Blogs

Anders Erikson's book Peak gives us deliberate practice as a very practical application of the way the brain learns. In picking up some of the loose threads from the book, this post will cover a bit of ground, sorry for that.

---

tldr; Neuroscience shows how the brain efficiently operates predominantly using predefined procedures. Erikson's book shows how these procedures can be reshaped using deliberate practice. The missing expert problem may be addressable using practice structures if we can find a way to refine and propagate them. Reflection is also reframed as a neuroplastic error detection device.

---

Intro

In the first part of my review of Anders Erikson's book Peak i opened up the topic of neuroplasticity as it relates to deliberate practice. It raised a few problems for our use case, chief among them the problem of the missing expert. But there is also a couple of other things he said that aren't clear to me. Especially his reference to 'mental representations', which i need to better understand. Here he is referring to the chess player that can instantly recognise thousands of board combinations, while having no better memory than anyone else, these shortcuts are somehow involved in the process of learning.

To understand all of this, i am going to need some help from the literature on neuroplasticity. BTW, plastic just means, that property of a material which is solid and holds its shape, but with the application of force changes its shape. In the brain, fittingly thats about learning constrained by some resistance to change. Another neuroplasticity concept we get to clarify along the way is, the brain as a prediction engine.

So before i can talk about reshaping relational behavior, i want to harvest some of what is known about the whole neuroplastic learning process.

What is procedural behavior?

Some key terms this blog has so far adopted are procedures, procedural learning, and procedural behavior. They give us a great navigational aid, by making it clear that will-power and cognition are not the most durable way to alter behavior. Cognition is far too slow for the speed that the world comes at us. Instead we can use our thinking mind more indirectly, for the specific purpose of first reshaping our procedures. The cynical conclusion is that we can mainly only change future behaviors, not present ones. Sad, right?

The slow cognitive mind can technically step in and try to interrupt a procedure after it has started. There are three provisos, firstly it has to catch the procedure in time, procedures begin execution at 150ms, while the cognitive override takes almost an entire second to begin. Secondly, either way it is a metabolically expensive and finite resource. Thirdly it makes no change to the pattern, so you will have to intervene in the behaviour every time. For these reasons reshaping our procedures wants to be a solid part of our change toolkit.

Procedural processing is fast, efficient and utterly dominates our life. So knowing something about it is a probably a good idea.

Alright, now we can dig in. Procedures function by drawing on procedural priors. Priors is the term for the maps that the brain has previously constructed about how the world works. They exist in the form of a neural network, which is just a fancy way of saying, a vast array of simple probabilities. Example: long, twisty brown things on the ground, are probably snakes.

Lets split what happens under the hood, into five easy to grasp steps:

1. Using sensory input information, the brain constantly pattern matches each current event against all its procedural priors. It quickly finds the nearest best match, regardless of how close the match or how confident it is. This is called prediction, but another name might be intuition. The prediction is an approximation, and necessarily imperfect.
2. Based on that prediction, the body immediately acts, automatically without recourse to cognition. This 'act first' approach kept us alive but results in lots of errors. As we know, this can cause some relational difficulty.
3. Next the brain calculates the difference between the expected result and the actual result. This is called prediction error. If the error is more than a certain threshold, it then incrementally updates the maps that comprise the relevant procedural priors. The update is a slight weight adjustment, not a complete rewrite. This is why it takes repeated exposure to new patterns, but over time it results in better future predictions. However small errors are ignored, so bad habits can still perpetuate.
4. After this all has taken place, the cortex eventually catches up and the brain's interpreter constructs a narrative. This comes after the fact and it basically tries to make the behavior make sense. The confabulated story is a fiction but feels like it caused the behavior. However it necessarily occurs after the behavior.
5. These are not seperate steps, its all one integral system that repeats very fast, over and over. And all except step 4 occurs below the level of awareness.

Understanding this makes it clear why we might have overestimated the impact of will-power and choice. The brain is a computer that works on auto pilot. The place we can make a difference is by reprogramming the computer, before hand. The matter of how, is the really interesting question, for another day.

Why this matters for relational learning

I'm now imagining relating as a very large set of procedures that we learned as children, and while growing up. If a person does X respond with Y, but they do M respond with N. By practicing new and inspired ways to respond, we are gradually building a new set of procedures.

Just to be clear, the prediction error isn't generated by a mismatch between the situation and the prior, but by a mismatch between the expected outcome of the action and the actual outcome. A full round trip, observe-react-observe loop has to complete before there's anything to learn. This makes the experience required. We learn by acting, and seeing what happens. The other person's response is a key part of the outcome information. In this way relationships are interesting, due to the richness and complexity of feedback. Is the feedback about you or about them. Is it suppressed, or over rendered? Is it accurate? It's not like target shooting where you either hit the target or don't. Its not like chess where you win or lose. In this way feedback only becomes more interesting for our application.

Relational practice is not about knowing about all the different kinds of behavior, or learning how to change behavior. Its about building a more capable muscle memory. This is primarily what supports new behavior.

But recall that learning is destabilising. Responding and behaving automatically from procedures is more comfortable, exactly because it doesn't require learning. But these are changing times so building capacity to learn seems rather sensible.

Our practice edges are exactly where all the prediction error occurs, and it is that which stimulates our growth and development. This reframes mistakes as an essential part of learning. Without the error there can be no rewiring. Its not, mistakes are OK, but its mistakes are needed. And for relationships, that means repair as well.

Feedback

Erikson distinguishes between high error visibility fields like athletics and music, where the gap between intended and actual output is immediately obvious. Low error visibility domains include teaching, management, therapy, and relating. Here the feedback is more messy.

