Tuesday, 15 April 2008

emergence and reduction

Over at BackReaction, Bee has been musing about the distinction between model and theory, and also about the concepts of emergence and reduction. She makes some good points, but I'm only going to quote one:

So I think I leave this domain over to philosophy.
My investigations of reductionism and emergence have so far focused on the three varieties of emergence I consider the most promising, one trivially true and two others whose value has been partially obscured by careless misuse of terminology: I call these “process” emergence and “multiple realizibility.”

Recent results of nonlinear dynamics have reinvigorated the debate over emergent properties. The claims of emergentists and their opponents have become confused by contradictory and inconsistent definitions of emergence. Emergentists usually claim that a whole can have properties that cannot be explained or predicted by reference to its parts and their organization. This definition is problematic: an ontological claim turns crucially on the epistemologically loaded terms “explanation” and “prediction.” As a result, even the most fundamental interactions (say, those between quarks and leptons) automatically count in this trivial form of emergence. Contemporary emergentists mean to do more than point out a conflation of terms common to simplistic versions of reductionism; the point of my recent investigations have been to discover what they do mean.

Most contemporary emergentists desire a strong version of emergence, one clearly ontological in nature. Some philosophers have tried to leverage the sensitive dependence on initial conditions exhibited by certain nonlinear dynamical systems into a general defeat for microreductivism, a conclusion that is largely repellent to practicing scientists and engineers. This is a mistake: in fact, sensitive dependence supports epistemological emergence but not ontological emergence. In fact, the epistemological or methodological defeat of microreduction does not imply anything at all about the ontological assumptions underpinning it. Careless arguments to the contrary are quite damaging to the reputation of emergentism, and have obscured several legitimate claims by emergentists.

One promising avenue of investigation is the computational irreducibility of certain discrete models of physical systems. If extensible to natural systems, a claim of “downward causation” is plausible because parts organize in such a way as to form a dynamically stable whole. This is “process” emergence (superconductivity is a strong candidate to fit this scheme). Such systems can be studied using different methods than the standard linear techniques common in much of science and engineering. Whether these results really do apply to physical systems is debatable, but this is a question that afflicts reductionism as much as emergentism: to defeat computational equivalence would largely defeat scientific realism. Any remaining disagreements between reductionism and emergentism would be purely procedural.

The most profound kind of emergence is also the most abstract—so abstract, in fact, that its proponents dispose entirely with any discussion of the material basis of the systems they study. This is because multiple realizations result in wholes that behave the same way even though they comprise different parts. Emergentists investigate the topological properties of attractors in phase space, searching for ultimate laws that govern their behaviour. If they are successful, certain methodological implications are likely to ramify throughout the scientific domain. At present, those implications are largely speculative, although proponents go on at some length in describing them anyway. These emergentists risk burying any important results they may have with their overeager claims.

Careful analysis of the arguments surrounding emergentism can help separate the wheat from the chaff, and this winnowing process can be quite worthwhile. For some systems, the whole is indeed greater than the sum of its parts. Emergentism complements reductionism by providing tools to investigate those systems for which reductionist methods fall short. Because of computational irreducibility, there is good reason to think that reductionist methods will ultimately fail for certain systems, but it simultaneously provides the test necessary to distinguish between epistemological and ontological emergence, and computational equivalence provides the basis for the rigorous investigation of complex systems. The future seems very bright indeed for emergentists.

No comments: