The Futility of Emergence

Prerequisites:  Belief in Belief, Fake Explanations, Fake Causality, Mysterious Answers to Mysterious Questions

The failures of phlogiston and vitalism are historical hindsight. Dare I step out on a limb, and name some current theory which I deem analogously flawed?

I name emergence or emergent phenomena—usually defined as the study of systems whose high-level behaviors arise or "emerge" from the interaction of many low-level elements.  (Wikipedia:  "The way complex systems and patterns arise out of a multiplicity of relatively simple interactions".)  Taken literally, that description fits every phenomenon in our universe above the level of individual quarks, which is part of the problem.  Imagine pointing to a market crash and saying "It's not a quark!"  Does that feel like an explanation?  No?  Then neither should saying "It's an emergent phenomenon!"

It's the noun "emergence" that I protest, rather than the verb "emerges from".  There's nothing wrong with saying "X emerges from Y", where Y is some specific, detailed model with internal moving parts.  "Arises from" is another legitimate phrase that means exactly the same thing:  Gravity arises from the curvature of spacetime, according to the specific mathematical model of General Relativity. Chemistry arises from interactions between atoms, according to the specific model of quantum electrodynamics.

Now suppose I should say that gravity is explained by "arisence" or that chemistry is an "arising phenomenon", and claim that as my explanation.

The phrase "emerges from" is acceptable, just like "arises from" or "is caused by" are acceptable, if the phrase precedes some specific model to be judged on its own merits.

However, this is not the way "emergence" is commonly used. "Emergence" is commonly used as an explanation in its own right.

I have lost track of how many times I have heard people say, "Intelligence is an emergent phenomenon!" as if that explained intelligence. This usage fits all the checklist items for a mysterious answer to a mysterious question. What do you know, after you have said that intelligence is "emergent"?  You can make no new predictions.  You do not know anything about the behavior of real-world minds that you did not know before.  It feels like you believe a new fact, but you don't anticipate any different outcomes. Your curiosity feels sated, but it has not been fed.  The hypothesis has no moving parts—there's no detailed internal model to manipulate.  Those who proffer the hypothesis of "emergence" confess their ignorance of the internals, and take pride in it; they contrast the science of "emergence" to other sciences merely mundane.

And even after the answer of "Why? Emergence!" is given, the phenomenon is still a mystery and possesses the same sacred impenetrability it had at the start.

A fun exercise is to eliminate the adjective "emergent" from any sentence in which it appears, and see if the sentence says anything different:

  • Before:  Human intelligence is an emergent product of neurons firing.
  • After:  Human intelligence is a product of neurons firing.
  • Before:  The behavior of the ant colony is the emergent outcome of the interactions of many individual ants.
  • After:  The behavior of the ant colony is the outcome of the interactions of many individual ants.
  • Even better: A colony is made of ants. We can successfully predict some aspects of colony behavior using models that include only individual ants, without any global colony variables, showing that we understand how those colony behaviors arise from ant behaviors.

Another fun exercise is to replace the word "emergent" with the old word, the explanation that people had to use before emergence was invented:

  • Before:  Life is an emergent phenomenon.
  • After:  Life is a magical phenomenon.
  • Before:  Human intelligence is an emergent product of neurons firing.
  • After:  Human intelligence is a magical product of neurons firing.

Does not each statement convey exactly the same amount of knowledge about the phenomenon's behavior? Does not each hypothesis fit exactly the same set of outcomes?

"Emergence" has become very popular, just as saying "magic" used to be very popular. "Emergence" has the same deep appeal to human psychology, for the same reason. "Emergence" is such a wonderfully easy explanation, and it feels good to say it; it gives you a sacred mystery to worship. Emergence is popular because it is the junk food of curiosity. You can explain anything using emergence, and so people do just that; for it feels so wonderful to explain things. Humans are still humans, even if they've taken a few science classes in college. Once they find a way to escape the shackles of settled science, they get up to the same shenanigans as their ancestors, dressed up in the literary genre of "science" but still the same species psychology.


Part of the sequence Mysterious Answers to Mysterious Questions

Next post: "Say Not 'Complexity'"

Previous post: "Mysterious Answers to Mysterious Questions"

125 comments, sorted by
magical algorithm
Highlighting new comments since Today at 6:13 PM
Select new highlight date
Moderation Guidelines: Reign of Terror - I delete anything I judge to be annoying or counterproductiveexpand_more

Hmm, interesting. I've never actually realized that people used "emergent behavior" as a model or an explanation for anything. In that context, I'd always treated it as just a description, with the meaning that an "emergent phenomenon" is a "complex or seemingly complex phenomenon arising from interactions of a large number of very simple subparts," or something of the sort. Never thought of it as a model or an explanation, but just as a reasonable descriptive word. But if it is used as an attempted explanation to end discussion, then it's just functioning as a curiosity-stopper and should be questioned further.

Eliezer: I generally like your posts, but I disagree with you here. I think that there's at least one really useful definition of the word emergence (and possibly several useless ones).

It's true, of course (at least to a materialist like me), that every phenomenon emerges from subatomic physics, and so can be called 'emergent' in that sense. But if I ask you why you made this post, your answer isn't going to be, "That's how the quarks interacted!" Our causal models of the world have many layers between subatomic particles and perceived phenomena. Emergence refers to the relationship between a phenomenon and its immediate cause.

So, for instance, suppose I'm on the interstate and I get caught in a traffic jam. I might wonder why there's a huge jam on the road. It's possible that there's a simple, straightforward explanation: "There's a ten-car pileup a mile further on, and five of the six lanes are shut down. That's why there's a traffic jam." Obviously we could get far more reductionist— both in terms of "why is there a pileup" and "why does a pileup cause a traffic jam"—but for the conceptual level we're operating on, the pileup is a full and complete answer. And thus the traffic jam isn't an 'emergent' phenomenon; it has one major identifiable cause.

In contrast, a lot of traffic jams 'just happen.' The previous sentence is false, strictly speaking; the jams come from somewhere. But you can't point to an individual cause of them; they arise from the local effects of millions of local actions taken by individual drivers. Removing any one of these actions wouldn't eliminate the jam; it's a cumulative product of all of them. So people searching for an explanation of why it takes two hours to dive ten miles in rush hour get really frustrated, because there's no good explanation to give them. And people trying to fix rush hour get even more frustrated, because there's no good angle to attack the problem from.

So emergence, in this sense, means that a phenomenon has many intertwined causes, rather than one or two identifiable and major causes. It turns out, of course, that most interesting phenomena are emergent (non-emergent phenomena are, by definition, boring, since their causes are straightforward). But "emergence" is useful as a shorthand for "the causes are complicated and interconnected, and I can't pick one out and tell you, 'here it is, this is why that happened.'" It's important not to get confused, and not to think an explanation of why we don't understand something is the same as an explanation of that thing. But as long as you remember that, it's a useful thing to remember.

In contrast, a lot of traffic jams 'just happen.' The previous sentence is false, strictly speaking; the jams come from somewhere. But you can't point to an individual cause of them; they arise from the local effects of millions of local actions taken by individual drivers.

I've actually seen a study on these types of jams, though I cannot remember the source. The results were pretty simple and surprising. The research discovered they could create a massive traffic jam on a full but still flowing highway by simply having a single car brake for longer than necessary.

The first person would brake for too long, causing the person behind him to brake for slightly longer (he isn't likely to brake for less time than the person ahead of him lest he risk an accident), which continued down the line, a chain reaction. Drivers in the lanes on either side of the initial brake chain would also begin braking as they saw people in the central lane brake, being sensibly cautious during rush hour, which would spread outward from their positions. Eventually traffic would halt, as the people ahead would have to stop completely before being able to move again.

I'm sure there was some kind of cutoff threshold regarding how long over the necessary length of time the first person has to break, but it wasn't very long, a second or two would do it during a non-jammed rush hour.

It also explains why, once a jam occurs for any reason, it is extremely slow to clear up even after the cause of the jam is long since removed.

Pretty shocking really, and certainly not an "emergent phenomena". That's why EY is against using emergence for everything - there absolutely must be a reason, and that reason cannot be "lots of stuff interacts and now we get a traffic jam!" Using emergence as an explanation encourages you to stop thinking about the problem, rather than dig in and figure out why what happened happens.

You have unexplained traffic jams - do you call it emergence or try to explain them? The rational thing to do is to try to explain them in a way that allows you to have expectations about future observations.

