Why Psychotherapy Hallucinates? Therapy and Wellness Specific Dunning-Kruger Effect

Neuroscience unskilled and unaware of it. On why oversimplified narratives thrive in therapy and self-help worlds.

The Dunning-Kruger Effect

People who know very little about something often believe that they know more than they actually do (“It is not rocket science, it cannot be that complicated”), whereas experts are aware of what they don’t know, how much more there is yet to be discovered and the weaknesses in existing knowledge and theories. Sounds familiar?

That is, in a nutshell, the idea behind the Dunning-Kruger effect. It is a cognitive bias describing the tendency of “people with limited competence in a particular domainoverestimate their abilities”. The graph above is typically used to illustrate this effect. It features the “peak of oblivion of how little we know” - I call it the peak of hallucination, then the valley of despair - when you begin to understand certain bits and realise how desperately little you actually know (I find myself mostly in that valley). From there, the graph gradually rises to the levels of confidence of someone who is an expert in the field. Note that the peak of hallucination always exceedes level of confidence displayed by an expert.

Saying it is a cognitive bias is also saying that there is no shame to it: it is human and we all do it. But it is good thing to become aware of it. I reckon.

The Psychotherapy And Wellness Dunning-Kruger

Now, what does this Dunning–Kruger business have to do with therapy and wellness? As you know, I am a therapist. I am also well versed and immersed in the wellness world through meditation, breathwork, yoga and other trends that I am never beyond trying. However, I occasionally do get to talk to neuroscientists. I also make an effort to read neuroscience papers and to watch videos in which neuroscientists discuss their discoveries.

One thing always strikes me as I compare these two experiences: the contrast between the levels of confidence that therapists/ mindfulness teachers/yoga instructors might display when compared with someone who actually works in the field of neuroscience, when making some neuroscience-based claims.

Here are just a few examples, off the top of my head:

My yoga instructor, when talking about the benefits of breathwork telling us how breathwork helps in ‘getting out’ of the hippocampus: “We live in our hippocampus, that is literally where we spend most of our life,” she said, and then added, “It is located in the back of the skull,” while pointing to the back of her head.

One mindfulness teacher: “Every time we do the loving-kindness meditation we literally spray the amygdala with molecules that calm it down”.

A therapist confidently explaining how moving the lower lip in a certain way would block my fear of public speaking and my stress response through the activation of the vagus nerve.

Literally seems to be an often used word together with some very specific descriptions of ‘mechanisms’ and apparent absence of any doubt. At least, in my experience.

So, in essence, it is the Dunning-Kruger effect, version psychotherapy/wellness, wouldn't you agree?

So let’s have a closer look into this idea for some of the psychotherapy favs.

[N.B. What follows is a random selection and I do not have any particular bone of contention with any of the people who created the content.]

Attachment

Here is an illustration how scientists specialising in attachment might talk about neurobiology of attachment:

From A functional neuro-anatomical model of human attachment (NAMA)

… but the therapists know better and are like:

Can be found here.

There are many other overconfident statements (‘hallucinations’) around attachment. I will regale you with one more: “The anxious-preoccupied person has the brain of a drug addict (!!)”.

Can be found here.

But now, let’s move to another usual suspect for neuroscience oversimplification.

Emotions

Where are emotions based and what are emotions? While we might feel we all know what emotions are, defining emotions scientifically is notoriously difficult and neuroscience of emotions is still very much developing.

This is how a neuroscientist of emotions might talk about emotions:

From The Neuroscience of Emotion – A New Synthesis

Note, that very importantly, scientists - even the leading ones - allow themselves to say “we just don’t know”, demonstrating epistemic humility. But therapists might just go ‘Nah’:

Can be found here.

Before you decide to throw me to the wolves for criticising our profession once again, let me say this: I am actually not blaming therapists (meditation teachers/other wellness practitioners) for the state of affairs. Let me explain.

Psychotherapy And Wellness Hallucinate

A paper that tries to explain why AI hallucinates has just been published. What has come to be called an AI hallucination occurs when an AI system generates outputs that are incorrect, misleading or entirely fictional, whilst presenting them as factual. Sounds familiar?

When looking why this occurs it is my understanding that the authors came to the conclusion that the reason lies in the way the AI is rewarded when generating content.

Every AI model is trained and will create the content that gets the highest ranking by the evaluators (you can think of it as the biggest number of ‘Likes’👍). Now the models have been given a bigger reward for giving an answer (any answer) rather than saying simply ‘I don’t know’ (display of epistemic humility). Sounds familiar?

According to the authors, this process is to blame for the effect of AI ‘hallucinations’.

Now think of the whole therapy/wellness authors and content landscape as a set of agents who put out new ideas as a result of the feedback that us, the consumers of the content (you can think of us being evaluators, from the previous scenario) give them.

If we are consistently rewarding overconfidence over epistemic humility (saying ‘I simply don’t know’), gross oversimplification over embracing the complexity and nuance (e.g. attachment is only about [insert your favourite neurobollocks] ) , then:

• therapy authors/content creators who oversimplify neuroscience or downright ‘hallucinate’ will thrive

• those who are sitting on the fence will be more tempted to give in to this easy way of getting gratification and validation for their work

• those who go against the trend might just give up creating content altogether

  
  

And I guess, seen like that, it is pretty clear how we got to this mess, when it comes the use of neuroscience ideas in therapy world, right?

Who’s to Blame??

If you think I am going to bash the therapists for this, you will be disappointed. I would never kiss and tell like that. Also, I don’t think we should take the rap for everything.

To me, it is the Dodo-bird verdict situation:

• Therapy authors/mindfulness teachers/content creators: for not being prepared to have a deeper dive and embrace complexity of brain research

• Audience (us): for rewarding and favouring ridiculously simple explanations over complexity and epistemic humility

• Neuroscientists: for being tucked away in what sometimes seems like ivory towers. Not getting their hands dirty by directly engaging with the therapy/wellness folks in making their science more accessible (some notable exceptions, of course)

  
  

To return to the idea of AI hallucinations and the parallels with therapy ‘hallucinations’ - in relation to neuroscience explanations: I don’t think someone intentionally made a master plan to ‘dumb down’ psychotherapy and stop it from accessing real neuroscience. It seems to me that the whole thing is better understood as a kind of self-organising system. From individual behaviours that are understandable, such as the pursuit of more ‘Likes’, the sense of being overwhelmed by the complexity of actual science or the preference for oversimplified content whilst knowing it is not an accurate representation of the underlying science (I often hear: “It does not matter, it is only a metaphor anyway”), we end up with the neurobollocks slop, ‘hallucinations’ and Dunning-Kruger effect running the show.

But I am hopeful. Next time you encounter a yet another ‘revolutionary’ or ‘groundbreaking’ approach to trauma treatment/‘healing’ of attachment wounds/nervous system regulation to safety, see if you can change your ‘voting’ habit and reward the one who says ‘We just don’t know” or “the problem is complex and does not allow for one simple explanation'“.

Yes, it might increase the demands on our tolerance of ambiguity and uncertainty - but, as therapists, we are so good at this when it comes to accepting those same qualities in human nature (shadow and all), right? Maybe we can apply that same skill when learning about neuroscience for psychotherapy.

As always, thank you for reading 🙏.

---

Aknowledgments

While the content and the writing of this piece is mine my thinking about these topics forever evolves and matures through conversations with my friends and co-authors Dr. Chad Luke and Dr. Pascal Vrticka (sometimes I am not entirely sure which ideas are mine and which are theirs).