Emergence & the Mind-Body Problem

Mikio Akagi. 09, 16 April, 2008.

"Emergent Systems: A Discussion" Working Group. Center for Science in Society. Bryn Mawr College. Bryn Mawr, Pennsylvania.

 

ABSTRACT

I will investigate the prospects of an "emergentist" solution to the mind-body problem. The presentation will have three parts. In the first part of this presentation I hope to lay the conceptual groundwork for the talk. First I will describe a contemporary reading of the mind-body problem as it interests me, then articulate a conception of emergence which seems promising for the emergentist project. In the second section, I will defend a dynamical model of emergent explanation. In the third part I will examine the application of dynamical models of explanation to the mind-body problem. Drawing on the insights of the embodiment perspective in cognitive science, I will claim that a dynamical account of the mind-body problem will need to set up a very wide scope.

The bulk of this presentation is based on my half-baked paper "Emergence for the Mind-Body Theorist." Comments are welcome.

 

1. Concepts

1.1 The Mind-Body Problem

The mind-body problem historically concerns the connection between the soul and the body.

In contemporary discussions, it is usually taken to be the problem of accounting for the relation between the mental and the physical. Most current approaches to the mind-body problem attempt somehow to reconcile the existence of the mental with the metaphysical doctrine of physicalism, usually token physicalism.

The mental: psychological and intentional phenomena, or events under psychological or intentional descriptions. Incl. beliefs, desires, being pleased, worrying, supposing that p. I should also like to include within the realm of the mental such events as calling up a friend, riding a bicycle or making a sandwich, at least when they are identified according to those descriptions.

The physical: the phenomena described by an ideal physics, of which our present physics is taken to be our best approximation.

Token physicalism: The view that all instances of real objects and events are identical with or supervene on physical objects or events.

The physicalist world picture: a "package" of views including token physicalism and usually the doctrine of the causal closure of physics, as well as commitments to, e.g., methodological naturalism, the unity of science, explanatory reductionism, the generality of physics, empiricism, nominalism, atheism, &c. (Stoljar, 2001)

The causal closure of the physical: the view that every physical event [with a cause] has as its cause another physical event.

A theory or account of T is reducible to a theory T' if and only if the phenomena explained by T are explainable in T'.

 

1.2 Emergence

The notion of emergence is meant to account for recalcitrant natural phenomena, without rejecting token physicalism. If emergence is to do more than lip service to physicalism, though, it cannot consist in an enumeration of brute facts (e.g. about chemical relations).

There are at least three axes used to describe emergent concepts: synchronic/diachronic, strong/weak, and ontological/epistemological. Embracing instances of strong emergence usually require abandoning certain elements of the physicalist world picture, particularly the commitment to reductionism. In what follows, I take the concept of emergence I use to be strong, ontological and ultimately diachronic.


Emergent property: A property P is an emergent property if

  1. P is described in an ideal theory T,
  2. T explains a class of systems SC,
  3. the systems in SC are composed of structures of entities described by an ideal physical theory T' of those entities,
  4. the systems in SC supervene on the phenomena accounted for in T',
  5. it is epistemically impossible to identify occurrences of P with any property finitely definable in T',
  6. each instance of P is an instance of one of a definable set of properties PC in T', and
  7. each set in PC is epistemically indistinguishable in T' from some other set.
  8. (liberal paraphrase of Newman, 2001: 185)


Emergent phenomenon: a class of objects or events EC which are taken up as objects of explanation by a theory T, where the members of EC necessarily exhibit one or more emergent properties.

 

2. Explanation

P1. If being in the basin of attraction of a strange attractor is a sufficient condition for being an emergent property, then chaotic nonlinear dynamical systems are emergent systems.
P2. Chaotic nonlinear dynamical systems can by explained through dynamical systems theory.
P3. Being in the basin of attraction of a strange attractor is a sufficient condition for being an emergent system.
C. Therefore, some emergent systems can be explained through dynamical systems theory.

The first premise is certainly true.
The second is controversial, and I shall return to it later.
The third is argued for by Newman (2001):

 

2.1 CNDSs as emergent systems

CNDSs are taken to be physically-determined systems described by DS theory, which feature strange attractors. They exhibit unpredictability and sensitive dependency on initial conditions.

If we take P to be the property of being in the basin of attraction of a strange attractor in a chaotic nonlinear dynamical system, then the rest of our definition of emergence is redundant.


Emergent property: A property P is an emergent property if

  1. P is the property of being in the basin of attraction of a strange attractor in an ideal theory T,
  2. T explains a class of systems SC which are described by dynamical systems theory,
  3. the systems in SC are composed of structures of entities described by an ideal physical theory T' of those entities,
  4. the systems in SC supervene on the phenomena accounted for in T',
  5. it is epistemically impossible to identify occurrences of P with any property finitely definable in T',
  6. each instance of P is an instance of one of a definable set of properties PC in T', and
  7. each set in PC is epistemically indistinguishable in T' from some other set.


(1) and (2) are our assumptions.
(3) is redundant because CNDSs are physical systems.
(4) is redundant because CNDSs are physically determined, ex hypothesi.
(5) is redundant because of sensitive dependency on initial conditions.
(6) is redundant because CNDSs are physically determined.
(7) is redundant because of the unpredictability of CNDSs.

Therefore:

Emergent property: A property P is an emergent property if it is the property of being in the basin of a strange attractor in an ideal theory T, and T describes a class of systems SC which are described by dynamical systems theory.

 

2.2 DS theory as explanatory

There is a charge that DS theory describes, but does not explain. So what is the difference ?

It is certainly not the case that DS equations are like entelechies or the infamous "dormative virtue," because they are counterfactual-supporting (and predictive, Mark). So they do not merely re-label phenomena. Do they, however, explain ? Or do they still merely describe ?

The charge leveled at DS theories is identical to the charge sometimes leveled at quantum mechanics, or classical genetics. The debate in quantum mechanics is contentious enough that discussing it won't really clarify our present position. The charges against classical genetics were finally allayed, however, when a physical interpretation of the equations was discovered in the 1950s (Maienschein 1998). 

 

2.21 What is an explanation ?

This story can be accounted for with a semantic approach to explanation (van Fraassen, 1989), on which an explanation consists of a formal object (usually logical or mathematical) and an interpretation (norms of application). A theory is empirically adequate insofar as it is associated with a class of models which can be used to represent all the phenomena.

Then DS models are not explanations, but formal elements of explanations. If there are satisfactory interpretations of the equations, that is to say norms of application which link equations and physical systems, then DS models figure in successful explanations.

So what are satisfactory norms of application ?

It depends on the context (van Fraassen 2007). This is not a satisfying answer but it is probably true. Anyway, so long as we are concerned with the scientific enterprise we have constrained the context significantly. There is good reason to think that scientific explanations must invoke causes so long as there are any (Pasnau, July 2007), even if causal talk is dispensed with (Norton 2001).


So do DS models track causes ? They can (cf. Clark 1997 and van Gelder 1995). Tim van Gelder uses the example of the Watt centrifugal governor, whose function is well-modeled as a dynamical system. The trick here is that an explanation featuring DS equations produces causal understanding, but that the causal influences are complex and continuous, and do not lend themselves well to serial models.

A Watt centrifugal governor. Digital image from the Wikimedia Commons (source).

A Watt centrifugal governor. Digital image from the Wikimedia Commons (source).

If DS models do, in fact, explain, then P2 holds.

Therefore, some emergent systems can be modeled using DS theory.

 

3. Prospects

3.1 Back to the Mind-Body Problem 

Newman (2001) points out four interesting properties of mental states.

1. They are ecological. This is to say that they are extended in time, and subsist in the context of a rich (high-dimensional and topologically complex) phase space which includes, among other things, other mental states.

2. Mental states are not noticeably periodic. Even those feelings which are most closely associated with circadian cycles are not strictly periodic, and circadian periods can be easily disrupted.

3. Mental states do not tend toward a particular mental state. Even in sleep or with various kinds of physical deprivation (sensory, social, sleep, sustenance) mental states do not become increasingly static.

4. Mental states are not easily predictable, nor random. They are arbitrarily sensitive to small factors either in the brain, body or environment of an organism, but are nevertheless reliably responsive.

Newman abduces that mental states can be accounted for by DS theory (1), that they do not have predominant limit cycle or torus attractors (2), nor point attractors (3), and in fact exhibit chaotic dynamics (4).

So it is likely that mental states can be modeled as a chaotic system.

However, here Newman reasons (hastily) that since mental states exhibit chaotic dynamics, the brain must be a chaotic physical system. This may be true, but it's not the whole story unless one swallows some assumptions which have proved troublesome.

 

3.2 Embodiment

The concept of embodiment is at least as controversial and confused as that of emergence. As I see it, however, a central insight whic gave rise to the current wave of embodiment theorizing is that not all computation happens in the brain. E.g. Barbara Webb's model of cricket phonotaxis, where much of the "computation" not by complex neural circuitry, but by physiological features of the cricket's body and some very low-level neural sensitivities.

In the case of humans, however, it is more complicated. Cf. Alan's (April 2006) talk on "Socially Extended Cognition." Humans don't just use our bodies to augment the computational powers of our brains; we use sticky notes, algorithmic procedures, other people, beer bottles and so on (Clark 1997, 2001; Clark & Chalmers 1998, &c.).

