"Squeeze your glutes"
"Don't let your knees go over your toes"
"Push your bum back"
"Keep your back straight"
"Look up"

These are some of the most common phrases you'll hear most fitness trainers utter at a person whom they are trying to teach how to squat. But are these helpful? What if they were at best a waste of time, and at worst, sabotaging people's progress and a borderline danger to them?

As Fitness trainers, we are told from the start, our job is to teach our clients the right technique. Most people would agree with this. The idea of correct technique is everywhere from #formcheck to the highest levels of sports training. But what is technique? [Pause and think] [1]

Some common definitions of technique are:

a way of carrying out a particular task [Oxford]
a particular method of doing an activity [Collins]
a way of doing a thing that works better than others [my mom]

First, lets notice how all these definitions mention "A way" of doing something, whereas in physical training, technique is usually described as "THE way".

THE correct way to do a squat is to not let the knees cross the toes
THE correct way to lift something is with a straight back
THE correct way to do a push up is with elbows in/out

But is it? What does it mean to be "the correct way"? Firstly, correct for doing what? Maximizing performance? Mitigating injury? We all want to give our students what's best for them, right? So as coaches, we try to search for "the best way" of doing a thing. But did we ever stop to ask ourselves, is there a single best way? How do we know? [Pause and Think]

These are clearly important questions, and if you'd like to take a stab at answering them, read on. We'll try to find some answers through some very clever experiments in a field far outside of fitness. Let's start by asking ourselves some questions.

Question 1: Are there any single techniques that are 100% effective? [2]

[Pause and Think]

Taking our example of the squat cues again, it's quite clear from everyone's experience that no single cue/technique helps a person squat 'perfectly'. I don't think anyone would contest this. So then we have to look at the next possibility.

Question 2: Are there any single techniques that are significantly more effective than others?

[Pause and Think]

While we're all inclined to think of our personal favourite techniques as more effective than others, we have to ask ourselves two important questions about those:

Question 2a: Is the technique I have in mind statistically more effective that others? ['statistically' : holds true when applied to a larger sample of people and independent from person of delivery (you)]

Question 2b: Are there other different techniques that may be similarly effective? [Pause and Think]

Often we're inclined to see the techniques we personally use as more effective because A) we're all biased whether we like it or not and B) because we may only (or preferentially) use and teach the techniques we like, and if we see success from it, then assume that this singular technique could be the effective one. And without trying or teaching other techniques, we may never know or acknowledge how effective they are in comparison.

Another simple argument that can be made to answer Question 2 is that IF there was a technique that was undeniably better, it would be used extremely commonly out of pure selection pressure. But we don't really see that happening, so we can assume that no single technique like this exists.

I know that it may seem a little strange and disheartening to see the things that we used to help people being regarded as not particularly effective, but if you read on, you'll see why there may still be plenty of hope.

Question 3: How does the effectiveness of that technique change over time?

Initially, an effective technique may help improve performance significantly, but soon, as we've all experienced as coaches, it doesn't remain as useful as it did on day 1. Either the student understood it and imbibed it as much as they could, or they didn't and don't find themselves able to.

Here's what that would look like on a graph

Now we're going to ask a bit of a difficult question.

Question 4: What is the effectiveness of no technique?

By no technique, I mean telling them what the task is, but not HOW to do it. For example - the squat, the task would be to sit down and get back up, without losing balance from the feet. For a deadlift, the task would simply be to pick the object up (also, without losing balance). How effective would that be? To rephrase this question, how good would the people who received no technique be, in comparison to the people that did? [Pause and Think]

I'll admit, it's hard to just think of an answer to that question. The better way to find out would be to actually do this as an experiment and find out some more real answers. If you're a coach with access to students, I would highly encourage that you try it, just to see what happens. It can be a bit daunting at first to not give any instructions (most students basically expect it at this point), but if you can both overcome the initial apprehension, something interesting may come of it. But it's also clear that most people can't do this on a large enough scale to yield really broad answers. So what do we do?

