Levels on Levels on Levels
A brief discursion on the idea of emergence
Let’s go on a tour of the brain. Don’t worry too much about the names of things as we go. Focus on having fun. Stick your arms out of the car.
Let’s take a simple neuron, the fundamental unit of neuroscience. It fires electrical signals through your brain, but itself is is composed of innumerable chemical and electrical processes. Neurons are attached to and supported by glial cells, such as astrocytes, which provide nutrients and regulate blood flow.
You can break the neuron down, into the main axon and dendrites, and many other fiddly bits, which in turn contain litanies of molecules bouncing around, sending signals and repairing damage and storing information.
Those molecules in turn can be decomposed into atoms and quarks and all the fuzzy quantum electrodynamics below that.
If instead of zooming in, we zoom out, we can see that each neuron is linked together to other neurons. When enough of these are connected together, and if they fire together in a particular way, we group them into a region, and give it a name, like the dentate gyrus. This is then where memory encoding takes place.
And in order to do memory encoding, the dentate gyrus has a bunch of close friends that it likes to hang out with, like the hippocampus and the entorhinal cortex, so we group those into something called the hippocampal system.
The hippocampal system rarely works alone. It is often interacting, at range, with other regions of the brain, encoding memories and later transferring those memories into the neocortex. These sorts of processes occur for every process, not just memory formation, and almost every process seems to use large chunks of the brain.
And that’s just inside the brain. The brain is attached to the body, and must factor in huge quantities of different information about things like hunger, whether we’ve had too much sugar and should in fact not eat a fourth muffin, and whether we feel sad.
If you zoom out further, from the body, we are gregarious creatures, and like to talk to other people. This entails a whole host of new dynamics. Imagine we’re at a party and we hear some speech. A lot of things happen.
We hear the words. We hear the tone. We combine them with context. Sometimes we tune 95% of the speech around us out completely, honing in on a small fraction, as in the cocktail party effect. Sometimes the way someone’s mouth moves can determine how we actually hear speech at all, as with the McGurk effect, where subtle changes in mouth shape mean the same sound can be interpreted differently by your brain.
We also change our response to speech based off of whether we think a person is attractive, whether they work in corporate law and whether they’ve recently fought a gorilla.
Interpreting speech alone is inordinately complicated, let alone all the stuff that actually underlies the process of speech interpretation. At some level, we know that speech processing is occurring in a set of neurons, some of which are located in a big juicy region at the front of the brain called the dorsolateral prefrontal cortex, and some of which are located elsewhere.
Here’s another way of seeing what I just described:
And here’s something I’ve found frustrating for a long time. Why bother studying individual neurons at all? Isn’t it impossible to go from the level of neurons and neurotransmitters to the level of the cocktail party effect? How could you ever tie those two together?
The Solution
There is an answer. It’s that you can’t. At least, not easily, and doing so often generates its own complications. The problem is that this is mostly a fool’s undertaking.
And just to digress here, before you decide that I’m pottier than a kiln - I’m not about to say studying particles and molecules and neurons is pointless. I know the idea is to study say, a brain circuit, and learn lots about that brain circuit, and then once you know enough, intervene. But some thinking about the kinds of thinking that are helpful here is useful. I’m trying to learn to think with problems, rather than against them.
I’ve been reading I Am a Strange Loop, by the philosopher and scientist Douglas Hofstadter. In one of his chapters he concocts a wonderful analogy, from which I’m going to quote at length, because it saves me having to do any typing. I’ve stripped out the less relevant parts:
Imagine an elaborate frictionless pool table with not just sixteen balls on it, but myriads of extremely tiny marbles, called "sims" (an acronym for "small interacting marbles"). These sims bash into each other and also bounce off the walls, careening about rather wildly in their perfectly flat world - and since it is frictionless, they just keep on careening and careening, never stopping…
… The sims are also magnetic (so let's switch to "simms", with the extra "m" for "magnetic"), and when they hit each other at lowish velocities, they can stick together to form clusters, which I hope you will pardon me for calling "simmballs". A simmball consists of a very large number of simms, and on its periphery it frequently loses a few simms while gaining others.
There are thus two extremely different types of denizen of this system: tiny, light, zipping simms, and giant, ponderous, nearly-immobile simmballs.
The dynamics taking place on this pool table — hereinafter called the "careenium" — thus involves simms crashing into each other and also into simmballs.
Why have I block quoted a long and seemingly facile analogy about marbles? Because this is a really good way of thinking about what is happening in the world.
It is true that we are glommed-together-conglomerations of particles, all of which obey very strict rules about how particles should move. But this doesn’t particularly matter for us.
Once you move between the levels of the Simm and the Simmball, the rules of the game you are playing change. I think the analogy game is useful for understanding what I’m driving at, so let’s play it again.