In other words, "Emergence" is an answer looking for a problem.

"Emergence" here would be a reference to the non-linear result of the braking. Like what Henry_V said.

Yes, the point is to be sure you aren't using "Emergence" or "Emergent Phenomena" as stop signs. That you recognize that there is in fact a cause (or causes) for what you are seeing, and if the total seems to be more than the sum of its parts, that there is some mechanism that exists that is amplifying the effects.

Emergence is not an explanation by itself.

In line with previous comments, I'd always understood the idea of emergence to have real content: "systems whose high-level behaviors arise or 'emerge' from the interaction of many low-level elements" as opposed to being centrally determined or consciously designed (basically "bottom-up" rather than "top-down"). It's not a specific explanation in and of itself, but it does characterise a class of explanations, and, more importantly, excludes certain other types of explanation.

I would think that something like "life/intelligence is an emergent phenomenon" means "you don't need intelligent design to explain life/intelligence".

Okay, but that's really not how I have understood emergence. It delineates a subject matter, and does so in an abstract way that includes many specific examples which are purportedly alike in some important way. But I don't think this use necessarily implies that the explanation has thereby been given. It is, rather, usually an attempt to delineate a subject matter which can be further investigated. I believe that the hope is that a general theory of emergence is possible, though my impression is that there isn't even a generally agreed-upon definition of it, let alone a commonly accepted theory.

One common element that I have sometimes noticed is that an emergent phenomenon can be idealized and a simplified mathematical model constructed of it, which is not precisely correct but which is a very good approximation. The existence of such simple and very good models is remarkable and extremely lucky for us.

For example, an actual fluid such as water is really made up of molecules that interact, but there is a simple mathematical model for fluids which treats fluids as absolutely continuous and smooth all the way down, not composed of atoms but fluid at every scale. As I vaguely recall, this simple mathematical model can be adjusted by plugging in values for viscosity, compressibility, and so forth. It is not exactly the same as actual fluids (the resemblance breaks down completely on the scale of molecules) but it is very close at the macroscopic scale.

Similarly, we have the concept of the "ideal gas", which is only an approximation to real gases but on a macroscopic scale a very close and useful one.

That we can vastly simplify and idealize something without losing all that much predictive power seems to be a characteristic of many so-called emergent phenomena.

Eliezer apparently travels in different circles than I do, and encounters people who use the word "emergence" very differently. Here is the kind of situation where I usually hear the word "emergence" used:

Me: Well, I think I'll build an AI that understands Chinese this weekend.

Philosopher: Build it from what?

Me: NAND gates, I suppose.

Philosopher: That's impossible. Searle proved it. NAND gates don't understand Chinese, even a little. So a collection of lots of NAND gates can't understand Chinese either.

Me: Huh? Searle and you don't get it. The understanding of Chinese is going to be an emergent property of the whole complex system.

Philosopher: "Emergence! Aaarghh!"

Me: Would you like me to explain the code to you?

Philosopher: No, thanks. I don't know anything about programming. But I do know that the word "Emergence" is a sure sign of messed up thinking.

In other words, I don't consider "emergence" as an inoculation against curiosity. I consider it an inoculation against stupidity. It is a claim by a reductionist that a high level phenomenon can be constructed from low-level machinery which is different in kind.

Most scientists I know use 'emergence" as I do. But I have to admit that most philosophers I know use it as Eliezer does. I guess we will just have to agree to dis- ... err, agree to miscommunicate. But I do wish that Eliezer would stop pretending that the word "emergence" has to have an explanatory function in order to be useful. It has a classificatory function. It collects together a class of models which have in common the property that naive reductionists fail to understand them. It is classification, not explanation. Putting the shoe on the other foot:

Philosopher: Your argument is fallacious.

Me: Aaarghh! "Fallaciousness" How does that explain how my argument is wrong. You are just trying to stop conversation.

Philosopher: But... But... You don't understand. Fallaciousness is not being used as an explanation.

There were some good comments on this thread, but I needed to add my own two cents.

I am glad that there are people other than myself who find the notion of "emergent phenomena" to be code word for "magic" or "ignorance." Quite frankly I am mystified by how learned people, presumably taught the scientific method, could be so enamored with so called "emergent phenomena" but then I recall that practically the whole discipline of modern cosmology is little more than dressed up religion (anyone here know what possible cosmological consequences "Dark Energy" or "Dark Matter" have). But in that vein it bares mentioning that if it is impossible to discriminate between two things, then they are logically identical. Magic = achieving work through methods that do not involve "work effecting activity." Emergent phenomena = quality achieving methods that do not involve "quality achieving activity."

In order to deal with the topic of Emergent Phenomena one is required to comprehend what is called "Supervenience." I will leave it to the readers to explore Supervenience at their own discretion. (Standford encyclopedia of philosophy has a good article on it). Suffice it to say that some people have managed to convince themselves that a form of Supervenience can exist whereby "Top down causation" can occur.

Strong claims to emergent phenomena (which are essentially arguments regarding causation) require that the Supervenient qualities occur out of nowhere. This would be like in real life saying that Kinetic Energy occurs without there being any potential energy. Potential energy is not itself energy; hence the POTENTIAL part of its name. So can anyone point out why we "assume" the existence of potential energy in matter but that "potential intelligence" is not present in neurons?

Weak claims to emergent phenomena (which are essentially classification arguments) reference sets which include all objects in the universe (as Eliezer has pointed out), and as such are completely devoid of explanatory powers. Technically all things in the universe are "emergent properties" of physics. But we could potentially explain everything in the universe "if we had perfect knowledge of physics." We would need something like Laplace's daemon, but with perfect predictability we could derive neurochemistry. Once you have perfect neurochemistry knowledge, you can derive economics or any other set of behaviors you wish.

Example: When two triangles are brought together a square is the result (technically a quadrilateral, but I am keeping it simple). There is no property of triangles that is equivalent with "squareness." So "squareness" can be said to be an emergent property of triangles. Right now every person reading this should realize that something fishy is going on. Each triangle has "potential squareness" as part of its "list of qualities that it possesses." So when two triangles are brought together you get a square. You do not get something for nothing. The only difference between the triangles in this example and everything described as "emergent" is the degree of complexity.

The "proper way" Emergent Property is supposed to be used is when you have a universe of discourse that is random and note that "small sections of apparent order" occur. These properties are then "emergent" as they are not connected with any law that the universe itself possesses. The problem is that any application of this to a level of reality short of applying it to reality itself is entirely inconsistent with all non-paraconsistent logics.

A truly random "thing" does not exist. Such a "thing" could not interact with any thing we are aware of, since any interaction involves a two-way quality exchange/copy/removal,etc. Does anyone here have any idea what "mono-interactive interaction" looks like? The only way to achieve that is to have a "nothing" do something.

The very definition of "Nothing" is that it lacks any and all qualities. No qualities means no abilities. No abilities means "doing" is impossible. Nothing does not result in a change in anything.

Just because you are not aware of all the causal factors DOES NOT mean that there do not exist any causal factors. You can use this for any possible "emergent phenomena." Just because you are not aware of selection mechanisms for quantum physics does not mean that there are no selection mechanisms. Repeat ad nauseum.


"(anyone here know what possible cosmological consequences "Dark Energy" or "Dark Matter" have)"

Ok, this is the point where I started to question your logic (incidentally, apology for the tangent).

I agree that Dark Matter and Dark Energy feel like epicycles and phlogiston. HOWEVER, they also feel like that or felt like that at one point to all actual physicists.

Therefore, if you claim that they do not exist, you must both know what the standard answer to that question is (for if there is no standard answer science would have abandoned those concepts long ago), and also why it is wrong, or in short you must know more about physics than every physicist on earth.

That is not quite so hard to do as it seems on first glance; Einstein did it. Maxwell did it. Planck did it. But it is important to realize that the chance that every scientist on Earth is wrong about thing X is significantly greater than the chance you just don't understand thing X.

As Eliezer requested, I offer my view on what emergence isn't: emergence is not an explanation. When I say that a phenomenon is emergent, I am using a shorthand to say that I understand the basic rules, but I can't form even a simple model of how they result in the phenomenon.

Take, for example, Langton's Ant. The ant crawls around on an infinite grid of black and white squares, turning right at the centre of each white square and left ant the centre of each black square, and flipping the colour of the square it's in each time it turns.

