Chapter 8: Dynamical Systems Become Extended and Embodied

Welcome to Last Minute Lecture.

This free chapter overview is designed to help students review and understand key concepts.

These summaries supplement not replaced the original textbook and may not be redistributed or resold.

For complete coverage, always consult the official text.

Welcome back to The Deep Dive, where we take a stack of dense, fascinating source material and really try to extract the essential insights for you.

And today, we are really undertaking a critical deep dive.

Our mission is to forge a strong, testable, theoretical foundation for what has become, I mean, probably the hottest debate in cognitive science today.

We're talking about 4E cognition.

Exactly.

That's embodied, embedded, and active and extended cognition.

And the central question here, the one this deep dive really orbits around, is whether the cognitive system, the thing that thinks and perceives and acts, is a self -contained entity that's locked inside the skull.

Or whether it necessarily reaches out, out into the body and even into the environment.

Right.

And the core thesis we're pulling from our sources today is revolutionary, but it's also, I think, beautifully simple.

It's that we really want to understand how the mind works.

We have to stop treating cognitive systems like they're closed boxes.

Absolutely.

We have to see them for what they are.

Open, far from thermodynamic, equilibrium systems.

And that's not just some, you know, philosophical hand -waving.

No, it's a real shift.

It's a huge theoretical shift.

And it's powered by something called dynamical systems theory, or DST.

And this is what makes the really radical claims of 4E cognition, not just plausible or interesting, but actually empirically testable.

Okay.

So let's unpack all of that.

Because the starting point for this whole framework is, surprisingly, not in psychology or neuroscience, but way back in a mid -century chemistry lab.

It is.

And we have to start there.

It's an essential historical analogy that really sets the stage for everything that follows.

So where are we going?

We have to go back to 1951.

To a Russian chemist, a guy named Boris Pavlovich Bilisov.

He's working in Moscow and he makes this just amazing discovery.

He finds a chemical reaction in a liquid solution, a homogenous mixture, so it's to colorless and then back again.

A chemical clock.

We know about these today, but I guess back then.

Back then it was unthinkable.

Bilisov tried to publish his findings in major Russian chemistry journals and he was rejected.

Not just once, but twice.

Why?

What was so shocking about it?

His paper was just deemed too improbable.

The prevailing scientific dogma, I mean the absolute bedrock of chemistry at the time, was this unwavering belief that all chemical reactions had to be understood as closed systems.

Let's make sure we're super clear on those terms.

What does closed system mean here?

A closed system is one that exchanges no energy or matter with its environment.

Think of a perfect thermos flask.

Totally insulated, nothing gets in, nothing gets out.

An equilibrium.

Thermodynamic equilibrium is the state where everything just stops.

All the macroscopic processes cease.

And disorder, what scientists call entropy, is at its maximum.

And the second law of thermodynamics says a closed system has to move towards that state.

It must tend toward maximum disorder.

It cannot, on its own, spontaneously create more organization or order over time.

It's a one -way street.

So if you mix chemicals in a closed box, they should react, find their lowest energy state, and just stay there.

So a color -changing clock, which is a spontaneous, repeating, organized cycle,

is a form of increasing order.

It is just chemically impossible under that worldview.

The journals literally thought Belisov had faked his data or made some fundamental mistake.

His reaction, which was later rediscovered and explained by Anatole Zabotinsky, was a perfect example of a system operating far from thermodynamic equilibrium.

It wasn't closed.

It wasn't closed.

And it wasn't until the 1970s that science, largely through the work of Ilya Pregagine, had the theoretical tools to explain how systems that are open to a constant flow of energy can spontaneously organize themselves.

So Belisov's discovery was dismissed as magic, because the science of his day didn't have a box to put it in.

And you're saying this leads us right to the core problem in cognitive science today.

It's the exact same mistake.

It's in a different field.

Decades later, we so often treat cognitive systems, the brain, the mind, as if they are closed, self -contained, near -equilibrium systems.

It's our default setting.

And that creates a paradox.

Right.

Because living cognitive systems are so obviously ordered.

They perform these incredible feats of organization, of growth, of development.

They do.

And if you assume this highly ordered system is self -contained, the classic brain in a vat thought experiment is the perfect example, then all that order seems mysterious.

Where does it come from?

You end up with what philosophers call explanatory gaps.

We struggle to bridge the gap between the messy physical brain activity and the organized subjective experience of consciousness.

Right.

If the brain is just a self -contained computer, you almost have to assume there's some kind of pre -written program inside that creates all that order.

But we can never seem to find it.

The mystery only exists because we're starting with the wrong assumption, the closed system assumption.

So the 4E perspective just flips that on its head.

It completely flips it.

It argues the order isn't pre -programmed at all.

It emerges.

It's the product of what our sources call soft assemblies of many, many components.

What does that mean?

It means they're not rigid, fixed structures.

They are temporary functional groupings that are defined by their interactions across scales, from neurons to hormones, muscles to the environment, not by some built -in pre -specified blueprint.

So if we just drop the idea of the brain as a closed box and instead embrace the idea that its order is emergent, that it's constantly maintained by interaction with the outside world, the whole problem looks different.

It does.

And that's the moment dynamical systems theory walks onto the stage.

It provides the mathematical and theoretical tools that just elegantly dissolve the mystery of Belisov's reaction.

And now it can do the same for cognition.

Exactly.

It gives us a non -magical empirical way to understand how order emerges in living cognitive systems.

And that paves the way for all the claims of 4E cognition to actually be successful.

OK, so if we're going to use DST to build this new foundation for 4E cognition, we have to be specific.

I mean, dynamical systems theory is a massive field.

We're not using the whole thing.

That's a critical point.

We have to be really clear to avoid confusion.

For our purposes, we're defining DST for cognition very narrowly.

It's based on two really crucial empirical hypotheses.

And these are hypotheses about what exactly?

They're about observation.

The version of DST we care about depends on the observation that one, the features of a cognitive system are not well defined independently of one another.

And two, cognitive systems themselves are not well defined in isolation from their environment.

They exist in this continuous dialogue.

So these are the two big claims.

Yeah.

And they're formalized as the interaction hypothesis and the openness hypothesis.

Let's start with interaction.

The interaction hypothesis states that any state or behavior within a cognitive system can only be properly characterized in relation to other states and behaviors within that same system.

So you can't just define, say, the state of one neuron without referencing its relationship to another neuron or to the hormonal environment it's in.

You can't.

And as you mentioned before, this isn't just a snapshot in time.

The temporal aspect is absolutely paramount here.

Interaction extends over time.

Right.

So a picture of all the states in my brain right now is not the same thing as my cognitive system.

No.

It needs a history.

It needs a trajectory through time to be meaningful.

Without that history, that persistence,

an exact copy of your current states might not even count as a cognitive system at all.

It's like pausing a movie.

You have the still frame, but the story only exists in the movement.

That's a great way to put it.

And this hypothesis, this idea interaction, is what we use to actually define the system's boundaries.

It gives us an empirical way to draw the line.

Yes.

We use these dependence relations to figure out where the system begins and ends.

A system by this definition includes all and only those elements that depend to a specified extent upon one another.

So some things are just more connected to each other than to other things.

Way more connected.

We can identify these neighborhoods of high interdependence.

The brain and the spinal cord, for instance, are so deeply dependent on each other that they clearly form a system.

OK.

That makes sense.

Now let's move to the second one, the openness hypothesis.

So if interaction is about the internal stuff, openness must be about the relationship with everything outside.

It is.

Openness is an extension of interaction.

It's applying that same logic to the relationship between distinct systems like the cognitive system and, say, the atmospheric system.

But it's also a refinement.

A refinement.

How so?

It adds a crucial requirement.

The non -equilibrium requirement, which we saw was the key to understanding Belisov's chemical clock.

Right.

It has to be in that far -from -equilibrium state.

Correct.

The openness hypothesis states that the states and behaviors of any cognitive system can only persist when they are not at equilibrium with the other systems they interact with.

Their boundaries have to be semi -permeable, allowing for flows of matter, energy, or information.

So why is that so important?

Why is being far from equilibrium so critical?

Why can't a cognitive system just be stable and near equilibrium?

Because that state is what makes self -organization possible.

And self -organization is?

It's the appearance of order, of complexity, or structure without any kind of pre -written blueprint or template.

