My fellow bloggingheads John Horgan and George Johnson took some time on their latest science talk to dissect my New York Times article on swarms (you can jump to that section here). John wonders if I'm just discovering all the complexity stuff he and George were writing about back in the 1990s. I think it's always good for John to keep everyone aware of the dangers of hype, of the need to ask how important or new scientific research really is. He's been particularly tough on the science of complexity, if there is such a thing. In 1995 he wrote a piece in Scientific American that practically brought tears to the eyes of many scientists who thought complexity was the Next Big Thing.
While John makes some decent points, I think he ends up playing a game of bait-and-switch. He doesn't have anything to say in particular about the research of Iain Couzin, the subject of my article, such as his investigations of army ants or locusts. He's not actually interested in ants or locusts, as far as I can tell--the complex systems "we're really interested in," as he puts it, are human societies, economies, etc. In other words, if some scientists make a big deal about how they've discovered the hidden rules of human societies but don't have more than simple computer models, then I guess the whole field becomes tainted. Everything becomes a simulation with a vague resemblance that could be a coincidence.
If John is not particularly interested in real swarms, so be it. If he doesn't think it's useful to figure out the rules that help trigger locust invasions that destroy vast areas of cropland, c'est la vie. But I think he needs to look more closely at Couzin's actual research before decrying the failure of complexity to become the next Newtonian revolution. It's true that scientists in the 1990s were trying to understand swarming behavior, as they are now. But does that mean that nothing's changed? No. Just compare Couzin's work to Craig Reynlod's work in the 1980s on "boids." While these flocking simulations were provocative and intriguing, they were simple and didn't correspond precisely to any real species.
Now people like Couzin are learning how to go into the field, observe real animals, extract the rules they follow, create simulations on computers with the help of new mathematical equations, and come up with experiments to test predictions of those simulations. So now it's possible to figure out how Mormon cricket swarms are fundamentally different from locust swarms. The latter are just trying to move together. The former are on a cannibalistic forced march. That's not just vague resemblances or coincidences. That's the real, albeit slow advance of science.
Complexity as a field, it seems to me, is a lot more mature than the hyped field which had hits hayday following the publication of "Chaos." (In other words the backlash was merited, but that doesn't mean that there aren't some core ideas which are now forming the basis of very solid research.)
Interestingly one of the big NSF programs being rolled out this year, Cyber-Enabled Discovery and Innovation (CDI) has one of its three prongs square in the "complexity" setting:
Understanding Complexity in Natural, Built, and Social Systems: deriving fundamental insights on systems comprising multiple interacting elements;
My paycheck comes from the New England Complex Systems Institute, so I'm doubtlessly biased in this, but I think there's a good argument to be made that what we call "complex systems" did not exist as a field until 1999. The one thing which everybody agrees is part of "complex systems study" is the investigation of networks, and the modern approach to networks got kick-started by the Barabasi-Albert model of preferential attachment. This was a re-discovery, as so many things are, but it brought an idea forth at just the time when large-throughput data collection made that idea applicable to real problems.