The first few hundred steps create simple patterns that are often symmetric, but after that the patterns Langton's Ant produces become pseudorandom. If left to run for around 10000 steps, the Ant builds a highway - that is, it falls into a pattern of 104 of steps, and at the end of each cycle, it has moved diagonally and the cycle repeats. After millions of steps, the grid has a diagonal streak across it. As far as we know, the Ant always builds a highway.

Highways are emergent by the definition I use - that is, I know exactly how Langton's Ant works, and therefore, in theory, know why it builds a highway, but I can't form a model of its behaviour that I can actually use. I simply do not have a good enough brain to actually run Langton's Ant. By this definition, conciousness is an emergent phenomenon (I know it's caused by neurons, but I have no idea how) but the behaviour of gases is not (I know the ideal gas law and its predictions seem reasonable if I imagine a manageable number of molecules bumping about).

By my definition, emergent is much like blue. "It's emergent!" and "It's blue!" are both mysterious answers if I asked for an explanation, but useful answers if I asked for a description.

Aren't superconductivity and ferromagnetism perfect examples of emergent phenomena?

Yes. So are non-superconductivity and non-ferromagnetism. That's the problem.

Uh. No. Non-superconductivity is not usually considered as an example of emergence. Because the non-superconductive system is composed of smaller subsystems which are themselves non-superconductive. Same goes for non-ferromagnetism. Not "emergent" because nothing new is emerging from the collective that was not already present in the components.

And even if what you wrote were true it would be a problem only if emergence were being used as an explanation. But, outside of the philosophy literature, it almost never is used that way. You are tilting at windmills here.

The apparent disagreement here, comes from different understandings of the word "non-superconductivity".

By "non-superconductivity", Yudkowsky means (non-super)conductivity, i.e. any sort of conductivity that is not superconductivity. This is indeed emergent, since conductivity does not exist at the level of quantum field.

By "non-superconductivity", Perplexed means non-(superconductivity), i.e. anything that is not superconductivity. This is not emergent as Perplexed explained.

Non-superconductivity means that moving electrons through it will result in the atoms moving unpredictably. It is a product of how electrons and atoms interact. It is less emergent than how, if they interact a different way, the atoms will not start moving unpredictably.

It's made up of non-superconductive subsystems in that if you take a little piece of it, that will be non-superconductive, but the same applies to a superconductor. You can't just take one atom and say whether or not it's superconductive. A current can't flow through one atom in a relevant sense.

I think that the point is that emergence is in the mind of the observer. If the observer is describing the situation at the particle level, then superconductivity is not there regardless of the size of the collection of particles considered. But, when you describe things at the flowing-electric-fluid level, then superconductivity may emerge.

Aren't the labels arbitrary?

Let's use sharpness.

Non-sharpness is not usually considered as an example of emergence. Because the non-sharp system is composed of smaller subsystems which are themselves non-sharp. Same goes for non-ferromagnetism. Not "emergent" because nothing new is emerging from the collective that was not already present in the components.

Let's use bluntness.

Non-bluntness is not usually considered as an example of emergence. Because the non-blunt system is composed of smaller subsystems which are themselves non-blunt. Same goes for non-ferromagnetism. Not "emergent" because nothing new is emerging from the collective that was not already present in the components.

That humans say "sharp", "blunt", "conductive", and "non-conductive" in English is due to circumstances of culture, technology, what minerals are abundant on Earth, etc. At least, I don't know the word, if there is one, for non-conductive.

To the extent "sharp" and "blunt" are not opposites, I apologize for the imperfect example.

Conductivity isn't there either unless you describe them at the flowing-electric-fluid level.

It seems to me that emergence is the opposite of rigorous structure. Take human brain function (similar to your intelligence comment in the article). Claiming that brain function is emergent versus rigorously ordered allows you to make a prediction, namely that a child who has a portion of their brain removed will retain all or a large portion of the functionality of the removed portion, or they will not. A child with half of their brain missing would be expected to be extraordinarily impaired. A simple search of the literature should prove it one way or another.

Thus, when one says that some property is emergent, it means that it is not limited by the macro form, but by the conditions effecting the micro components from which the property emerges. This should allow for all manner of predictive ability. Of course, there are plenty of people who latch on to the word, just like there are plenty of people who latch on to the word "evolution", and don't think or use it to make predictions, and in that, your point is well taken.

Sorry for commenting 5 years after the fact, but this place seems to have at least some ongoing discussion.

A child with half of their brain missing would be expected to be extraordinarily impaired. A simple search of the literature should prove it one way or another.

According to this, a child with half of their brain removed can sometimes do just fine. It has a lot to do with age, though, given that children have more neuroplasticity–a fully functioning adult would probably lose a lot of their normal abilities.

"Holistic" seems to label that phenomenon more clearly, for my money.

I do think there is a good deal of commonality among the reasonable comments about what emergence is and also feel the force of Eliezer's request for negative examples.

I'll try to summarize (and of course over-simplify).

When we have a large collection of interacting elements, and we can measure a property of the collection as a whole, in some cases we'd like to call that property emergent, and in some cases we wouldn't.

I can think of three important cases:

  • If we can compute the property as a simple sum or average of properties of the individual elements, then it is not emergent. So e.g. mass or temperature are not emergent properties.

  • If we need to analyze long chains of structurally specific causal interactions to explain the coarser grained property, then it is not emergent. So e.g. the time telling properties of a mechanical clock, or the arithmetic computing properties of a calculator are not emergent.

  • If we can compute the property as a function of the properties of the elements, and it depends sensitively on specific characteristics of their behavior and interaction, but is robust under local perturbations (i.e. doesn't depend on structurally specific causal chains), then the property is emergent. So e.g. percolation is emergent. Also we have some warrant to say that flocking, thinking (as brains do it), social interaction, etc. are emergent.

I'm not claiming these three cases cover all the legitimate positive and negative examples of emergence -- I don't think the concept has crystallized that completely yet. But I do think they answer Eliezer's challenge.

Another, less crisply defined question is whether we should be using "emergence" so defined, and relatedly, whether people are mostly trying to use it in this sense, or whether they are, as Eliezer fears, just using it as a synonym for "magic".

My own feeling is that many users of the term are groping for a clear definition of this general sort, and that they are doing so precisely to avoid having to explain a large class of phenomena by "magic".

Eliezer: Here's another example similar to ones other people have raised, a story I heard once, that might explain why I think it's an important and useful concept.

Supposedly, in the early nineties when the Russians were trying to transition to a capitalist economy, a delegation from the economic ministry went to visit England, to see how a properly market-based economy would work. The British took them on a tour, among other things, of an open-air fresh foods market. The Russians were shown around the market, and were appropriately impressed. Afterwards, one of the senior delegation members approached one of his escorts: "So, who sets the price for rice in this market?" The escort was puzzled a bit, and responded, "No one sets the price. It's set on the market." And the Russian responded, "Yes, yes, I know, of course that's the official line. But who really sets the price of rice?"

The Russian couldn't conceive that an organization as complex as the open air market could have assembled itself; he was sure someone must have designed it in order for it to work. It had to have been set up. But markets and prices are an emergent phenomenon; the price isn't set by one person and doesn't have any one cause. And yet the markets function.

Similarly, a lot of people seem to have a mental model of democratic institutions that says it's a non-emergent phenomenon: if you write a constitution and hold elections, you get a democracy with the rule of law. Others (including myself) claim that democracy and rule-of-law are emergent phenomena: if they don't exist, there's no specific set of actions a central actor can take that will cause them to exist. They exist because of millions of decentralized and uncoordinated actions of individuals without specific direction. If you hold the first view, projects like the establishment of the new Iraqi government make sense: we set up a government with a constitution and elections, so it should become a free democratic state. If you hold the second view, the project is insane: freedom and democracy require millions of individual and low-level cultural shifts that can't be imposed from above, so there's no way for us to turn the nation into a democracy. My point here isn't that one view is right or wrong, although I have a firm belief. My point is that it's highly relevant to our foreign policy to ask whether democracy is emergent or not.

Usually when you say, "You can't just impose X from above," you're claiming X is an emergent phenomenon; the hallmark of a non-emergent phenomenon is that it's possible for a single actor to take a series of actions that either cause or prevent it.

I don't buy the analogy between emergence and phlogiston or vitalism. Offering up "emergence!" as an explanation of a phenomenon is a category mistake, to be sure, and is a semantic stopsign when misunderstood this way.