If so much "mental" activity supervenes on the body and, indeed, the proximal environment of persons, then the interpretation of a dynamical model of mental states cannot just be the brain. Rather, it must include the body of which the brain is a part, and the environment in which the body is situated. The relation of the mental to the physical, then, will be a relation of individual mental states not to brain states, but to a host of spatially distributed physical structures.

 

3.3 Second nature

But an account of human mentality cannot stop even there, for humans are discursive creatures endowed not only with a canny nature but a cannier second nature. Humans are social creatures who are sensitive not only to physical properties of their environments, but social situations in them.

Moreover, the ways in which humans are so responsive are not crudely innate, but depend upon the conceptual proclivities fostered within a sociocultural context. So a dynamical model of the mental must take as its interpretation not only physical structures, but social structures.

But here is a worry: that we have finally lost the physicalist thread which was the lifeline of the emergentist account. If emergentism does not preserve our commitment to token physicalism, it has lost its virtue as a solution to the mind-body problem.

 

3.4 And so on

If the emergentist account of the mental is to be saved, the complexity of the matter will continue to multiply. The social structures which constrain individuals are not normally taken to be fixed rigidly. In fact, emergence is often cited as the kind of concept which could account for social structures, in terms of relatively uniform cognitive structures in communities of people. Another, nested, dynamical model could in theory account for the production of social structures from human behavior. It is not viciously circular that such structures should be both caused by human behavior and causally influence the same behavior, for the same reason that the Watt centrifugal governor works. The situation is not one of impossible, circular causation; rather it is one of non-serial, complex, reciprocal causation. Most real causal influences are not lexically-ordered and don't behave like geometric idealizations of billiard balls.

By now the solution we have envisioned is highly speculative, and it is difficult to marshall strong arguments to favor it rather than another approach. However, a powerful upshot of such a view, should it prove in time to be plausible, is that the phase space of sociocultural dynamics may turn out to be a natural way to account for social norms in a solidly naturalistic framework. Norms may be simply the topological vectors which serve to constrain individual actions. Such norms may include behavioral norms which demarcate intelligibly meaningful actions, as well as linguistic norms which serve to fix reference and produce strange phenomena such as intentionality.

In the practice of cognitive science, there is a concept of a "head problem." A head problem is a theoretical puzzle of such complexity and dense interconnections with other cognitive puzzles, that in order to solve it you would need to have a full-blooded theory of human cognition and consciousness. If this series of considerations is right, however, then the project of naturalizing the mental (that is, solving the mind-body problem) is not a "head" problem at all. Rather, to settle the matter we should need to smuggle in not just a head, but a structured natural world of norms. The project of naturalizing the mental may require not just the solution of the hardest questions of cognitive psychology, but also the most central puzzles of metaphysics.

 

References

Baker, Alan. 2006, 5 April. "Socially Extended Cognition." Emergent Systems: A Discussion. Center for Science in Society. Swarthmore, PA.

Clark, Andy. 1997. Being There: Putting Brain, Body, and World Together Again. Cambridge, MA: The MIT Press.

---. 2001. Mindware: An Introduction to the Philosophy of Cognitive Science. New York: Oxford University Press.

Clark, Andy & David Chalmers. 1998. “The Extended Mind.” Analysis 58: 10-23.

Maienschein, Jane. 1998. “The Gene: Historical Perspectives.” In Evelyn Fox Keller & Elizabeth A. Lloyd, eds. Keywords in Evolutionary Biology. Cambridge, MA: Harvard University Press, pp. 122-127.

Newman, David V. 1996. “Emergence and Strange Attractors.” Philosophy of Science 63: 245-261.

---. 2001. “Chaos, Emergence, and the Mind-Body Problem.” Australasian Journal of Philosophy 79: 180-196.

Norton, John D. 2003. “Causation as Folk Science.” Philosopher’s Imprint 3:4 <http://www.philosophersimprint.org/003004/>

Pasnau, Robert. 2007, 11 July. "Does Science Seek Causes? Science, Knowledge and Causality." Colorado Summer Seminar in Philosophy. University of Colorado, Boulder. Boulder, CO.

Stoljar, Daniel. 2001. "Physicalism," in Edward N. Zalta (ed.), The Stanford Encyclopedia of Philosophy (Winter 2005 Edition). Stanford University. Accessed 20 September, 2007. <http://plato.stanford.edu/archives/win2005/entries/ physicalism/>.

Van Fraassen, Bas C. 1989. Laws and Symmetry. Oxford: Clarendon Press.

---. 2007, 23 July. "Values in Science: What Are Theories or Models?" Colorado Summer Seminar in Philosophy. University of Colorado, Boulder. Boulder, CO.

van Gelder, Tim. 1995. “What might cognition be, if not computation?” Journal of Philosophy 92, 345-381.