Imagine the Tower of Babel, which all of the Earth’s people worked together to build, before it was destroyed by a wrathful God who hated teamwork. In this analogy, we’ve sent God on a corporate away-day, so he’s learned the power of teamwork and the Tower remains standing.
However, while the tower is standing, it is propped up by lots of different groups, who all in work in different ways. Let’s pretend at the bottom floor, people use stilts to prop up the tower. One floor up, the floor rests upon the stilts, but people on that floor decide to use granite. Another floor up, people have used cups. If we go up again, there’s just a dude called Kevin holding up the floor. We can go on.
Different floors have different assumptions, different theoretical backgrounds, and rely on different parts of underlying floors in order to support them. This is for me somewhat like the landscape of science, which contains bewildering arrays of devices, formulas and tricks that prop each other up, creating a vast edifice of possibilities.
If you start really trying to unpick why the clown floor holds up the cheese floor, you realise that actually, there’s a lot more to understand about the clowns and their motivations. If you want to understand why the floor as a whole remains standing, you have to figure out the relationship between the clowns and the strange material just to their left. But the clowns are also supported by the crosshatched stilts, in turn balanced on the cups- you get the drift.
What’s that emerging over the hill?
This idea is called emergence. Emergence is a term we use to describe situations where the rules change as you zoom in and out, and how things are defined by what they are related to as much as by what they are composed of.
Often the lower levels are prioritised as somehow deeper or more meaningful. You may have heard the expression that “psychology is explained by biology, which is explained by chemistry, which is explained by physics, which is explained by maths”, plus or minus a few sciences. It isn’t really true. Granted, maths is unreasonably effective, but as you move between levels, emergence gets in the way, and messes up the sums. Maths gets less and less unreasonably effective as you move up the levels towards interactions at a societal level.
To keep analogising, here are some examples of emergence that I came across or thought about writing this post:
1. Imagine you were to climb up an extremely tall but real tower, such as the Petronas Towers. I choose these because they are the tallest tower that I have ever been up, not because I am sponsored by a Malaysian oil company.1
If you look down, you notice a few surprising things. Firstly, you remember you have extreme vertigo and that you shouldn’t be climbing any towers. Secondly, people disappear. Kuala Lumpur becomes a city of cars and houses and skyscrapers. You can see architectural structures of wealth and class that are invisible from the ground. They even have a model of this on the observation deck, which obviously doesn’t contain people. All the people who built these things and operate them and are the sole reason for their existence evaporate into nothing.
2. In 1972, Philip Anderson wrote an article called More is Different, where he pointed out that as we move to different levels of enquiry, the rules change. He points out that when large numbers of particles come together, the collective system can display entirely new properties that are not present at the level of the individual components.
Symmetries that hold at one level can break at another, giving rise to new kinds of order — for example, the way superconductivity or magnetism emerges only in aggregates of particles, not in any single atom.
3. In a celebration of Philip Anderson’s life, Steven Strogatz writes:
In 1665, while confined to his room with “a slight indisposition”, Christiaan Huygens noticed that two pendulum clocks he had recently built were keeping perfect time together. When one clock’s pendulum swung to the right, the other’s swung to the left, exactly 180 degrees out of phase. When Huygens tried disturbing their oscillations, he found to his astonishment that the board from which both clocks were suspended began to jiggle. That jiggling gradually nudged the pendulums back into antiphase synchrony. In letters to his correspondents, he described this “odd sympathy” of clocks as “marvelous”.
Huygens’s work launched the study of synchronization, a phenomenon that pervades the natural and technological world, from congregations of fireflies that flash in unison to arrays of superconducting Josephson junctions. Yet although more than 350 years have passed since Huygens’s observations, we still don’t fully understand the sympathy of pendulum clocks mathematically. The main obstacle is that synchronization is a nonlinear phenomenon, which makes the governing equations impossible to solve explicitly. On top of that, the equations include non-smooth, impulsive effects, stemming from the sudden jolts imparted by the clocks’ escapement mechanisms. Even though we know the laws for individual pendulum clocks, that isn’t enough to tell us how two or more of them will behave together.
4. In the Franck-Hertz experiment, which demonstrated the quantum nature of atoms, the electrons move at around 6.5x10^5 m/s, and the mercury moves at 230 m/s. This means the speed of the mercury is irrelevant and can basically be ignored in calculations. Some properties are relevant at different scales and some aren’t.
5. George Cronk writes of George Herbert Mead’s work: “The world in which the self lives, then, is an inter-subjective and interactive world—a ‘populated world’ containing, not only the individual self, but also other persons. Inter-subjectivity is to be explained in terms of that ‘meeting of minds’ which occurs in conversation, learning, reading, and thinking.”
6. From Herbert Simon: “In a rare gas, the intermolecular forces will be negligible compared to those binding the molecules - we can treat the individual particles, for many purposes, as if they were independent of each other.”