As other commenters have noted, however, there is a proper understanding of emergence that is useful. (In philosophy, for instance, it's an admittedly sloppy but still useful term to classify different kinds of explanations of consciousness). This doesn't seem true of explanations that appeal to phlogiston or vitalism. Vitalist explanations aren't category mistakes. They're simply vacuous explanations, full stop.

Sean, that's a useful link. The "irreducible-pattern" epistemological version of emergence, described there, is one I'd heard before. It definitely wouldn't fit everything (if I had to bet, I'd bet it fits nothing).

In most of the contexts in which I have seen the word "emergent used", it has signified a lack of understanding of the underlying causes of the phenomenon being described but rather than acting as a semantic stop-sign seemed to be used as its exact oppposite - as a marker for an area sufficiently interesting to be deserving of further research with a view to eventual full explanation. But perhaps I've been misunderstanding the intent of the authors.

Creeping into his soul, he felt the first faint tinges of despair.

After all these posts on how the strength of an idea is what it excludes, forbids, prohibits, people are still citing positive examples as proof of the power of emergence? Tell me what it isn't!

To respond to a really old comment -

Emergence does exclude some possibilities. For example, if consciousness is emergent, it means that it's not ontologically basic, it's not caused by something outside the system, and that it exists.

I could that if your epistemology is popperian, negative evidence is all you can have.

The even/odd attribute of a collection of marbles is not an emergent phenomenon. This is because as I gradually (one by one) remove marbles from the collection, the collection has a meaningful even/odd attribute all the way down, no matter how few marbles remain. If an attribute remains meaningful at all scales, then that attribute is not emergent.

If the accuracy of fluid mechanics was nearly 100% for 500+ water molecules and then suddenly dropped to something like 10% at 499 water molecules, then I would not count fluid mechanics as an emergent phenomenon. I guess I would word this as "no jump discontinuities in the accuracy vs scale graph."

I don't understand this comment.

Two posters had offered only positive examples. Five had offered at least one negative example in contrast to a positive example.

Emergence is NOT the sum of the parts.

I'm curious, Eliezer, what you think of Alex Ryan's and Cosma Shalizi's definitions/formalisms of emergence?

The both seem to be claiming that emergence is more than you are, but that could be an illusion...

You ITprimer seems to disagree with your statement:

Emergence is NOT the sum of the parts.


(3) the non-trivial interactions result in internal constraints, leading to symmetry breaking in the behaviour of the individual components, from which coordinated global behaviour arises;

(4) the system is now more organised than it was before; since no central director nor any explicit instruction template was followed, we say that the system has ‘self-organised’ ;

(5) this coordination can express itself as patterns detectable by an external observer or as structures that convey new properties to the systems itself. New behaviours ‘emerge’ from the system;

Non-trivial interactions of individual components -> Self organization -> New behaviors labeled to have 'emerged'

Where did they emerge from? The non-trivial interactions. This description runs counter to your discription "Emergence is NOT the sum of the parts." It is the sum of the non-trivial parts by the above description and a loose definition of sum.

I remember when Warren Spector & Harvey Smith were going on about emergence in videogames. I think their definition was something like "a non-obvious [it may even surprise the designers] outcome of a system of rules rather than something scripted". That's a rather subjective definition but it seems to fit as well for the things that are described as "emergent" in real life. Since life is not actually a videogame but has universally valid rules, it would not be a very useful concept for that domain. I think Wolfram has written a lot about that sort of thing, but I don't actually know much about what it is he says other than that its an important idea.

What are phenomena that aren't "emergent"? I guess Eliezer is right when he says "a single quark". I think Eliezer makes a good case that the word is overused, and doesn't enlighten the discourse.

It might be more useful to describe things in reverse " X are the components of phenomenon Y". Such as "Neurons firing are the known components of intelligence". Because when we observe something, it can be useful to ask "what are its components"?

It contrast, everything observed IS the component of some bigger system, but it can be also useful to ask, what is the next biggest ordered system it is a part of, etc. That's where "emergent phenomena" might legitmately come in. Because an ant colony might be the next biggest ordered system that an individual ant is a part of, and that does seem like useful information.

Most of this is specific to videogames and probably will not be applicable anywhere else:

An emergent property in the context of videogames is one the designers of the game did not intend, [more strictly: yet is not a programming error].

Excluding the possibly, since this example is ambiguous using it:

In the game Super Smash Bros, jumping is not emergent, since the designers programmed it into the game specifically.

Wavedashing [dodging into the ground so that you will be able to move while attacking] (and in fact, every single bit of strategy for every character) is emergent; it's not programmed into the game, it's just that if you put together all the intended rules of the game, wavedashing appears also.

What does "this is emergent" tell you, in this context?

It tells you first of all it's unintentional, which then tells you it has a vastly greater chance of being unbalanced or broken.

Using the stricter definition, it also tells you whatever it is profits the player in some way, because if it did not profit the player in some way it would not have emerged; someone would have found it, not used it or told anyone, and it would just fade away. (But this is only valid for emergent things when they're structured in a certain way. This part can be generalized to, say, the economy, but not, say, to traffic jams, because traffic jams are more tragedy of the commons types of things.)

It also tells you, most importantly, that it is probably not possible to know all the specific causes of this thing and instead to try wide and general causes. (i.e: "World War I happened because of a general attitude among nations that military force was a good way to solve problems." It's possible to say it happened "because Alice thought... and Bob thought... but Carol thought... and Dave thought....", but this is going to be either much less accurate or not worth the effort to make it accurate.)

What does "this is non-emergent" tell you?

There are one or more obvious specific causes that it would not be worth breaking down. (In the case of videogames, the developers, but it also works for cases like "there is a big dent in the front of my car because I crashed it into a tree" [but wait, you say, isn't that also phrasable as "because the force from the tree caused this molecule and this molecule and this molecule to move backwards"? Yes, but it doesn't matter; the only cause is still the force from the tree.])

(Finally, random other example I thought of after reading the Go example:

In chess, the position of the pieces at the beginning of the game is not emergent: there is one cause for that: because it's part of the rules of the game.

The fact that the best first move for white in most cases is pawn to e4 is emergent. Nobody wrote that into the rules of chess; it's just a consequence of the positions of the pieces.)

What's the difference between a "programming error" and an "emergent consequence of the program as written", other than whether the programmers decide they like the result? Is it just a question of whether the rules involved can be described intuitively at the level of user-interface objects rather than lines of code?

Answer to your question: Honestly, I should not have included that line about errors in there at all; it doesn't need to be special cased out because most errors are emergent. (Not always; a missing negative somewhere is not emergent. But when you get to the complexity of a video game, most errors that will make it through QA are emergent.)

But also: I actually have thought about this a bit since I wrote this, and I think I can come up with a decent general definition for emergence: (don't worry, I'll get to your question in a moment)

Something is emergent when it is caused by a rule that works similarly to the second law of thermodynamics. (More specifically, the property of the second law that it isn't actually a hard law at all; it's just that when you crunch all the probabilities for all the particles involved, it is vastly more likely that the result will obey the second law then will not.)

Similarly, the ways economies develop aren't hard laws; it would be entirely possible for an economy to develop in such a way that it lets you get a free lunch. It's just that that, considering all the actors involved are out to find and take those free lunches, that you are about (using about very broadly here) as likely to find an actual free lunch as you are to find your foot has suddenly turned into gold.

(Also: I think it's a mistake to point at some finished product of laws of emergence and say it's emergent. "The economy is emergent" is just a short and slightly misleading way to say "The laws that govern an economy are laws of emergence".)

But going back to what this predicts: It predicts mainly that there is something equivalent to atoms in thermodynamics or actors in economics; some small unit of behavior that you can test for. It also predicts (in very complex systems it might not be possible to do any actual math on this, but in theory it predicts) how often the law will fail. (As noted, sometimes all you can say with confidence is "it might fail sometime"; of course if it fails OFTEN it doesn't have enough predictive value to justify keeping around.)

I'm getting the feeling that Eliezer is starting to get overly eager to attack semantic stopsigns. I recommend magic oil in the evening and emergent phenomena in the morning.

My impression of "emergence" was that it's closely related to pattern recognition. You have atoms A, B..ZZZZZZZZZZ, and you recognize that these atoms form a certain pattern. So you say that a supercluster of galaxies/bar stool/intelligence "emerges" from a bunch of atoms.