If you were at equilibrium, any little bit of structure that appeared would just decay back into disorder, back into entropy.

It's a constant input of energy, that constant fight against entropy that lets complexity build up and be maintained.

That feels intuitively right for a biological system.

Life itself is the ultimate example of complex, structured order.

It is.

And we can even test this with a thought experiment.

The oxygen thought experiment.

Okay.

Imagine you're observing a healthy, living, biological, cognitive system.

Now if you force that system toward equilibrium, you isolate it, and most importantly, you cut off its oxygen supply.

What happens?

The cognitive features, they would vanish.

Instantly.

The openness is broken.

It's not that oxygen is cognition, of course.

But cutting off that necessary external system, the atmosphere, it starves the system of metabolic resources.

And with fewer resources, the parts of the system, the neurons, the cells, they become less dependent on each other.

The strong, highly correlated interaction you need for structured thought just breaks down.

The system has become closed relative to a necessary gradient, and its ability to do cognitive things just disappears.

So the claim is that cognition, as we see it, only happens under these two conditions.

Strong internal interaction and constant external openness.

Exactly.

And we have to stress these are empirical hypotheses.

They are claims about cognitive systems as we actually observe them in the universe.

Right.

This isn't metaphysics.

Not at all.

Every cognitive system we've ever seen is biological, and every single one exists in a far from equilibrium context.

We're just describing the conditions under which cognition actually happens here on Earth.

Okay.

Let's drill down into what these two ideas, interaction and openness, really mean in practice.

What are their measurable effects?

Let's go back to interaction.

Interdependence and correlation.

Right.

So interrelated causes just means the parts of a system are highly dependent on one another.

To understand component X, you absolutely have to know what's going on with Y and Z.

Our sources use a set of abstract equations to model this with the variables X, Y, and Z all being coupled together.

They are.

And if you imagine this as a real -world system, it becomes clearer.

Let's say it's a simple ecological model.

X is the rabbit population, Y is the amount of grass, and Z is the fox population.

Okay, so you can't just study the rabbits in isolation.

You'd get a completely wrong picture.

To predict or understand the rabbit population, you have to be measuring the grass and the foxes at the same time.

Their fates are intertwined.

And the really big conceptual jump for cognitive science here is realizing that the individual components, the rabbits, the grass, the foxes, they do not matter more than their interactions.

That's the key.

If you just focus on variable X, on the rabbits, you miss the most important feature of the system, which is its long -term behavior.

And this is where that idea of percolation comes in.

It is.

A small fluctuation in, say, variable Z, maybe a few foxes get sick, that little change will cascade or percolate through the entire system.

It will affect the rabbits and the grass, and it will shift the whole system's behavior for a long time.

The coupling is so tight that those little ripples don't just fade away.

They ripple indefinitely, which leads directly to the idea of correlation.

Right.

And correlation is just a measure of how much that ripple matters over time and space.

It's the extent to which the behaviors of the components are quantitatively similar.

And we can see this correlation length change dramatically depending on how open a system is.

The classic physics example here is the laser system.

Right.

At low energy, before you're pumping enough power into it, the photons are just being emitted randomly.

The components are basically independent.

This is a system with an extremely short, temporal correlation length.

Meaning that what one atom is doing right now tells you basically nothing about what another one will be doing a tiny fraction of a second later.

Exactly.

Like 10 to the minus 100 seconds later.

It's negligible.

They're all acting on their own.

But then you crank up the energy.

You make the system highly open to an external power source.

And the mutual coupling just skyrockets.

The system self -organizes.

The atoms start emitting light that's increasingly in sync, increasingly similar.

The whole thing starts acting like a unified whole.

The correlation length gets much longer.

Hugely longer.

Maybe a whole second or more.

The system becomes coherent.

And it shows how energy and openness create that long -term mutual dependence.

So how do we take this idea from lasers and abstract math and actually measure it in a living, breathing person?

Well, this long -term temporal correlation is.

It's everywhere and far from equilibrium biological systems.

It's how we characterize them.

For real.

You can measure this.

Absolutely.

For instance, in humans,

tiny seemingly random fluctuations in the timing of your heartbeats or in your footfalls when you're walking.

