Copyright: 2001
Publisher: Scribner
ISBN: 068486875x
Steven Johnson explores the phenomenon of emergence in Emergence: The Connected Lives of Ants, Brains, Cities, and Software. Emergence theory is the study of how disparate local knowledge and local decisions can lead to higher level (or global) intelligence and global structure. We are programmed to think that without an overseer or man-in-charge, groups cannot self-organize. Drawing on examples as diverse as ant colonies to the rise and structure of the city of Manchester, UK, Johnson shows how emergence is not only possible, but is a very important part of our everyday life.
Between Formulas and Statistics
Scientific inquiry is generally divided into one of two camps. Either you devise a formula from which you can calculate an expected result or you use statistical analysis to predict an expected result. In the case of simple systems with only a few variables, scientists and mathematicians are quite adept at reducing problems to their formulaic equivalents. In the case of disorganized complexity, statisticians are veritable magicians at reducing vast expanses of data down to a handful of descriptors.
Emergence theorists however are concerned with the middle ground between these two extremes. Not simple and not disorganized: organized complexity. This is the domain of emergence and the common tool used by practitioners is modeling software. StarLogo is the system most often discussed in this particular book but there are others as well. The idea is that you model the behavior of individual agents, turn them loose in a system with rules that guide rather than direct and watch to see what happens.
To describe how emergence works, Oliver Selfridge used a concept pulled from literature: Pandemonium, the shrieking demons of Dante's Inferno. He envisioned layers of demons (agents) that passed their best knowledge up a hierarchy of demons. Higher level demons aggregated the best knowledge from below and passed only the best matches up the hierarchy. This method of pattern recognition at a very basic level that passes it's learned knowledge up the hierarchy underlies much of the software today that does hand-writing, voice or character recognition.
What Does a System Need to be Emergent?
Systems where macro-intelligence and adaptibility derive from local knowledge and local decisions follow a handful of principles:
- More is Different (Systems change as they grow)
- Ignorance is useful (Agents don't need to know everything)
- Encourage Random Encounters (Agents need to communicate with many other agents)
- Look for patterns in signs (Patterns of signals can be as important as the diversity of signal types)
- Pay attention to your neighbors (Local information can lead to global wisdom)
Random Notes
- Cities were the original "repositories" for human knowledge. Now we have the web and other repositories.
- Positive feedback is only marginally useful. Negative feedback however can turn a complex system into an adaptive complex system.
- The early days of slashdot.org were instructive in the usefulness of negative feedback in self-organization.
Conclusion
This was a good little book to serve as a primer to the field of emergence. Like many tech writers, Johnson gets in over his head in the last third of the book. (In fact, most books I read seem to get in over their head as they try to eek out another 75 to 100 pages... but that's another story.) Johnson at one point makes fun of earlier tech writers who predicted the demise of the city because we would all be working from home at our ranches in Wyoming by the turn of the century. He then turns around and predicts emphatically that by 2006, all TV's sold would come with built-in TiVo and there would be "millions of channels" on TV deliverd over the web. We are sort of getting close(ish) but it's now 2011 and well... it hasn't happened yet. All in all though if you ignore his confident and wild predictions about where emergence is going to take us, the book is informative and useful.