I once had a prolonged debate with an anticognitivist. He, as usual, argued that no matter what kind of AI you build, if you take it apart, it's just "switches flipping". In that debate, I maintained that intelligence does not require anything else - it emerges from switches flipping the same way that Firefox emerges from switches flipping. Both are just human names for patterns that arise in a sea of subatomic particles.

[That was one frustrating debate... Both of us were equally bewildered that the other refuses to get it.]

Isn't his usage of "switches flipping" basically another 'literary genre' switch--i.e. he had attached some sort of negative connotation to the phrase which he could not conceive of attaching to intelligence?

The example of emergence that comes to my mind most readily is a simple observation that Douglas Hofstadter made in Godel, Escher, Bach -- a book which definitely does not use "emergent" as a synonym for "magical":

In a game of Go, once there are two separate open spaces -- "eyes" -- in the middle of a connected group of stones, that group becomes invincible (because the opponent can't fill both holes with one move). There's no official rule in Go that says "Patterns with two eyes can't be captured", the rule just says that to capture a group you have to surround it completely and leave no open spaces. Thus two-eye invincibility is an emergent consequence of the rules of Go.

It's important that this new emergent rule is a significant simplification: once you realize that two eyes are invincible, you no longer have to do any complicated analysis about how close a two-eyed group is to being completely surrounded. It's safe, full stop (at least as long as you don't stupidly fill in the holes yourself).

The game of Go has very few rules. In practice, the two-eye invincibility "rule" is a very important and useful one, if you want to play the game well. To try to force Eliezer to talk about "emergence" or something equivalent, I would ask: where did the two-eye invincibility rule come from?

-- Okay, now, so what isn't emergent? There's a another Go rule, the "ko" rule, which says you can't play in such a way as to get the exact same board position back after two moves: no capture followed by immediate recapture unless it changes the board. There's nothing "emergent" about that rule that I can think of -- it helps keep Go games from going on forever, but it has no simple-but-unexpected high-level consequences.

There are a lot of strategic patterns of play in Go that are not emergent, either -- e.g., there are no simple rules of thumb that can tell you, in all cases, whether a group with one eye or no eyes can be captured or not. Often the answer depends on a single apparently unrelated stone way over on the other side of the board. No simplifications available here, therefore no emergence.

There are many other more involved examples of emergence (and non-emergence) -- gliders and spaceships in Conway's Game of Life come to mind, and Herschels and random ash densities -- but this blog comment is too narrow to contain a good summary of them all...

Two other books that do a fine job (in my opinion) of describing the concept of "emergence" as distinct from "magic" are Cohen and Stewart's The Collapse of Chaos and Figments of Reality.

i think by 'emergence' you just mean 'implication'

Eliezer: "Don't y'all find it a little suspicious that so many people think "emergence" is a useful concept, yet have different definitions of what it is?"

That's a non sequitur. Different people define intelligence differently, so what? The fact that they don't understand intelligence doesn't mean that it isn't a useful concept.

When people actually can't agree on the meaning of a word, the signnal to noise ratio drops from using it. But in that case, instead of discarding the word, people just need to standardize it.

An excellent example of a published paper against reductionism, using "emergence" in exactly this way such that it is indiscernible from "magic", is here:

Well, I agree that that fake explanation is used too often, and that it only gets any cred because it's from the right literature genre. But I don't think the whole of work in emergence can really be reduced to a mystery to worship. Certainly "emergence" is a stupid noun, just like "Red-hood" is a stupid noun. And that's a wonderful exercise to shut up the anti-reductionist movement based around emergence.

But "emergently arising" and "arising" can be given useful different meanings without stretching things too far, specially if "emergent" is contrasted with "resultant".

The origin of the modern concept of emergence can be traced to the middle of the nineteenth century when realist philosophers first began pondering the deep dissimilarities between causality in the fields of physics and chemistry. The classical example of causality in physics is a collision between two molecules or other rigid objects. Even in the case of several colliding molecules the overall effect is a simple addition. If, for example, one molecule is hit by a second one in one direction and by a third one in a different direction the composite effect will be the same as the sum of the two separate effects: the first molecule will end up in the same final position if the other two hit it simultaneously or if one collision happens before the other. In short, in these causal interactions there are no surprises, nothing is produced over and above what is already there. But when two molecules interact chemically an entirely new entity may emerge, as when hydrogen and oxygen interact to form water. Water has properties that are not possessed by its component parts: oxygen and hydrogen are gases at room temperature while water is liquid. And water has capacities distinct from those of its parts: adding oxygen or hydrogen to a fire fuels it while adding water extinguishes it.

-- Manuel Delanda

Delanda is one of those anti-reductionists that I was talking about, but nonetheless I still think he gives useful and viable meanings to "emergent" and "resultant" here , though I think his arguments against reductionism are just plain silly. They leave room for reducing any "emergent property" of a whole to the interactions of its parts, so as far as I can tell his arguments leave plenty of room for reductionism.

The word 'emergence' is an accent, not an explanation. It shifts focus to the idea that the system itself contains enough power or complexity to produce the effects wanted, when the mistake is to assume that the system doesn't have it. Let's show a simplistic example with ant colonies:

Ant colonies exhibit intelligence.
Ant colonies exhibit intelligent emergent behavior.

In the first statement, there is an easily ambiguated idea that intelligence is part of the ant colonies. This could mean one of many possible things in common speech:
- The ants themselves are intelligent
- The ants are not intelligent, but the colony is well-organized and they act as an intelligent body together.

Both are likely (I'm not an ant specialist) wrong. The problem is that ordinary speech emphasizes the two above interpretations, and deemphasizes the intended one below of "emergence":

- The ants may not be intelligent, and the colony is probably not well-organized. However, the simplistic rules in the ant interactions create a complex system that behaves just like an intelligent body.

That's a lot of space to save by just saying that the colony shows emergent intelligence. A similar accent describes what happens with evolution and emergent traits. The word is intended to shift the focus from the idea that "the two lifeforms developed similar traits as a /direct/ consequence of such and such conditions" to the idea that "under such and such conditions, the two lifeforms happened to develop similar traits".

The system, in this case evolution, more properly describes the mechanism than the situation, in this case the habitat. The word emergent only brings to attention that fact.

Eliezer, although the comments did eventually get better, don't despair for the early comments on this post. Remember yourself, all you are finding in the comments is evidence confirming the belief that no one reading this blog is learning anything. I conjecture that those who have learned something just don't get excited enough to post because they don't disagree with you strongly enough or aren't sufficiently surprised to thank you publicly.

Of course, I still suspect, as you probably do, from years of experience that most readers of this blog believe they are learning to overcome bias when in fact they are just convincing themselves that they are learning to overcome bias because they have read about it and believe it's a virtue to overcome bias. And I don't exclude myself from this group either, because although I don't feel as though I'm thinking this way, that doesn't mean I might not secretly be and I will be revealed when I discover a serious error in my behavior and beliefs.

Thanks, Eliezer. Regarding your questions:

  1. Is the property objective or subjective? The coarse grained property is objective -- e.g. the largest connected component in percolation. The meta-property that a coarse-grained property is emergent is as objective as the entropy of a configuration. It is model dependent, but in most cases we can't come up with a model that makes it go away.

  2. To the extent "emergentness" is subjective, it is because it is relative to a model. So in some cases it could possibly be the result of ignorance of a better model. But we can't claim "emergentness" due to ignorance of any workable model, we can only say "I don't know".

  3. Conjecturing that a property is emergent is a guide to inquiry. It is saying "Let's look for a model of this property that's robust under perturbation of the elements of the ensemble, but where the value of the property changes dramatically due to small changes in the average value of some properties of the elements. The model will be based on some highly simplified view of how the elements interact."

  4. Some observations about models of emergent properties:

    • We don't have a very good "toolbox" for building them yet. We're getting better, but have a long way to go before we know how to proceed when we conjecture that a property is emergent.

    • We are even weaker in the design of emergent systems. That is why even very simple designs with emergent properties, like flocking, seem so striking. This is a serious disability because we depend on systems with major emergent properties, such as markets, and we don't know how to manage them very effectively.