Here's where I want to take you on a journey outside the world of fitness - and into artificial intelligence research. Why? Because artificial intelligence research is all about trying to teach computers to do what humans do. And maybe by observing how they do it, we can learn a thing or two ourselves. (You don't need any background knowledge in the area of course, I'll try to keep it as simple as possible)

Artificial Intelligence

First, artificial intelligence is nothing mystical. It is essentially a computer program. All programs are just machines following instructions. Artificial intelligence programs called this because they try to exhibit 'human-like' capabilities (recognising something from a picture, understanding speech or making decisions). We're going to look at a sub-field of artificial intelligence called 'machine learning' - the field which gives computers the ability to learn by themselves without being exactly programmed.

Regular programs are created with detailed instructions about what to do in all the situations they encounter. Your web browser will show you what you type in, your text messaging app will send the message you type to the person you want, etc. And most importantly, those instructions can't be changed by the program itself. Sometimes circumstances do come up where the program ends up stuck doing something wrong, but it is helpless there and can't do much about it, even if it reaches the same state again.

This works well enough for most things that we use our computers for these days, but there are some tasks that are practically impossible for us to create precise instructions for: recognizing faces, having a conversation, or driving a vehicle. The tasks are too complex, with too many different situations and possibilities.

So how can humans do all these things? This is the beautiful process that we call LEARNING. Somehow, humans that are born with no skills (think of how useless babies are) can develop into adults that are able to competently do all these things - and more. So, computer researchers tried to take this idea over to the computer programs. Just like us, the computer starts out with the task not knowing how to do it, but through trial and error and improving from those trials and errors, it gets better at the task, eventually good enough to be considered competent (the technical term for this is called reinforcement learning).

Now let's take an example of computers learning to do something this way. I'm going to choose the game of chess, because computers have been playing chess for a long period of time, and in many different ways. We can also compare the efficacy of each of those ways very easily - by making them play against each other. Most people have heard of some computers that play chess as well. The 1997 match between reigning human world champion at the time Garry Kasparov vs IBM's Deep Blue was the first time a computer defeated the strongest human under tournament conditions, and they've been getting better ever since.

The interesting thing is, computers were first taught to play chess in a very technical way. Top human players would basically encode strategies that they had developed, when and how to use them. The computers would then search through enormous possibilities to find the best moves, but using the information provided by their human trainers to make decisions. Furthermore, these programs, can only specifically play chess. To try to play another game, you would have to start redesigning the entire program from the ground up.

Along came a group called Deepmind, that generally works in machine learning to solve problems. They wanted to create a program that could play chess. What was unique in their work, was that instead of teaching the program how to play chess, they taught it HOW TO LEARN BY ITSELF. The program was only given the rules of chess, and then made it play against itself to practice. All that the program was set up to do, was change itself after playing each (set of) games - to try to be a little bit better. As it started with no knowledge of chess, it played pretty randomly (and therefore extremely poorly) at the beginning. How good do you think it could get just playing on its own with no guidance whatsoever?

To the amazement of everyone not only in the artificial intelligence community but also the chess community, after just a few hours of self play, this program (called AlphaZero - because it had zero human knowledge) was able to beat not only the best humans, but even state of the art computer programs. By the end of the training period, it was the strongest chess entity in history.

What's more amazing however, is HOW it played. When the games it played were analysed, AlphaZero played in an exceptionally UNIQUE and CREATIVE way, with ideas that were never seen before in human or computer chess play. By learning without any human interference, it had learned most of what we know, but also discovered NEW and powerful ideas. Chess, a game studied by the brightest humans for hundreds of years, and this computer was able to surpass it all from breakfast to dinner. All by practicing with itself, and trying to improve.

And there's more. This same program was - without any additional help - able to also repeat this historic feat in the games of Go and Shogi.

Ok, that's all we're going to hear about machine learning details.

But what does this mean for us?

What does this mean for learning? And what does this have to with fitness techniques? Well, if an entity can develop the same (and even greater) competence as collective human knowledge with just self practice, is that the thing we should be prioritizing in our teaching?

Why do we always start with technique?

