Friday, October 12, 2012

My position on emergence, as a physicist

Rationally Speaking is starting a new series on emergence.  So far, I like it merely because it starts out by talking about Renormalization Group theory, which comes from condensed matter physics.  Finally, the philosophers are talking about my field, rather than all that stuff about cosmology and particle physics.

I should take this opportunity to explain my position on emergence as a condensed matter physicist.  And yes, here I am speaking as a physicist, not because my study of physics has compelled me to view it one way or another, but because physics has greatly influenced my view.  I can definitely imagine another physicist coming to the opposite conclusions as me, but surely their opinion would also be greatly influenced by their study of physics.

In a way, condensed matter is all about emergence.  Condensed matter is about throwing ~100,000,000,000,000,000,000,000 atoms together, and trying to predict what they will do.  It is not an easy task.  Consider: a single helium atom is already an unsolvable problem, because it's just too hard to solve the quantum mechanical equations for the nucleus and two electrons.

How do we solve the problem?  Approximation upon approximation upon approximation upon approximation etc.  Really, there are too many approximations stacked up to fully comprehend them all at once.  One could say that condensed matter physics is the art of making approximations, and using experiments to test that they're sufficiently approximate.

An approximation is all about throwing out information, clearing out some of those irrelevant details so that we can zoom out and see the big picture.  In fact, Renormalization Group theory is literally about zooming out.  In the theory, we zoom out by some scaling factor--say, if we removed every other atom, and the remaining atoms were 1 micrometer apart from each other, instead of 0.5 micrometers.  What laws governing this sparser atomic lattice would cause them to behave just like they did in the original atomic lattice?  That is, how are the parameters, such as the coupling strength between atomic neighbors, affected by zooming out?

Renormalization Group theory explains the behavior of phase transitions (like gas to liquid) by showing that when you zoom out on a liquid, the parameters change in a different direction than when you zoom out on a gas.  It's a very powerful theory.

On Rationally Speaking, emergent behavior was defined in the following way:
The idea being that a phenomenon is emergent if its behavior is not reducible to some sort of sum of the behaviors of its parts, if its behavior is not predictable given full knowledge of the behaviors of its parts, and if it is somehow new.
I'm sort of on board with emergence, but I'm sort of not.  Emergent behavior is everywhere, and in particular it's right in front of me on my desk every work day.  Call emergence an illusion if you will, but by that standard so is superconductivity, which IMHO is a pretty hard-sciencey thing.  I suppose I'm taking the slippery slope here:  I think superconductivity is a real thing, therefore emergence is a real thing, and therefore even highly-emergent patterns like the stock market and the internet are real things.  Yes, I'm willing to bite the bullet and say that the internet is real, despite appearances to the contrary.

But at the same time, I think the standard way of describing and understanding emergence is all wrong.  As with the above definition, emergence is about something new that appears in the big picture.  But having worked with emergent systems, I do not think it's about adding something new.  It's about taking information away.

Perceived patterns do not indicate a greater amount of information, they indicate less information.  Usually, a pattern would consist of many things repeated over and over.  Repetition and redundancy do not convey more information.  Repetition and redundancy do not convey... oh, you get it.

When you find emergent pattern that is difficult to predict, the problem isn't that you can't reduce it to fundamental physics.  The problem is that you have to reduce it far, far beyond fundamental physics.  You have to eliminate large swaths of useless information in increasingly creative ways.  You have to do experiments on all levels to make sure that you didn't lose any of the important information.

Alternatively, you can take a more phenomenological approach, and "guess" the result of the information-elimination based on empirical observation and intuition.  I get the sense that this is what many people call emergent behavior, because by taking guesses they're adding something new.  Guessing is an tried-and-true practice used everywhere in science, so I'm cool with that.

1 comment:

Larry Hamelin said...

One might say that emergence is something that's very difficult to talk about precisely at the micro level but is much easier to talk about at the macro level.

Of course, my view of emergence is strongly conditioned by my training as an economist.