    • Often when people claim properties are "mysterious", we could dispell these claims if we could respond with an intuitive account of how those properties emerge. Lacking such an account, we are often vulnerable to mystification.
    • I ll try a silly info-theoretic description of emergence:

      Let K(.) be Kolmogorov complexity. Assume you have a system M consisting of and fully determined by n small identical parts C. Then M is 'emergent' if M can be well approximated by an object M' such that K(M') << n*K(C).

      The particulars of the definition aren't even important. What's important is this is (or can be) a mathematical, rather than a scientific definition, something like the definition of derivative. Mathematical concepts seem more about description, representation, and modeling than about prediction, and falsifiability. Mathematical concepts may not increase our ability to predict directly, but they do indirectly as they form a part in larger scientific predictions. Derivatives don't predict anything themselves, but many physical laws are stated in terms of derivatives.

      The concept of emergence is useful as a guard against certain errors, such as, for example, conspiratorial theories which explain phenomena as the product of intentions (malign or benevolent). Order does not always arise from intention. If society is lawful, that is not necessarily because there is some commander dictating that it be lawful. The lawfulness of society may be a phenomenon with a mostly dispersed, decentralized cause (e.g., lawfulness may be in large part enforced by ostracism of transgressors and thus enforced by all members of society rather than by an elite and privileged police apparatus applying the decisions of a legislature). Similarly, a sudden rise in the price of something may not be the product of a secret cartel or of a particular government policy or politician, but may be the result of a non-obvious confluence of disparate causes. When we say that prices emerge from the marketplace rather than being set by some authority, we are denying the idea that there is some person or group who has chosen that the price should be what it is, a person who can be appealed to to change his mind, or blamed for his decision.

      Don't y'all find it a little suspicious that so many people think "emergence" is a useful concept, yet have different definitions of what it is? (Though more important is what it isn't.)

      Next stage in the gauntlet: Why is this a useful concept? Why does it increase your understanding of the universe, and your predictive power? Can you force me to talk about emergence or a concept isomorphic to it?

      I think you are railing against the way emergence is used as opposed to itself as a theory.

      Emergence theory can be helpful in breaking down complex systems to gain a better understanding of how they work. If the behavior of a system is dominated by a decentralized and internal control mechanism, then it is emergent. If the behavior is dominated by a centralized or an external control mechanism then it is non-emergent.

      Let's take for an example the complex movement of a bee swarm. From casual observation, it looks to be a single entity that has an intelligence of itself. As a layman, upon looking at this, I could take a guess that perhaps it is magic, or the Queen bee is in full control somehow.
      If somebody tells me that it is in fact emergent, then I can understand and predict the bee swarm much more accurately. I can see that each individual bee's behavior influences it's neighbor's in such a way to produce a highly complex and sophisticated system. I will be able to model and predict the swarm much more accurately than if I thought it was controlled by the Queen via pheromones, for example.
      This theory also has the luxury of being testable.

      Also, we can use emergence as an approach engineering problems as well. Since video games were used as an example before, let's use them again. Say we wanted to make a game where the player is being chased by formations of monsters.

      One approach would be to create a computer AI to direct the monsters in a line and move towards the player. This would be non-emergent.

      Another would be to create the monsters as individual objects that move towards the next nearest monster and also towards the player but at a slower rate. This would tend create clusters of monsters that charged towards the player. Now, what if had the monsters move towards the player and the next nearest monster, but away from the second nearest monster. This would tend to create lines of monsters that moved towards the player.

      So, to force you to talk about emergence, I would like to propose a theory about your chosen profession, computer AI. (Caveat, I don't really know anything about it) Software AI* is non-emergent.
      Neural Networks are emergent.

      • This does not count software that is modeling the human brain on a neural level.

      Does an emergent bee swarm follow the same rules as an emergent market?

      That is, can you take the rules of emergence and apply them to both as is - including only the necessary details (like how many bees or the number of stocks)?

      If so, you have a testable theory. If not, you have phlogiston.

      As Wikipedia describes them, "weak emergence" seems to be what the people above are talking about - that is the properties of the whole can be described/predicted by the interactions of its parts, and "strong emergence" which says the properties of the whole cannot be described/predicted by the interaction of its parts - seems to be what EY is talking about.

      The notion of strong emergence, to me, seems nonsensical. I also see no formula for emergence of any kind, which suggests to me that it is not intended to make any predictions - only describe events after they occur. That sounds a lot like phlogiston to me.

      "Tell me what it isn't!"

      I'll go with TGGP's domain (video games), since that is what we blog about at Kill Ten Rats. The gaming blogosphere uses the term "emergent gameplay" more or less as TGGP defines it. Going back to my first online game, Asheron's Call, an example of what is not "emergent" gameplay is characters slaying monsters and leveling up. Monsters have the same code, but rarely win, so they rarely level; an example of emergent play was having characters sacrifice themselves to bunnies, who would gradually level up to a Night of the Lepus situation. Non-emergent was the use of "pyreals" (gold pieces) as currency; emergent was the economy that arose using alternate competing commodities as units of exchange.

      As others have stated, I usually see "emergent" used as a shorthand for something like "the whole is not predictable from the parts." No cell in your brain understands Chinese, and neither does any part of Searle's Chinese room, but the system as a whole does. Is there a better term akin to the fallacy of composition or ecological fallacy?

      I'm pretty ignorant on this, but I always thought that the phrase related to complex outcomes that result from surprisingly simple systems, so that the complexity is "emergent".

      One example is chaos. One can have chaotic non-linear dynamic systems and non-chaotic non-linear dynamic systems.

      But, again, I could have misunderstood.

      There are a few examples of non-emergence. For example, if we tessellate many small equilateral triangles to create a larger equilateral triangle, the resulting figure will not show any emergent properties.

      Outside of mathematics, though, the concept is vague and I can't see much use for it as applied to specific phenomena.

      Aren't superconductivity and ferromagnetism perfect examples of emergent phenomena? I'm not saying that calling something an emergent phenomenon adds any deeper understanding of it. But I think there certainly are phenomena that can be fairly called as emergent.

      Saying "X is emergent" is conveying some information, if there is someone in the room that does not already know this fact. Here is an example:

      Quarks are emergent.

      This is not an explanation though. It is more like a anti-explanation. I just claimed that there is an underlying explanation to quarks, and then stopped. I told you to make space for an explanation, in you mental world model, and then I left you with that space empty. If you believed my statement, and if you don't already know how quarks emerges and from what, I just made an explanation shaped hole in your mind. This is not nice of me.

      But at least you now know that there is an explanation to be found. When you thought quarks was fundamental, you did not even know to look, because fundamental things can not be explained, only described.

      Whilst I appreciate the validity of criticism offered here of the use of the word emergence (by itself) as if were an explanation sufficient unto itself - I think it a little harsh. To call it "futile" is almost acting as semantic stop sign itself for the term.

      We need to take a little time to properly understand what is meant by emergence when used properly.

      First that it is an observation rather than an explnation. But an observation with useful descriptive power since it observes that the phenomena under consideration is a process with properties whereby larger entities, patterns, and regularities arise through interactions among smaller or simpler entities that themselves do not exhibit such properties.

      Therefore not at all properties that arise from interactions or combinations of smaller components are emergent (e.g. putting a whole bunch of magnets together just gives a larger magnetic field). So not all things arise are emergent.

      So, while "emergence" is hardly an explanation - and one is obliged to look for the mechanisms that lead to the emergent behaviour - (such as how the polar hydrogen bonds in H20 give water surface tension - a property that a single H2O molecule does not exhibit) - nevertheless it's use as an observation has power since it points us to look for (and ask question about) how properties which do not exist in the sub components come to be via the interactions of the components (often multi-factor) - and also to see if there are simple factors or descriptive rules than have predictive power (e.g. flocking phenomena of birds)

      I would suggest reading "More is Different", an excellent paper on the topic of emergent phenomena and the limits of reductionism. (

      The "More is Different" approach essentially notes that even when the basic bottom-up rules are known, they cannot be efficiently used to predict large-scale interactions. Those behaviours have to be studied as though they have their own set of rules (ie, the laws of chemistry), even though they emerge from a more fundamental set (the laws of physics). It seems to me that labelling a phenomena as "emergent" is a useful and descriptive term, which means "you need to study this phenomenon empirically to figure out its rules, because trying to make predictions based on fundamental principles won't get you anywhere".

      that description fits every phenomenon in our universe above the level of individual quarks...

      What do you know, after you have said that intelligence is "emergent"? You can make no new predictions.

