Emergence: From Chaos to Order by John Holland, Addison Wesley/Longman, $25, ISBN 0201149435
JUST what is a 鈥渓evel of description鈥�? Consider the lily, from the bottom up. You have your quark, nucleon, chemical, genetic and protein level. But we are in mushy territory: does gene regulation form a distinct level from the 鈥渞aw鈥� genome of proteins? Paddle on upwards, waving your arms with increasing vigour to keep your balance. You might identify cell morphology, cell function, metabolic regulation and reproductive 鈥渟trategy鈥� as levels. Because you are exhorted to consider 鈥渢he lily鈥�, there is no reason to stop at 鈥渁 lily鈥�: speciation and ecology, references metaphoric and poetic, and theology all enter the picture.
Now consider the entity which is considering the lily. Is this 鈥渁n emergent phenomenon鈥�, or many? And what is one of those, other than a string of letters which evokes an image of the acme of arm-waving? In Emergence, John Holland of the University of Michigan and the Santa Fe Institute (and the father of the genetic algorithm and founder of complexity studies) makes a determined and provocative attempt to introduce some rigour.
Advertisement
Intuitively, the concept of emergence is not very difficult to grasp but analogies 鈥� the game of chess is an emergent property of the rules of chess 鈥� seem to lack depth. In principle, chess could be reducible, leaving aside small problems such as the mass of the Universe being insufficient to build the necessary computer.
If emergence means anything, it鈥檚 more than a synonym for 鈥渧ery hard to describe reductively鈥�. But saying that consciousness is an emergent property of the functioning of neurons is too big a leap to be meaningful. As Holland notes, it can serve as an excuse for vitalism or mysticism. Somewhere in between lies a phenomenon that really can鈥檛 be reduced to the behaviour of its components if viewed separately, and really can be pinned down to a specific pair (or set) of closely related levels.
In attempting to define what 鈥渁 level of description鈥� is, Holland introduces us to the 鈥渃onstrained generating procedure鈥�, or cgp, within which a 鈥渕echanism鈥� is a transition function operating on a set of states and a set of inputs to produce a new set of states. So far, so thoroughly reducible. By chapter 10 he is ready to conclude: 鈥淲e can produce hierarchical definitions of cgps, using cgps defined early in the process as building blocks for later, more complicated cgps. We gain, thereby, a precise notion of level.鈥�
Really? Certainly, the picture of mechanisms built of mechanisms, all the way down, is helpful. But to this reader, happily corrupted by flaky French philosophers and Douglas Hofstadter鈥檚 Aunt Em, the word 鈥渉ierarchical鈥� demands 鈥渜uestion me!鈥� What if our mechanisms define each other in a tangled web? Then we will often find self-reference in the skeins of definition, and will need a more subtle notion than a hierarchy.
The reader鈥檚 corrupting influences do seem to be relevant. As for the author鈥檚, Holland includes Gell-Mann鈥檚 The Quark and the Jaguar, Hesse鈥檚 The Glass Bead Game and Hofstadter鈥檚 G枚del, Escher, Bach. He does not include Kauffman鈥檚 The Origins of Order or Wittgenstein鈥檚 Philosophical Investigations, either of which might have given pause for thought about the assumption of hierarchy.
The problem of level permeates Holland鈥檚 writing. After 30 opening pages of genteel schoolmasterly exposition, he hits the reader with: 鈥淟et X = {x1, x2鈥�, xn}鈥� and suddenly we鈥檙e into the main course. His closing chapter risks us losing our lunch on a roller coaster of level switching. This breathless chapter provides half-a-dozen research programmes, a plea for a millennial defence of the scientific method, and a handful of titbits for those who still enjoy ridiculing 鈥渟ystems theory鈥�.
The point is that reading, and arguing with, Holland is as likely as anything to produce progress on these crucial questions. Do it.