Dialogue with Liam Graham about Emergence

Emergence, a word whose Latin roots are ex = “out” + mergere = “merge, sink”, has 19 different meaning according to Oxford’s English Dictionary. While emergence is also widely used in quantum statistical mechanics, Liam Graham argues in his new book “Physics Fixes All the Facts” that we should abandon it and are better off replacing it by “physicalism”. In this post I discuss with him why.

About Liam Graham: Liam holds a bachelor degree in theoretical physics, a master degree in philosophy and a PhD in economics. After working as an associate professor for macroeconomics at University College London, he decided in 2018 to become a science writer.

Required background: None.


Philipp: Dear Liam, thanks for agreeing to discuss with me about the concept of emergence.

When I get asked by laypeople what I am researching on, I often try to explain what an emergent phenomenon is. The standard example (and probably the simplest one) is the emergence of life (such as us humans) from lifeless constituents such as atoms. If people can follow me, I then try to explain issues closer to my research, such as the emergence of an arrow of time (few laypeople know about time-reversal symmetry) or the quantum-to-classical transition.

Would you say these are good examples for “emergence” or do you have another favorite? How would you explain emergence to a layperson?

Liam: Dear Philipp, thanks very much for inviting me.

Those are three great examples. But for me the problem is rather finding an example of something which is not emergent. To see this, take the standard definition of emergence, “more is different”. Then think about a single body in an empty universe. Add in a second and suddenly you’ve got new properties: orbits; mutual centres of gravity; Kepler’s laws. These are emergent; two is different from one.

If such a simple system has emergent properties, then so does (almost) everything. “More is different” just says that interactions matter. Emergence labels phenomena which aren’t fundamental. This is what most physicists mean by emergence and I think it fits your examples of time and classicality. As an aside, I doubt we can ever know what is fundamental. But let’s leave that for another time.

The main aim of my book is to argue for a “physicalist” view of the world in which the final explanation of everything is in terms of interactions governed by physics. Clearly emergence-as-interactions fits this view. But this usage is not my target. The trouble with the term emergence is that it has acquired an edge of mystery, a suggestion that emergent phenomena lie outside the scope of physics. It’s often seen as a middle way between physicalism and belief in the supernatural. Outside physics this meaning is pretty much ubiquitous. And some eminent physicists agree: Ilya Prigogine, Robert Laughlin and George Ellis among them.

You can see this in your third example. Life emerging from lifeless stuff already has a whiff of magic. I’d tell this story firstly by pointing out that “life” is an arbitrary conceptual distinction which only exists in the minds of humans. From another perspective, creatures are the latest way the universe has stumbled across to configure physical systems to maximise the rate at which entropy increases.

Living systems are on a continuum with whirlpools and crystals, the result of a complex set of interactions. There’s a very neat paper which explains the chemical processes which enable bacteria to swim towards food. These bacteria process environmental information and respond to it in a way that meets their needs. This is the basic characteristic of living things, but the paper shows that it’s no more than chemistry.

Similarly, you can give detailed accounts of how the arrow of time is related to statistical mechanics or classicality to quantum physics. But we can’t do this yet for many of the processes that constitute life, let alone intelligence and consciousness. When we describe them as emergent is it just a shorthand for “we don’t know how to do the physics yet”?

P: These are a lot of interesting points and to some of them I would immediately agree whereas other points require more scrutiny in my view—aside from the big question what is actually “fundamental”.

I guess the catchy sentence “more is different” (or, as the Greeks would say, “the whole is bigger than the sum of its part”) is too imprecise to properly define what emergence is.

Coming back to the example of gravitationally interacting particles, you are right that 2 is qualitatively more than 1, and probably also 3 is qualitatively more than 2 (since 3 gravitating particles are chaotic but 2 not). However, it seems that there is no qualitative difference between 3 and 4 or 4 and 5, etc. But then, if you come to 10,000 particles, you might observe robust patterns in your data that you can describe with new laws (say, hydrodynamic laws) and that are not even definable for, say, 5 particles (concepts such as local densities, pressure, temperature, etc. make no sense for 5 particles). Wouldn’t you agree that there is a big difference?

So, in a sloppy way (but probably more precise than “more is different”) you could define emergent phenomena as everything that requires a new degree (or at least a new course) at university. You need a biologist to understand life and not a physicist. The laws of life, even if rooted in physics, have quite an independent status and are, as a matter of fact, also completely different from the laws of physics.

L: I’d agree that “more is different” is imprecise. And you’re right, you could give a definition in terms of scientific disciplines. But this would be just as imprecise: boundaries between disciplines are notoriously fuzzy and also change over time. Wikipedia lists over 60 fields beginning with the letter “A”. It’s hard to imagine this tells us much about the structure of reality.

What you say about robust patterns is important. But such patterns don’t suddenly appear when a threshold is passed. Instead, as you add in particles, distributions become tighter and laws less approximate. Phase transitions are an example. The more particles there are, the sharper is the transition, but it is always continuous. While transitions in macroscopic samples seem discontinuous, in fact this is only the case in the thermodynamic limit.

If life seems to have a different status it’s because we don’t yet know how to do the physics. Just as, say, the law of stellar evolution, seemed to have an independent status until we understood fusion. When we can trace the development of life from chemical reactions on the early earth to self-replicating chemicals to protocells, the different status will disappear. This is physicalism: there is a sequence of physical processes leading continuously from chemistry to living systems. If instead the process is discontinuous, then at the discontinuity something different must slip in. Either this is new physics and we’re back to physicalism. Or it’s something non-physical and we’re dualists. In either case, there’s no place for emergence.

P: I totally agree that there is nothing fundamentally discontinuous about phase transitions. They appear discontinuous because of our own limited resolution with which we can glance at nature. Perhaps this is an important characteristic of emergence? For instance, we see classical (and not quantum) physics around us because we view things at a coarse temporal and spatial scale.

I am not so sure, however, about two other interesting points you raised. First, assume there are around 60*26 ~ 1500 scientific fields/disciplines. This seems a lot, but we know that absolute numbers have no meaning. I think one could compare it with the ~10^80 atoms in the observable Universe (neglecting photons for simplicity). If we make the cruel approximation that each atom stores one bit of information, we find that there are “only” 1500 disciplines for understanding 10^80 bits of information. I think this tells us something about reality…

Second, I agree that the definitions of scientific fields is notoriously fuzzy and changes over time, so perhaps our view on what is emergent also changes over time. This is fine for me, but I think I object to the idea that the status of fields changes once we know how to do the physics. Take thermodynamics as an example. Even though we are by now convinced that the laws of thermodynamics can be derived from a microscopic statistical mechanics description (even though there are some gaps to be filled in), the discipline of thermodynamics didn’t vanish. Physicists, engineers and chemists who did research on thermodynamics were not replaced by physicists working on quantum statistical mechanics. And for good reasons: stat mech physicists don’t understand more about thermodynamics and their methods are of little help for thermodynamic problems.

Now, my prediction is that also at the end of this century (if humans still exist by then), we will have people researching thermodynamics. Perhaps they disappear at some point, but I would claim that’s a strong point for the concept of emergence.

L: I think the interesting points you raise all relate to human cognitive limits. Let me start by returning to the bacterium. We describe it as swimming towards food because it is hungry. This is fine for practical purposes. A biologist might go further and says that it senses a chemical gradient and adjusts the motion of its flagellum accordingly. But both are gross approximations to a complex set of chemical processes.

Similarly, if we’re designing a fridge we use thermodynamics since for that purpose it is a good approximation to the underlying statistical physics. And so on for all the other branches of science. They are coarse-grained descriptions, shortcuts which, as you say, efficiently compress information and are essential for practical purposes. But they compress information in a way that is right for human cognition and human purposes. It’s the nature of conceptual thought to abstract from detail and use coarse-grained descriptions, not the nature of reality.

Your point about different scales emphasises this. Some phenomena seem distinct because we are too big to see the details. Again, this is about the structure of our minds rather than the structure of reality. If we could build an intelligent system on the nanoscale, it wouldn’t have much use for thermodynamics since it would experience the statistical physics directly.