      It is true that I can make no new predictions, and you can make no new predictions, but other people do make new predictions from the explanation "emergence".

      "Emergence" is like "atheism" or "naturalism".

      If a religious person learns only that a set of phenomena is the world is explained by naturalism, that person has had his or her anticipations constrained. That person now knows that the actual explanation is one of those that don't involve the supernatural. This is only a benefit to this person because he or she had been unnaturally focused on the narrow set of states of the world in which there would be a theistic explanation of those phenomena.

      Someone who believes all or many things are explained by real magic will describe something's being explained entirely reductionistically as being explained by "emergence". It is people recognizing and appreciating the mundane magic of some phenomena while believing in real magic as the cause of other things. To tell such a person that the cause is emergence is to tell them to make models at different levels and not bother looking for the supernatural as a cause.

      As was said, "I don't know" can be.the best way to communicate, so too can be "emergence".

      I just came from a debate with a friend of mine about emergence, so here's a simple example of what emergence is and isn't that I just told to him. (That he rejected anyway.)

      Let's take, as an example, a car. Motion is an emergent property in cars. (I'm talking about motion on the level we live on that allows whole objects to move great distances.)

      The pieces of the car, gathered into a pile, could not move. So motion was not a property in the parts making up the car. Motion emerges when the parts are built into the complex relationship that makes up the car.

      That's an example of what emergence is. Here's just as simple of an example of what it isn't.

      Let's say I strapped a bunch of cars together to make a long train of cars traveling down the freeway. One car might be towing the whole train. Half of the cars might be contributing to the motion of the car train. All of the cars might be running and contributing to the motion of the train. But the motion is not an emergent property of this car-train.

      Motion emerged at the level of the individual cars. Motion is also a property of the train, but that's not where the motion emerged. Motion is therefore an emergent property in cars, but not an emergent property in car-trains.

      Emergent properties being "Properties possessed by the whole that the sum of its parts did not possess" basically means "This property emerged here, and not a level higher. And it was not present a level lower."

      This creates a huge number of falses for a huge number of car-train configurations when testing for motion as an emergent property.

      This thread appears to be missing references to support the notions that "emergence is commonly used as an explanation in its own right" and "many people think emergence is a useful concept, yet have different definitions of what it is".

      "Intelligence is an emergent phenomenon" is a valid response to Searle-followers - who ask questions about how brains can be intelligent when no neuron is intelligent. AFAIK, the response doesn't pretend to be a complete theory about how brains work.

      Some other vague concepts people disagree on: 'cause,' 'intelligence,' 'mental state,' and so on.

      I am a little suspicious of projects to 'exorcise' vague concepts from scientific discourse. I think scientists are engaged in a healthy enough enterprise that eventually they will be able to sort out the uselessly vague concepts from the 'vague because they haven't been adequately understood and defined yet'.

      "As a result, the theory wasn't scrapped;"

      By "the theory" you mean general relativity, which is one of the most well-confirmed theories in all of physics. You can't just come up with a slightly modified version of GR to accommodate weird observations; the Einstein field equation is a unique solution because of all the demands placed on any reasonable theory of gravity. If you assume:

      • Spacetime is flat in the absence of matter;
      • Spacetime curvature is linear with respect to the density of matter;
      • The standard principles of mathematics (eg, two matrices with different dimensions cannot be equal);
      • The laws of physics are invariant under coordinate transformations (no preferred coordinate system); and
      • Spacetime does not have an a priori curvature not affected by matter;

      you are forced to use general relativity.

      In line with previous comments, I'd always understood the idea of emergence to have real content: "systems whose high-level behaviors arise or 'emerge' from the interaction of many low-level elements" as opposed to being centrally determined or consciously designed (basically "bottom-up" rather than "top-down"). It's not a specific explanation in and of itself, but it does characterise a class of explanations, and, more importantly, excludes certain other types of explanation.

      This comment hits the bullseye. The general idea of emergence is primarily useful is in pointing out that when we don't understand something, there are still alternative explanations to those that superstitiously posit a near-omniscience or that pretend to have information or an ability to model complex phenomena that one does not in fact have. So, for example, a highly improbable organism does not imply a creator, a good law does not imply a legislator, a good economy does not require an economic planner, and so on, because such things can be generated by emergent processes. To come to such a conclusion does not require that we have first reasoned out the specific process by which the object in question emerged. Indeed if we had, we wouldn't have to invoke emergence any more but rather some more specific algorithm, such as natural selection to explain the origins of species.

      For this reason, I strongly disagree with the following definition

      Let K(.) be Kolmogorov complexity. Assume you have a system M consisting of and fully determined by n small identical parts C. Then M is 'emergent' if M can be well approximated by an object M' such that K(M') << n*K(C).

      Because it is just in situations where a phenomenon has a not highly reducible complexity -- where M is not fully determined by n small identical particles, or where it is but K(M') is not substantially smaller than n*K(C) -- that the idea that a phenomenon is emergent, rather than the product of a near-omniscient or near-omnipotent creator, is most useful.

      I'd add that the belief that any important phenomenon is highly reducible, or that even if it is reduceable that humans are capable of undertaking that reduction, are two other species of superstition. These are just as pernicious as the related superstition of the near-omniscient creator. In many, perhaps most cases of interest we either have to be satisfied with regarding a phenomenon as "emergent" or we have to superstitiously pretend that some being has information or a capability of reduction that it does not in fact have.

      Emergence is an annoyingly vague concept, but that doesn't mean it's an empty one.

      One meaning of emergence is "decentralized control". In a free market economy, prices and other properties are emergent from a large set of transactions among distributed agents, in contrast to a centrally planned and controlled economy. So there's something that is not emergent, or less emergent. Similarly, it used to be thought that a bee colony was controlled by the queen, but now we know that its activity is also the result of the work of distributed agents. It didn't have to be that way, so we have found an instance of emergence where we might have found something else. In AI, there are emergentist approaches like neural nets or genetic algorithms, or the old-school symbolic approach which is top-down and designed.

      So it's clear that emergence is not meaningless. But you are right, it is not an explanation for anything. If you've found that a bee colony is decentralized, you still have to tease out the rules and interactions of the various agents involved.

      "It seems as if every new observation is "surprising," or "shocking," or "baffling." "

      Surely the underlying reality of this "seeming" is that a large fraction of new newsworthy observations is "baffling" etc.

      Jadagul's example seems to me to be a clear place where the term emergence is useful. Phil Goetz has given others in the past. OTOH, it still seems that in most of the cases where emergence is used as a synonym for "magic" much too often. 'Emerges from' seems to be less strong evidence for a legitimately useful term than than 'emergent', as 'chaotic' seems to be a perfect synonym for the latter.

      Even in the case of 'chaotic', the tendency to use the term as a stop-sign is serious. A great deal of understanding of chaotic systems is possible (they are probably most of what we understand in the world, after all), just not precise long-term prediction of their configurations.

      Vitalism seems to clearly constitute a "stop sign", but I'd want much more expertise before confidently asserting that proto-chemists didn't use Phlogiston to make novel predictions similar to those we would make with Oxygen. It seems to me like Phlogiston is a conflation of Oxygen and Energy similar to Newtonian "mass" as a conflation of gravitational and inertial mass, or pre-Newtonian "weight" as a conflation of weight and mass.

      If we take EY's example of a market crash, the non-emergent hypothesis is that of a random process. In other words, price action is like the action of gas molecules, or Brownian motion. To say that a market crash is an emergent phenomenon is to say that it displays more order than gas molecules bouncing off one another, which do not display emergent properties. That is not an empty distinction, as far as I can see.

      You can also add the adjective 'emergent' into any sentence, and the product will be comparative garbage.

      "Intelligence is an emergent phenomenon!" means that intelligence didn't happen on purpose, or that intelligence doesn't need to be intentional in order to happen. Emergence as a term doesn't add a reason for a thing, but it does rule some out.

      An important point that I think this raises is that answers which are otherwise not mysterious can be made mysterious if they are used as such. One good example is the way that some laypeople use "quantum mechanics" interchangeably with "magic". This doesn't invalidate quantum mechanics in its correct usage. The same could be said of emergence, which, given the comments below, seems to at least be disputed as a mysterious answer in and of itself.

      EDIT: Also, this is my first post (reading through the Sequences, getting acquainted with the basics, etc.), so feedback is appreciated. I hope my contribution, well, contributes!