I asked myself this question in my teaching, and the answers I came up with all had something in common. I wanted my students to learn fast. I wanted my students to be successful. I didn't want my students to get hurt. I wanted to teach more students together. I wanted to show how much I knew about the subject. I wanted to teach them. But were they learning?

It's important to remember that teaching happens in the teacher, but learning happens in the student.

And since they happen in different people, they are different processes. We have to ask ourselves: how much teaching is really required for learning? And how much are we doing just because "a teacher has to teach something"? This brings us back to Question 4: What is the effectiveness of no technique?

AlphaZero shows us that technique helps, but is not as crucial as we thought. We, as humans, have some handicaps that we have to work with, like a finite amount of energy, degradation with fatigue, and quite a fallible memory. We also have a fairly low input-output speed and limited time to work with, so we need to be somewhat efficient with our learning processes. But in over-optimizing for efficiency, are we losing out on a vast landscape of possibility? By teaching everyone how to play the game the same way, are we missing out on other approaches? By over-teaching, are we wasting our time as teachers? By micro-managing the learning process, are we robbing students of control, responsibility and independence?

What CAN we take away?

The first time I reached this stage, I felt a little silly. I felt silly for potentially wasting my time and my students’ time. I felt bad for potentially robbing them of their self-sufficiency. But what do we do?

I think the first thing we can take away is a sense of relief that students will eventually learn by practice, despite whatever we teach them. Once we let go of the need to hold their hands (or bodies), we can start to think ourselves not as deliverers of information, but as guides. Not TEACHers, but facilitators of the learning process. Instead of the programmers that specified what to do in every situation, we can be more like the Deepmind team - creating a structure where the students can learn through practicing, evaluating their actions, and then trying to improve them. We will definitely have to share ideas with them, give them suggestions and feedback. The experienced eye of an expert will often help reveal things that the learner may miss. We must be there to keep them on track, keep them motivated, and of course, keep them safe. But we must remember that the learning happens in them, not through us.

The good news from the latest artificial intelligence research is that they also agree with this approach. If you look at the strongest modern chess programs today, they are indeed a hybrid of the two approaches.

Taking this light-handed approach is not easy, but remember, we never hold a babies legs or tell them to flex their quadriceps while teaching them to stand or walk. One of the most complex bodily movements that we make. It involves a multitude of joints (the ankle, knee, hip, twisting of the spine, arm swing, etc...). It involves strength but also balance, stabilization and multitasking with other movements: walking is really complex. Then we decide to add momentum to the mix and try to run as well. As if that wasn't enough, we also rapidly change direction and orientation on surfaces that are sometimes extremely uneven. And how do all the humans on the planet learn how to do this? Do we teach everybody through technique? Do we tell them : lift left knees, then push heel ahead, then lean forward, then put the foot down, then push with the back leg toes pushing from the calf and glute, then lift the foot and kick back... Does this seem practical?

How do babies learn to walk and eventually run? We show them, and if they are interested, then they try it. Of course, they fail at the start, but since they're interested, they'll try again. And they will probably manage a second of stumbling before they fall. And do that 20 more times immediately after. But they will learn something from each of those 20 times, so that by the 21st, they now are wobbly but upright for a good 3-4 seconds before they plonk down again. To the baby, it may seem like a small feat, but you know the parents will be over the moon watching this.

It may take a while, but the baby eventually learns to stand, walk, and even run really fast after a few months and years. As teachers of both physical and intellectual disciplines, I think that we can actually take solace in that we don't need to explain every action, but inspire them, encourage them when they succeed, create feedback when they fail, and let them LEARN. The hardest thing we have to do is stand back and watch this beautiful process unfold.

[Pause and think]

How can you as a teacher/practitioner try this out in your own practice or teaching? Drop in your ideas as comments or send me an email if you found this interesting and would like to chat about it.

Annotations

[1] This article has a lot of places where the way to get the most from it is to pause and try to think of your own answers to the questions before going forward. I hope you will try it here, and I'll leave this little note at all the other places where I think it will be useful.

[2] Let % effectiveness be number of people that when instructed with the technique achieve the desired change in performance