Setting scale aside, would different sorts of intelligence slice up knowledge in different ways? A 2022 paper has an AI analyse physical phenomena and finds it comes up with explanatory variables different from those in standard models. But this is just one paper and it’s early days.

Calling all of this emergence hides what’s interesting. When we want to talk about the structure of science, our scale or our cognitive limits, let’s talk about them explicitly. Returning to physics, rather than calling phase transitions, quasiparticles or convection cells emergent, let’s think about them as three distinct types of useful macroscopic approximation. Does it add anything to cover all these very different ideas with the blanket term emergence?

P: Okay, I think we are touching some really nice question here. But let me first summarize two points on which, I think, we agree.

First, what the scientific communities calls “emergent phenomena” is nothing mystical or outside physics, at least as far as the majority of scientists is concerned (I would say). Second, a precise or eternal definition of emergent phenomena seems difficult since it relies on experimental capabilities and human-made concepts and interests that change over time. And maybe you are right that “emergence” is not a good word and we should talk about “macro-approximations” or something similar, even though I like the word emergence…

Now, an interesting question you raised is how subjective is emergence? That’s a very difficult question in my view. The quick intuition here seems to suggest that emergent phenomena have a strong anthropocentric flavor as a nano-sized intelligent creature moving close to the speed of light might come up with completely different emergent phenomena.

Alas, I believe this conclusion is a bit too quick. My argument is that any creature capable of doing physics at a level comparable to us would necessarily come up with similar theories, macro-approximations or emergent phenomena (whatever we like to call them). Let’s consider thermodynamics. Unless there is something really bizarre that we have missed, we know that all living creatures require access to free energy, which is turned into waste heat (aka increases the entropy of the Universe). In particular, creatures able to build a civilization and think about science will have an even higher need of free energy, so it seems inevitable that they cook up some sort of theory that helps them to understand heat engines.

I also remember a nice post by Matt Strassler about the question why is the speed of light so fast. I think he quite correctly points out that, first, the actual question is why are humans so slow and, second, any living atomic creature will necessarily view the speed of light as fast (a creature moving at high speeds simply couldn’t live very long). So it seems likely that any naturally evolved intelligence first comes up with Newtonian physics before discovering general relativity. I further recall a Mindscape podcast episode with Arik Kershenbaum, who argues that it is quite likely that extraterrestrial life will look not too different from us. In my view, these are strong arguments showing that emergent phenomena are not completely arbitrary, rapidly fluctuating, and subjective concepts.

L: Newtonian physics and thermodynamics are examples of effective (field) theories (EFTs), approximations valid over some range of energy. Yes, this is sometimes used as a definition of emergence. Much of physics is about creating EFTs appropriate to the question being studied. But there’s a problem. To see it, my favourite example is the dozen or so EFTs based on quantum chromodynamics (you could do a similar exercise for cosmology or CMP). This proliferation does not reveal the structure of reality but is about our computational resources and the purposes for which the EFTs are used.

I would contend that the same is true for your two examples. There are other approximations to statistical physics (caloric theory, engineering thermodynamics) or to general relativity (Kepler’s Laws, the rule of thumb we use to catch balls). Once again, these are conditional on our knowledge and our purposes.

I don’t think we can say much about other intelligences: reality has a striking ability to surprise us. An empirical approach, and I think a fascinating research question, would be to ask AIs to reinvent science. Until we meet some aliens, this is the only way we might be able to step outside our preconceptions.

And we definitely disagree that most scientists use emergence in a benign way. Even in physics, describing phase transitions as emergent tends to brush over the interesting dynamics around the transition. Outside physics, in current debates over things like the origin of life, AI, consciousness or stock market crashes, the word usually means “we don’t have a clue what’s going on but labelling it as ‘emergent’ makes us sound clever”. Outside physics, it’s often seen as a way of escaping from the view that everything is physics. And some eminent physicists share the responsibility for this.