This might be the biggest challenge for relational practice. The feedback is noisy and erratic. One answer is that the practice architecture therefore needs to increase feedback visibility. Some ways to do this include the simple practice of reflection, enhanced feedback processes, and slowing things down enough that the unnoticed becomes noticeable.

Its worth noting that interpersonal feedback occurs continuously, whether we know it or not. Multiple channels of information tumble at us, for instance verbal, nonverbal, tone, pitch, timing, facial response, posture and body movement. The information might be kind of encrypted, but theres a lot to see if you know how to look. Maybe the problem is that we were just never trained to read it. Slowing the process down might serve to locate more of the quiet richness of the feedback.

While practitioners may benefit from a benevolent external source pointing out the error, Erikson shows us that we can also learn to self detect the error. This develops as a result of the learning itself. But new practitioners probably can't reliably see their own errors.

Erikson's expert suffers all of the same sorts of problems that every human does, and isn't necessarily able to see things neutrally. Being victim to the same gamut of cognitive biases. Just like the rest of us. However where interpersonal feedback is currently weak, we can incrementally improve, as part of the overall practice of relational capacity building.

The feedback rich environment that is interpersonal relating can be kind of wild. In chess this would be like each game we decide on a brand new board layout, with different numbers and types of pieces. It would be like the rules changing every move, and the pieces randomly moving around when we aren't looking. Total mayhem, oh the joy of relating. I guess thats why relational practice in particular attempts to create some structure.

The fact that there is more than one player in the game also adds a non-intuitive and actually quite dramatic complexity problem. If a single person's complexity is N, then in a dyad that becomes N squared. For groups of N size, its gets worse, something like 2 to the power of N. This asks us to think twice before adding every extra person to our practice spaces.

Lastly Erikson's book largely presupposes the individual practitioner. Social Baseline Theory, and indeed Polyvagal Theory turn this around, and say that the default or normal situation is a social situation. So social feedback is something the brain has a ton of experience with, so much so that Robin Dunbar argued that its the very reason we have such big brains in the first place. Social baseline adds weight to our desire to improve feedback processes. Indeed relational learning is probably the singularly best placed kind of learning to actually iteratively learn better feedback, such that it scales and promotes emergent collective mind.

Reflection as error detection

That now brings us to reflection. Reflection, in addition to being a device that supports listening empathy, now has an additional role, that of error detection. You say XYZ, and i heard XYW. Theres two possible kinds of error here, as listener i mis-heard, and as speaker i was vague and imprecise. Both conditions are an error. Both are chances to learn. Oh i need to listen better, or, oh i need to be more clear. Either alone, or together, both seem like especially fruitful raw materials for learning, when seen in the light of neuroplasticity.

On the speakers part, they have a challenging task to compress the hugely complex world of their experience into language. I don't think we should underestimate how hard this is. Especially because most of the data of our experience is encoded as priors beneath the level of our awareness. Our words and sentences are themselves constructed procedurally, and it seems like something more akin to trying to describe a cloud or a dream.

The listener has to try to make sense of that stream of consciousness, encountering each sentence as something rather novel, something it has never heard in exactly this exact form before. All the while, having to pass everything through the listeners attention span, their own filters, biases, and relational priors. This is equally a miracle. Both processes are lossy and inaccurate.

The reflection presents the total accumulated transmission error for examination, and the practice partners get to iteratively approach that magical moment, when the speaker says i am complete, thanks for listening, i feel received. This comprises a matched pair of brains both learning at the same time. Quite interesting really.

When speaker and listener together reach that felt sense of, yes yes, that's it, you got it, that's no small thing. It might be one of the more amazing things that humans can learn to do. Over many iterations, our listening and reflecting procedures improve, layering skill on top of skill. Eventually constructing a shared mental model, and a shared vocabulary that points at shared truths. Truths about what it is to be human.

Lessons for the container

This model also adds a useful piece to our understanding of what practice structure is. By structure, i mean, as an example, the skill of reflection is part of the practice structure that was handed down to us. Part of the job of such structures is to present and highlight errors, in a domain where errors are less easy to see clearly.

In Erikson's framework, he considered the experts role as naming what to pay attention to, so the practitioner can build better procedures. If the practice structures are well designed, they can do the same thing. The practices, such as reflection, carry and hold the right conditions for neuroplastic change.

In this sense the practice traditions are a repository of wisdom. Utilising and reproducing the structures faithfully, serves both practice and institutional memory. Whether they are propagated and held only in practice, or also orally or in written form, it all supports an environment that in turns supports practitioners to reshape old relational procedures.

Structural evolution

The relational practice movement has proven capable of constructing and keeping alive its own defining structures. Taken further it suggests the potential for a pathway to develop this practice structure in an ongoing manner, incrementally over time. This is the practice of practice, and learning to learn. If the structure is to be a viable replacement for the expert, we are reliant on the structure being accumulated, handed down through the medium of practice, and then refined as capacity builds. This is a spiral that has to somehow escape the gravity well of the chicken and the egg. In such a way that accumulated wisdom can survive to become yet more accumulated wisdom.

The movement has no founder, and no founding text. It is a dispersed practice community driven by bursts of inspired passion and leadership from random people, that learned things through trial and error. A lot of error. They kept what worked, tossed out what didn't, and handed it all down not in writing, not even orally, but in the practices themselves. Whatever documentation of our models of practice might accrue, can only be an improvement, and one that serves that propagation and refinement.

The spiral requires enough structure to create the conditions for practice, enough capacity to use the structure well enough to generate useful feedback, and enough useful feedback to keep refining the structure. While early in the spiral the coherence and capacity of the structure may be low, however, any structure that supports even minimal forward movement will in time improve, and create unknowable things.

I see that as an expression of the growth imperative. And it takes us one step closer to a solution to the missing expert.