They have measurable effects minutes later.

Wow.

They are correlated across very long time scales.

It's proof that the human biological system is highly interactive.

Those tiny fluctuations don't just vanish.

They percolate and they affect the system's entire trajectory over time.

It's a unified whole.

Okay.

That covers interaction.

Now for openness.

Yeah.

You said that for self -organization to happen, we need at least three systems and a gradient.

Yes.

A gradient is just the difference.

The pot of water on the stove is the perfect analogy.

Right.

You have the hot burner, the cooler water, and the cold air around it.

Three distinct systems, all differing in temperature.

When the burner is on, you have a temperature gradient, a flow from hot to cold, and that maintains a non -equilibrium state.

The constant flow of energy keeps the water system open.

If you turn the burner off, the gradient vanishes.

The system reaches equilibrium and you just get a pot of lukewarm water.

The water molecules themselves, they act more or less independently in that state.

When you have that strong gradient, when the system is pushed far enough from equilibrium, something incredible happens.

This is where the order comes from.

The dependence relations among the water molecules extend exponentially.

You get this qualitative shift.

Random independent movement becomes structured.

You can actually see persistent rotating convection cells form in the water.

And that structure wasn't programmed into the water.

It emerged from the interaction under the pressure of that constant energy flow.

And that's why openness is so vital for biology.

The complex systems that enable heartbeats, gate patterns, the very organization of life,

they can self -organize because they are open to energy and information from their surroundings.

It's really the only way to square it with thermodynamics.

The second law says order should decrease in a closed system.

But in an open system with a sufficient gradient, order and information can persist and even increase.

I mean, just look at the earth.

Think about it.

The earth is this hugely complex system of open subsystems, geological, atmospheric, biological, all running on massive energy gradients, mostly from the sun.

If you ignore that constant flow, that openness, life, and cognition start to look like magic.

Just like Belisov's clock looked like magic to his colleagues.

So to sum this up, the dynamical view says cognitive systems are self -organizing collections of highly dependent parts.

And they organize non -magically because they are open to the world.

This is the foundation we need to even start talking about for e -cognition.

And now we get to the real payoff.

How does accepting these two ideas, interaction and openness, actually support the specific claims of embodied and extended cognition?

Okay, so the first big practical effect is that DST gives us a way to make system boundaries an empirical question.

It's not just a matter of intuition or philosophy anymore.

Exactly.

System boundaries are defined by measurable interactions.

And this is key relative to a phenomenon of interest.

So if you're studying how someone learns a new dance, the system you're studying will include all the parts that are highly interdependent for that specific task.

Right.

You don't start by just assuming cognition is in the skull.

You test it.

If the phenomenon you're studying requires strong correlated interaction with, say, the dancer's legs or the rhythm of the music, then those things have to be included in your definition of the system for that task.

Because of openness, these boundaries can change.

They're variable.

Highly variable.

They depend on the energy and information available.

So as a researcher, you have to set an empirical threshold.

You have to ask how strong does the interaction have to be for something to count as a component of the system.

And that depends on what you're trying to do.

If you want to affect someone's behavior,

your definition of the system has to include everything that has a significant effect on that behavior.

And this just fundamentally challenges that default neurocentric assumption that cognition is always and only in the nervous system.

Which brings us straight to embodied cognition.

The claim that the cognitive system includes parts of the body outside the nervous system.

Because those parts are strongly interacting with it in a far from equilibrium context.

The philosophical argument that really drives this home is the re -examination of the brain in a vat.

Thought experiment by Thompson and Cosmeli.

It's such a powerful argument.

They say, look, it's not enough to just keep the brain biologically alive in the vat.

To sustain a brain that actually experiences anything, you have to maintain its incredible, spontaneous, endogenous activity.

All the metabolic demands, the hormonal processes, the constant busy regulatory loops.

And here's the kicker.

The brain itself regulates most of these processes.

It acts to get the raw materials it needs to maintain its own highly ordered state.

So if the brain needs constant inputs to maintain its own order and it's built to regulate those inputs itself, like blood flow or hormones, then those regulatory loops are actually part of the process of cognition.

Their conclusion is just devastating for the isolationist view.

They argue that any vat trying to sustain an experiencing brain would have to be functionally equivalent to a human body in a real environment.