I think we agree that the definition of emergence excludes nothing (except the fundamental level, if there is one); it changes over time and is extremely vague. I think we also agree that it is used to mean many different things including large-number approximations, EFTs, chaotic behaviour and the structure of knowledge. Is there another definition in science with so many undesirable properties?

But you’re right: physicalism is a ghastly word whereas emergence is somehow uplifting.

P: Unfortunately, I am not knowledgable enough about quantum chromodynamics and that stuff, but it sounds very interesting what you say. However, your reply about thermodynamics and gravity didn’t convince me. Caloric theory was, as far as I remember, a theory trying to explain thermodynamics and didn’t question its validity. And how far “engineering thermodynamics” differs from thermodynamics, and why this is a point against emergence, remains unclear to me. Similarly, I see no contradiction or problem between Kepler’s law, rules of thumb to catch a ball and Newtonian gravity…

I remember an interesting psychology blog about the fact that there are only nine possible reasons why people disagree. The first one was simply different experiences or “prior information”. In my life I never witnessed somebody abusing the word emergence. Coming back to the beginning of this conversation, before I start explaining what emergence is, I ask laypersons whether they have heard of “emergence”—and the vast majority (even those with a university degree) answer “No!”. In addition, at least in my field, people spend an enormous effort to precisely quantify violations from macroscopic-19th-century thermodynamics or the border between the quantum and classical world. I never met a physicist believing that emergence is outside physics or that understanding the precise nature of fluctuations around phase transitions is uninteresting. It is surprising to find out that you made such different experiences. Perhaps we can agree that those “eminent physicists” you are alluding to should not be called “eminent”? 😉

I would like to finish the discussion by you explaining a bit better what “physicalism” means, as I am not yet sure whether this is a useful and/or precise term. So, first of all, I like to know how you distinguish physicalism from realism or materialism. Second, I wonder whether you have a precise procedure to detect possible effects “outside physics”, i.e., is physicalism testable or a pure believe? And finally I wonder about the following. Suppose you have a microscopic physical theory with X axioms able to explain everything around you, and suppose also you have another theory about your mind with X axioms explaining everything around you. Which one is better? Is the latter one non-physical?

L: I thought you were claiming that effective theories like Newton’s laws and thermodynamics are non-arbitrary cases of “emergence”. I wanted to argue that they too are arbitrary, a function of our capabilities and purposes. Kepler’s Law or the rule of thumb for catching are just as much approximations to General Relativity as Newton’s Laws and have the same claim to the label “emergent”. So once again it turns out that everything is emergent. But I may have missed your point.

It’s not surprising you haven’t come across the wilder uses of “emergence” since you work right down at the heart of things. Lucky you! But the further you get from fundamental physics, the worse it gets.

Let’s start with the physicists: Robert Laughlin (Physics Nobel 1998) “A Different Universe”; Ilya Prigogine (Chemistry Nobel 1977) “The End of Certainty”; George Ellis (Hawking’s co-author) “How can physics underlie the mind?” and Paul Davies “The reemergence of emergence”. The latter two have also written numerous articles on the topic.

Further from physics, there’s the work of Stuart Kauffman, Harold Morowitz or Terrence Deacon. Or search for “emergence” followed by a field of your choice: biology, consciousness and AI all work well. There’s also lots of interesting work by philosophers of science which I discuss in the book along with a list of 75 (yes, 75) definitions. But I conclude that none of them help.

Then to physicalism. I use it as a synonym of materialism (which has fallen out of fashion). Realism I take to mean that the universe exists independent of thought. This could include non-physical things, so is distinct from physicalism.

Physicalism is just like any other scientific hypothesis. If, as in your example, a theory based on the mind could explain the physical world, we’d have to abandon physicalism or at least modify it so dramatically we’d need to change its name. It’s easy – and fun – to think of other things that would contradict physicalism. What if researchers found a sequence of DNA shared by all life and, after repeated checking, it was found to read “© Yahweh 4004 BC”? (I borrow this example from here). Or if ghosts were reliably seen? Or if you could change the result of an experiment just by thinking about it? In the absence of such phenomena, I think physicalism is the only game in town.


Comments

Leave a Reply

Your email address will not be published. Required fields are marked *