7. From the introduction to a series on statistical mechanics by David Tong:
Suppose you’ve got theoretical physics cracked. Suppose you know all the fundamental laws of Nature, the properties of the elementary particles and the forces at play between them. How can you turn this knowledge into an understanding of the world around us? More concretely, if I give you a box containing 10^23 particles and tell you their mass, their charge, their interactions, and so on, what can you tell me about the stuff in the box?
There’s one strategy that definitely won’t work: writing down the Schrödinger equation for 10^23 particles and solving it. That’s typically not possible for 23 particles, let alone 10^23. What’s more, even if you could find the wavefunction of the system, what would you do with it? The positions of individual particles are of little interest to anyone. We want answers to much more basic, almost childish, questions about the contents of the box. Is it wet? Is it hot? What colour is it? Is the box in danger of exploding? What happens if we squeeze it, pull it, heat it up? How can we begin to answer these kind of questions starting from the fundamental laws of physics?
Questions on the subject of whether a box is hot, wet or exploding are of much more interest to 99% of humans than solving a Schrödinger equation.
8. Eve Marder points out that the very same patterns of neural activity can have completely different underlying dynamics. One pattern of activity can be caused by a completely different set of transmitters.
Understanding at one level does not necessarily grant you understanding of the next level up. It may instead, by a sort of diabolical biological subterfuge, deceive you into believing that you understand them, but later turn out to have tricked you and left you stranded up a chimney with no trousers.
Grabbing another drink
I could go on, but let’s instead go back to the cocktail party for a second. When we’re here, talking to someone about whether you could beat a gorilla in a fight, it doesn’t matter so much what is going on under the hood at the lower level. These levels will have some impact, but your understanding of those impacts can be contained within the assumptions of the levels you’re working at - here, the cocktail party.
I think this is a useful idea conceptually, because a lot of discussions often get bogged down by people standing on different level of the tower sniping at each other. This happens in every field, but let’s stay in neuroscience.
I regularly see an argument that boils down to something along the lines of: “Why are those people bothering with that level, when they could be studying the brain at this level?”
“How can we possibly understand someone at the cocktail party without a complete understanding of the changes in dopamine transmission occurring in the dorsolateral prefrontal cortex?” snipes a nefarious neuroscientist that I’ve made up to make this point.
Another might shoot back with, “How can we possibly use the neurotransmitter architecture of the dorsolateral prefrontal cortex to help us understand what is going on at a complex event such as a cocktail party?”
In this piece, Michael Gazzaniga at his decision-making level, argues that other researchers cannot hope to understand humans from the level of neuronal electrical activity.
In this piece, a team of researchers argue that individual neuron studies are great and well done and that was spiffing stuff chaps, but we should also think about the brain from an ensembles of neurons level.
It’s not that any of these people are wrong about the value of their own particular level, its just that it isn’t really a competition. I think that much of the best work in science is working to expand a particular level. New insights at a particular level allow you to make new predictions at the level above it.
Having a clear model of the different possible levels you can work at in your mind is useful, and makes thinking easier. New research in neuroscience may also the ability to move between different levels become more relevant, as work at one level offers us the opportunity to influence higher level. Formalising these relationships can be helpful.
For instance, brain–machine interfaces now decode neuronal activity and can control neuroprosthetics. We can evoke memories by reactivating a neuronal ensemble in a mouse that were active in the animal’s brain for a given previous event.
These different conceptual levels of analysis are just a tool to think about the world. Feel free use them, and to pick them up and put them down as much as you want.
And I’m dreadfully sorry for using the word “level” enough to make you want to live in a bungalow.
Endnote
I experimented with explicitly labelling each “level” in this post with a specific term, such as a “noi” (I made up a word for each of the building blocks of a Tower of Hanoi) or a “leVel” (because you can play on level’s palindromic tendencies, and stack the words together, i.e. leVeleVeleVel). I opted not to bother with it, because I think it just ended up being more confusing.
However, you could see an explicit conceptual language for moving between different levels being valuable. For instance, in Donella Meadow’s excellent book Thinking in Systems, she introduces a bunch of concepts like “paradigms” and “stocks” and “flows”, which you’re meant to use like playmobile to build your conceptual model of the world.
Her framework doesn’t quite reach the goal I have here though - the word “paradigm” often does a lot of heavy lifting in these discussions, and belies the actual complexity of what is going on. There is rarely a clear dominant paradigm in a field that everyone subscribes to. The actual playing field is much messier, and each level, let alone the whole superstructure, can often be decomposed into different warring factions.
Thanks for reading! If you enjoy this stuff, I want to do more of it, and subscribing makes it easier for me to do that.
But hey, if you’re Petronas CEO Tengku Muhammad Taufik Tengku Aziz and you’re reading this, call me.
Great read, thank you.
Subscribed
Absolutely love this! Really, really nice read :)