      This is very curious. I never thought of emergent as an explanation but as a property. I roughly understood it to mean that the emergent quality was transferable. That is, intelligence is a product of neurons firing but it need not have been, it could also have been generated from transistors or whatever else.

      This is roughly the opposite of your ant example. Something is emergent if it can be explained/predicted with no knowledge of the lower level. A lot of properties of turing machines do not depend on the actual formalism of the turing machine.

      Edit: After browsing the other comments, I realize this is something that has been brought up before. My 2 cents for whatever it is worth, I guess...

      Wolfram has done a lot of fantastic work on emergent mathematical phenomena. (TED talk given to a non-technical audience, but still worth watching.) One of the highly counter-intuitive things that he has worked on is computational irreducibility. Irreducible functions are ones where you have to physically run the function to find it's outcomes, and the emerging patterns. For this class of function, the emergent patterns cannot be predicted in advance.

      There seems like the next step to build on older work on the halting problem, which states that some types of problems require unknowable amounts of computational power to solve. It's not possible to know whether such problems will be solvable in a finite time, or whether the program will run forever without finding an answer.

      This is relevant to Yudkowsky's criticisms of emergent phenomena, because it demonstrates that emergent phenomena are not just fake explanations in these cases. The term is used to describe a specific class of mathematically defined problems, which are irreducible. These problems can't be broken down to any more simple explanations. In these instances, the phrase "emergent" isn't a semantic stopsign telling someone not to ask any more question, but rather a useful marker letting someone know that they physically can't break the problem down any simpler without breaking the rules of mathematics.

      To be fair, Yudkowsky does go through great pains to specify that

      It's the noun "emergence" that I protest, rather than the verb "emerges from".

      But that's not quite the right distinction to make. Perhaps the noun form is more often used as a meaningless buzzword, and has been used to refer to larger and larger groups of things. Maybe you could argue that "emergence" should only be used to refer to this formal case where complexity arises from simplicity in an manner which may be impossible to predict in a finite amount of time. But, by definition, we can't actually know that any particular problem won't halt, because the test would take an infinite amount of time.

      So do we just use this "emergence" word to refer to any problems which haven't yet been computed from more fundamental principles, or do we use it more generally to refer to problems where the math is just really complex? Personally, I'd lean toward the latter. The term is already in common use to describe complex systems which arise from simple ones. The term does still become meaningless if we use it to refer to slightly complex systems emerging from fairly simple ones, but it's still a useful and descriptive word for other cases. We shouldn't stop using terms like "toxins", "energy", "quantum", or "exponential" just because they have been re-purposed and watered down, so why should we do so with "emergent"?

      In summary: emergence is sometimes an observation but never an explanation.

      "Emergent" is just an adjective describing an attribute. Other examples are complex, simple, generic, unique, random, predictable, valuable, politically inconvenient, unexpected, widely-known, and a few others. For example, saying "The behavior of the ant colony is the widely-known outcome of the interactions of many individual ants.", won't tell you much new about the ant colony itself, nor will it let you model it. It will tell you that ant colonies aren't successfully secretive, nor too complex for humans to understand, which technically is details about the ant colony's behavior.

      Likewise, knowing that ant colony behavior is "emergent", won't let you model the ant colony. It will tell you that ants interacting in simple ways can lead to seemingly complex and perhaps unexpected behavior. But it won't tell you what rules govern the interaction of individual ants, nor of the whole colony. It does tell you that if you replaced half the ants with other compatible ants, rearranged the ants, or similar disturbances, there'd be little difference in overall behavior of the colony. You see, it is a description of the model of ant colony behavior, and not a model of ant colony behavior. It's use as an explanation or curiosity-stopper is inappropriate, but not any more so than any of the adjectives in the above paragraph.

      It would be inappropriate to describe ant colony behavior as being and "emergent" property of atoms. You can't rearrange the atoms, the atoms will not continue to act like a ant colony. Conversely, rearranging and replacing cars and stock market participants, will do little to change the traffic patterns and stock market (minus a period of adjustment). As for human intelligence as an emergent property of neurons, it is a statement of fact but I do not know if it is accurate. The hierarchical structure and specialized regions, would seem to suggest more nuances than emergent behaviors are supposed to have. Neural rearrangement and replacement, on a scale of half the neurons at once, can't be done for practical nor ethical reasons, and would be expected to delete years of experience, and that without the development phase -- but it might still result in intelligence, especially if done on a fetus. I could easily see people hoping that intelligence is emergent from neurons, as it would make for a simpler model and thus cause an earlier advance in AI and intelligence related fields. I wouldn't fancy trying to figure out neuron-based intelligence if it is not an emergent behavior.

      I have always been intrigued that such a complex system like a living cell, could be reduced underlying physics and chemistry. Over time, my reductionistic curiosity was eroded by holistic views that embrace emergence phenomenon as an explanation of complexity. However, eventually, I was disillusioned with the emergence paradigm, as misleading and concluded that the popular interpretation of holism ‘The whole is more than the sum of its parts,’ is profoundly deceptive if used within scientific explanation. My current view is that emergence is a perception caused by part’s properties not observable in isolated parts. These properties become observable only during interactions in the system; a system acts as ‘litmus test’ or a ‘magnifying glass’ that just reveals the parts’ properties not observable otherwise.

      Regarding ant colony how much we know about individual ants to deprive them from ability of complex behavior? In my book the complexity of any system resulted from collective complexity of its elements.

      The typical rhetorical argument in favor of emergence is the question: ‘Is water more than one atom of oxygen and two atoms of hydrogen?’ The intuitive respond, is yes, because in our perception, water, the way we directly experience it, is very different from an abstract theoretical model of atoms of oxygen and hydrogen. However, from a scientific point of view this question is misleading. The correct question would be: ‘Is a molecule of water more than one atom of oxygen and two atoms of hydrogen, interacting among each other?’ This time the answer is no: the molecule of water is no more that the sum of its components.

      I would be interested in any example in which complexity of the system "emerges" from elements we know (or at least we believe we know) everything about.

      If anything, a single ant is a far more complex system than, say, a simple simulation of an ant colony of a computer.

      In this case it is obvious. However, in general a system could be less complex than complexity of its elements. We don't have agreeable way to measure complexity but for now I would argue that complexity of molecule of water (its complete physical and chemical description) is more complex than sum of those molecules in form of a drop of water. Unfortunately so far there's neither accepted approach to measure nor define complexity.

      I was under the impression that a property x was emergent if it wasn't determined by the set of property states of the components. IE, gravity isn't emergent since the gravity generated by something is the addition of the gravity of the parts. Intelligence isn't, because even if I know the intelligence of each of your neurons, I don't know your intelligence.

      Observation of individual neurons doesn't indicate they have intelligence however doe it means that intelligence of a human brain is emergence phenomenon?

      Observation of individual atoms and molecules wouldn't revel any gravitation like properties either however we don't call that gravity emergence phenomena. Instead we argue that gravitation like properties of atoms and molecules are not observable. Could you conceder that we may grossly underestimate an "intelligent ability" of individual neurons?

      What I meant by this is the gravitational influence of N particles is the sum of the gravitational influences of each of the individual particles, and is therefore a strict function of their individual gravitational influences. If you give me any collection of particles, and tell me nothing except their gravitational fields, I can tell you the gravitational field of the system of particles. If you tell me the intelligence of each of your neurons (0), I cannot determine your intelligence.

      I believe I first came across the term emergence in relation to Langton's Ant, in a book section loosely centered on game theory. In the book, the patterns formed by the progression of Langton's Ant were termed 'emergent' because they could not be predicted except by running the program. One could not, with full knowledge of the simple rules that governed the Ant's world, predict what the pattern would look like after x iterations, or where the ant would be. Given this, I would not call the location of a dropped object at time t after the drop an 'emergent property', because given the laws that govern how it will fall and the initial values of the system (height and mass of object, area perpendicular to motion, density of air etc etc) I can accurately predict the object's location and momentum without having to actually drop it and see where it is.

      So an emergent system is one wherein the simple base rules do not enable one to predict the large scale order or consequences (I suppose lack of order too), whereas a system which is not emergent is one wherein knowledge of the governing rules enables a prediction which can later be proven by running of the system. Prime numbers are thus an emergent phenomenon of how numbers related to each other (unless someone has found a way to predict them while I wasn't paying attention), whereas even numbers are not.