Back to CQ Homepage

Remarks on Teed Rockwell's "A Defense of Emergent Downward Causation"



I think this paper needs to distinguish more carefully between emergent and downward causation, and innocent from robust versions of each. I would urge the view that emergent properties do not entail downward causation, and would like a careful discussion of what it takes to get from one to the other. I would also urge care in distinguishing exactly which version of emergence and/or downward causation are supported by empirical evidence from dynamical systems theory. For concreteness let us call higher level properties M-properties and the lower level ones which reductivist's hope they reduce to the P-properties. (I speak of properties causing other properties, but for those who believe only events are causes, take this to be a facon de parler, to be eliminated in favor of corresponding event talk.

Innocent Emergence: Human epistemological limitations entail that humans cannot predict or explain the evolution of occurrences of M-properties given all information they can acquire about occurrences of P-properties.

Robust Emergence: In principle there is no way to explain the evolution of (occurrences of) some M-properties given perfect knowledge of all P-properties of events.

Innocent Downward Causation: When M-property M is causally relevant to the occurrence of a P-property P* there other P-properties which completely explain and predict P*. (This is causal overdermination, so not quite innocent according to Kim, but innocent in comparison to the next one:)

Robust Downward Causation: On some occasions an M-property M is causally relevant to an occurrence of a P-property P* and there are (in principle) no P-properties that completely explain P*.

The big difference between the innocent and robust versions of these theses is that in the robust cases, we lack supervenience of M properties on P properties. I am guessing that such failure of supervenience is crucial to your conception of downward causation. But my take on the literature on dynamical systems theory is that most accept (and should accept if I am right) some brand of supervenience. In fact, I think some sort of supervenience assumption is built into the DST approach. Therefore DST

does not support robust downward causation. Therefore I think it was misleading for you to say that there is "no reason why any particular molecule in the liquid should become part of either type of pattern" or (speaking of boiling liquids) there is " no reason why any particular molecule in the liquid should be part of either type of cylindrical pattern". Actually electro-magnetic forces between the molecules and the boundary conditions for the system predict which particle does what (it is just WE can't do the prediction). Typically a DST model begins with a set of deterministic differential equations describing the elements of the model which (in principle) predict fully the behavior of the elements. In a case where a quantum fluctuation carries the day, it is not as if the formation of a cell determines a particle's fate, for local electro-magnetic interactions between a particles neighbors determine (at least in a statistical sense) that the cell forms. To put it quickly: the CELL doesn't shove anybody around, over and above what you would get by standard physical forces.

So to my mind, examples from DST might be used to support Innocent Emergence, but they need not support Robust Emergence nor Robust Downward Causation. Kelso does speak of circular causation, as when parameters like the carrying capacity of the environment for predator-prey population are affected by the evolution of the populations of predators and prey. But I do not think that Kelso, or most Dynamical Systems Theorists would opt for robust versions of Emergence or Downward Causation. Say it is foxes and rabbits. The M-level of description is given by the dynamical systems model, which has variables for the size of prey and predator populations, a parameter for the carrying capacity, and equations relating prey to predators and an interaction between the population of prey and the carrying capacity. (When there are too many prey they destroy their environment.) The carrying capacity parameter (actually here it starts to take on the character of a variable) in turn figures into the nature of the (possibly chaotic) boom-bust cycles in the population. However, it is not as if the carrying capacity exerts a "new" force on the population, one that is not explained by lower level properties of the environment. I mean at the lower level of description, we could explain the whole show by tracing the fate of every blade of grass, and the munchings of each and every animal. Kelso's circular causation appears robust when we restrict talk to the language of the DST model, but that is mere appearance.

The same point should be made about the behavior of molecules in turbulent cells. It is not as if formation of a cell creates NEW physical forces that "take over" the behavior of the particles. For all behavior of each particle is locally determined by perfectly standard physical forces like electromagnetic ones.

Another analogy may help. Consider the emergent properties that we find in ant colonies. Here the creation of teams for food foraging, food collection when it is found, etc. are all emergent features of simple and local laws governing ants concerning the circumstances under which each one is to lay down and follow a pherome trail. The creation of a system exhibiting emergent features such as the organized teams performing different functions, does not itself argue for the creation in nature of a new force that keeps the ants in line. The lesson of DST, is that fantastically complex structure is the natural product of perfectly normal (but non-linear) forces operating at the micro-level. The moral is that downward forces aren't needed to generate this structure despite our folk intuitions that something new is needed to "keep everybody in line". True, new brands of causal talk may very well now leap to mind, as in: that team caused the ant colony to avoid famine. True, this brand of causal talk may be more than a facon de parler; it may be essential and eliminable. But it doesn't provide the sort of robust downward causation that undermines supervenience.

Another point: Suppose I cave in on supervenience and grant you Robust Emergence. It still would not follow that we had Robust Downward causation. Here is why. Suppose you take a 2 levels position something like Amie Thomasson's. Then you could believe in ineliminable causally efficacious properties at the M level, ones that don't even supervene on the physical (not that Thomasson likes this) but you have a levels view that disallows cross level causation.

I think there may be good reasons based on physics for disliking robust downward causation, even if one goes for robust emergence. One faces the old problem, namely violation of physical laws. Even in a non-deterministic physical theory, one assigns probability distributions to events at the P level. Now suppose there is robust downward causation. Then M properties change the probability distributions at the P level, in conflict with the laws describing the P-level. True there may be no class with P-laws when one considers individual events, but summing over statistical properties predicted at the P-level there will be a clash. (Of course when the P-level laws are deterministic the class will be overt event by event. Unless M-properties are epiphenominal, downward causation breaks physical law.

The only way out for you that I can see is to declare that there are P-properties whose occurence is genuinely unaffected in any way by physical laws, that is, no information at all is available about their statistics of occurrence. But in such a case what is the intuition that rates these as physical properties? Properties that we are not able to explain or predict even statistically would seem to lie outside science.

So I think the conundrum of downward causation is much more serious, and in any case the evidence for robust downward causation in PHYSICS is scant at best.

Back to CQ Homepage

James W. Garson

Department of Philosophy

University of Houston

Houston, Texas 77006-3785