The body isn't just a life support system.

It's an active participant.

It is.

An isolated, disembodied cognitive system is, from this DST perspective, a theoretical impossibility.

And we have real world data on this.

That functional human brain grown in a lab back in 2015 by Renee and Ann's team.

Right.

And while it was a huge biological success, he was very clear about the ethics.

He said, we don't have any sensory stimuli entering the brain.

This brain is not thinking in any way.

Which implicitly supports this whole viewpoint.

A brain cut off from a body in an environment is not a cognitive system.

It's not.

But for the real empirical proof, you have to look at studies that measure the coupling directly.

Like the 1998 study by Kelso and his colleagues on finger movements.

Oh, this is a classic.

Tell us about it.

So participants moved their right index finger in four different patterns paced by a metronome.

The metronome is the external non -equilibrium input.

And they measure the brain activity in the left sensorimotor cortex.

You expect that the brain activity for, say, flexing your finger would look totally different from extending your finger.

They're different motor commands.

That's what you'd think.

But the astonishing finding was that the cortical activity was nearly identical for all four movement types.

As long as the pacing rate from the metronome was the same.

Get out.

Really?

Really.

The brain activity wasn't being driven by the internal program for what kind of movement to make.

It was being driven by the external pacing.

So the brain, the nervous system, the muscles,

they were all forming a single,

temporary,

self -organizing, embodied system with the metronome.

Exactly.

The boundaries of that cognitive system for that task extended well beyond the skull to include the informational input from the metronome.

OK, now we take that same logic and we push it even further.

Out of the body and into the world.

This is extended cognition.

The claim here is that sometimes the strongly interacting components include parts outside the individual biological body.

This is the one that sounds the most radical, but the DST framework makes it a pretty straightforward, empirical question.

It does.

If a non -biological thing is strongly coupled with the biological parts and that coupling is necessary for the phenomenon you're studying, then it's part of the cognitive system.

And we see this in multi -agent systems.

We do.

Richardson and his colleagues did this great study with two people sitting in rocking chairs.

When they were told to cooperate on a task, they would tend to start rocking and phase with each other.

So think about what's included in that functional system.

You've got parts of two brains, two nervous systems, two musculoskeletal systems.

And crucially, the mechanical properties of the two rocking chairs themselves.

Right.

When they cooperate, the coupling between them actually starts to dominate their individual rocking behavior.

A new single extended system emerges and it's defined by the spatial and temporal correlations during that cooperative task.

And the really critical point is that the system's boundaries are task -dependent.

As the sources say, there are potentially as many definitions of cognitive systems as there are distinct cognitive phenomena.

Right.

If you change the task from cooperation to competition, that extended system just dissolves.

Then you have the second type, human tool systems.

This is where things get really interesting.

The study by Dotov, Nee,

and Camero with people playing a video game with a mouse.

Yes.

They wanted to prove, empirically, that the person and the mass formed a single dependent system, not just a person using a tool.

And how do you even test for that?

With a disruption test.

For six seconds, while they were playing, they secretly severed the connection between the mouse movement and what was happening on the screen.

The feedback loop was broken.

Ah, wait.

And they found that during normal play, the human plus video game formed a single system.

The moment they disrupted that coupling, the correlations broke down.

But the really amazing part is how they measured it.

They didn't just look at the big movements.

They look at what's usually considered noise.

They looked at the tiny, subtle fluctuations in the hand movements, things like physiological tremor that most experiments would filter out.

They were looking for long -range, temporal correlations in that noise.

Which is the telltale signature of a far from equilibrium, strongly interacting system.

It's that percolation effect we talked about.

It is.

When the mouse was working normally, those tiny, noisy movements showed predictable long -range correlations.

The noise wasn't noise at all.

It was structured feedback.

It showed the tight coupling.

And when they broke the connection, those correlations vanished immediately.

The evidence was right there in the noise.

The human and the tool were coupled into one open, strongly dependent system.

But only as long as that functional gradient, the task was maintained.

So after all that, what's the big picture here for cognitive science?

Yeah.

I think the key takeaway is that this isn't a win -or -takes -all argument where we have to say the mind is never in the head.

No, not at all.

DST actually paves the way for a kind of pluralism.

What do you mean by that?

It allows for a plurality of different cognitive systems to exist, even simultaneously.

The boundaries are dynamic.

They're functional.

They're determined by the interactions of all these parts in open context and by the specific thing we're trying to explain.

So it could be that for some really specific, localized phenomena like, say, memory encoding in the hippocampus, the system really is mostly bounded by the neural activity.

It could be.

But the DST framework strongly suggests that very few complex, real -world cognitive phenomena are going to be strictly confined to the skull because those things usually rely on the larger body and on external tools for their stability and complexity.

This really challenges our gut feeling about what an individual is.

It completely does.

We are spatially and temporally bounded things, sure.

But at the same time, we're parts of larger and smaller cognitive systems that are defined by these dynamic relationships.

The individual self is constantly part of a network that's always forming and dissolving.

Okay, but I have to push back on this a little.

Please do.

If the rocking chair is part of the system,

well, where does it stop?

I'm holding a coffee mug right now while we're talking.

Is my mug part of this cognitive system?

It feels like the boundaries just dissolve into everything.

That is a fantastic question.

And it's the most common one, right?

The key is that word from before, dependence.

Is your performance in this conversation highly dependent on the specific properties of that coffee mug?

Is there a strong measurable coupling?

Probably not.

But for the people in the study, their ability to synchronize their rocking was absolutely measurably depender on the physics of the chairs.

It's all about the strength of the coupling relative to the task.

So it's not arbitrary.

It's an empirical question.

It's an empirical question every time.

And if this pluralistic position still feels too wishy -washy, the framework gives traditionalists two very clear ways to prove it wrong.

OK, why are they?

First, you could show empirically that for any phenomenon we would properly call cognitive, the parts that account for it are only highly interacting brain -bound parts.

If you can find a truly complex cognitive task that relies only on internal neural coupling, the 4E claim gets a lot weaker.

And the second way?

The second way is to show that for any phenomenon that is accounted for by a system extending beyond the brain, like the mouse or the rocking chairs, that phenomenon is not properly cognitive.

You'd have to argue that what's happening isn't really part of the mind.

So until someone does one of those two things, DST provides a solid empirical basis for taking 4E cognition seriously.

And it lets us move beyond just asserting it, and it helps us avoid appealing to magic.

One last clarification.

Does DST get rid of the need for mechanistic explanations?

You know, the kind of gears and a clock

explanations that traditional cognitive science likes.

Not at all.

In fact, it complements them.

Even a discrete mechanism like the gears and a clock is itself a system that only exists because it's a well -bounded system of strongly interacting parts that is open in a non -equilibrium context.

The gears only turn because of the wound up spring.

That's the gradient.

That's the gradient.

So DST explains the occurrence of the mechanism.

It explains the conditions under which that ordered thing can even exist.

And then the mechanistic approach can explain the appearance of the specific cognitive phenomenon.

They work together.

So to recap our whole deep dive here, dynamical cognitive science sees cognitive systems as these strongly interacting, highly correlated collections of parts that are far from equilibrium.

And this non -equilibrium state, which is maintained by these constant gradients of energy and information, is what makes the incredible self -organization we call life and cognition possible in the first place.

Which means that all those big questions about embodiment and extension,

they're really just empirical questions about boundaries.

It's about identifying which collection of things forms the highly correlated system you need for a specific cognitive act.

Given interaction and openness, DST implies that forecognition isn't just possible.

It's probably the most rigorous way to describe how minds actually work in the real world.

And here's the final thought we want to leave you with.

Given that your cognitive system is defined by these dynamic dependence relations relative to a specific task, whether you're playing a game, working with a team, or just walking down the street, you have to accept the possibility that the boundaries of your cognitive system, the actual set of parts driving your successful behavior, are constantly forming and dissolving as you move between tools, tasks, and other people.

You aren't a fixed agent.

You're a dynamically boundary -changing entity, defined by the flows and interactions you engage in from moment to moment.

That realization fundamentally changes how we think about ourselves and the world.

Thank you for joining us for this Steam Dive.

We hope you walk away feeling a little more informed and a lot more curious about the systems you're a part of every day.

A warm thank you from the entire team.