never judge is as deficient as a therapist that always does.
- The client receives empathy from the counsellor. An AI therapist cannot share our feelings. Nor can they demonstrate they do.
- The client perceives acceptance and unconditional positive regard. The client may perceive acceptance and UPR - to some degree, but this is the issue my daughters raised, who are they receiving it from?
In summary, I think evangelists for AI, and its use in many fields, are over-claiming. Is there a market for a text-based 24/7 responsive engine that delivers machine-learned, individual-specific, supportive and kind words, perhaps some tools and exercises, and asks us how we’re doing? Maybe. But that’s not therapy.
The AI evangelists underestimate three things. One is the power of the relationship, and how central that is to any therapeutic journey, irrespective of the type of therapy. Second, is an overestimation of the power of tools and techniques. AI can deliver them but the very fact that AI is delivering them undermines their effectiveness. Therapy works with the art and the science of being human.
Thirdly, I think there is a general underestimation of how effective therapy can be. The nail in the coffin of therapy, delivered by AI, could be how poorly it succeeds in the spaces in which we, as therapists, currently operate.
I hope that those seeking therapy by this route aren’t put off therapy by AI and fail to appreciate how life-changing a successful therapeutic journey, with a congruent, empathetic, non-judgmental human therapist, who has skills, tools and techniques, can be.