« ANOVA | Main | Losing the night sky »

August 30, 2007

WNYC on emergence

Radio station WNYC did a 60 minute segment on "emergence" by Jad Abumrad (of WNYC) and Robert Krulwich (of NPR's Science Desk) which is quite well done. Well worth a listen.

The segment features many of the usual suspects. Steve Strogatz talks about firefly synchronization (for instance, did you know that fireflies in Thailand gather in thousands and lightup in time with each other). Deborah Gordon talks about the wildly complex behavior of large groups of ants as compared to the stupid behavior of single ants. Steven Johnson, author Emergence, features several times. And my colleague Iain Couzin makes an appearance to talk about collective behavior. Toward the end, Christof Koch makes an appearance to talk about his collaboration with Francis Crick to figure out how consciousness works. Quite a cast, and Jad does an excellent job of weaving the clips and segments into a compelling whole.

Throughout the piece, there is the consistent theme that the behavior of some systems seems to be more than just the sum of the parts, at least from the typical reductionist perspective that's served science so well in other areas. Ants and brain cells are examples of things that, by themselves, are pretty dumb, but when put together in large numbers, groups of them can acheive complex behavior that often seems highly intelligent. The fundamental question is, How do they do it? Scientists want to know because many of life's complexities (like your brain) are clearly examples of "emergent complexity" and engineers want to know because systems that are emergently complex seem to have other features that we'd like to build into our machines. Characteristics like robustness to failures, adaptability, etc. (Familiar themes, for sure.)

I suspect that physicists are right when they say that emergence has a lot to do with phase transitions, but I think there's more to it than just that. It's probably partly my own biases and training to believe that emergence has a lot to do with computation (of the distributed sort), and thus that computer science is likely to have a great deal to offer. That's the idea at least, and so far, so good.

posted August 30, 2007 04:24 PM in Complex Systems | permalink


Emergence remains a tricky subject, since it is largely in the eye of the beholder. Only the observer can discern levels of abstraction where complexity arises from a lower, more basic, level. With that said about the philosophical debate on emergence, two things you might find interesting. Firstly the idea of metasystem transitions. I presume this is analogous or even the same as phase transitions, but I'm no physicist so I can't tell.
Secondly I like the example of emergence in formal grammars as described in a paper by Kubik. The idea is easily grasped; two grammars can produce their terms, but when the grammars are combined, new terms can be produced that neither grammar could individually. The caveat is that the grammars need to be different to produce anything interesting when combined, but still it's a nice, semi-formal, interpretation.

Posted by: Max Hinne at September 2, 2007 04:47 AM

Alex Ryan has a nice paper with a normative definition of emergence that isn't defined in terms of "levels of description". This paper is part of ongoing work between DSTO,
CSIRO and a few others (notably Cosma Shalizi), most of the project is detailed here

Posted by: Matthew Berryman at September 2, 2007 04:37 PM