Dedicated to Zar and Gwumbldy

Originally a chaos of ideas. Those which were consistent with each other survived, the remainder perished and are perishing.

Friedrich Nietzsche

    Many of the ideas in Chapter 1 also appear in "Measuring Static Complexity," Journal of Mathematics and Mathematical Sciences 15, and in The Structure of Intelligence (Springer-Verlag, 1992).

    Chapter 2, parts of Chapter 1, and all but the last three sections of Chapter 3 are based on "Self-Organizing Evolution," Journal of Social and Biological Structures 15-1.

    Part of Section 3.4 is adapted from "What Is Hierarchical Selection," Biology and Philosophy 6

    Section 3.7 is based on a talk, "Structural Complexity of Sequences, Images and Automata," presented at the International Conference on Finite Fields, Coding Theory, and Advances in Communications and Computing, in Las Vegas in August 1991.

    Chapter 5 is based loosely on a talk, "Evolutionary Optimization," given at the AMS-MAA Joint Conference in Baltimore in January 1992.

    Section 5.7 is based largely on "Simulated Annealing on Uncorrelated Fitness Landscapes," by Malwane Ananda and myself, recently submitted for publication.

    Thanks are due to Jeanie for her help formatting the manuscript, and to Donna for her help with some of the figures

"The evolving mind" is an ambiguous phrase. At first, it brings forth images of the evolution of mind over historical time _ from reptile mind to lower mammal mind, through primate mind, and on to modern human mind. And indeed, we shall deal with this sort of mental evolution briefly, in the final chapter. But our central focus in these pages will be on a less obvious interpretation of the phrase "the evolving mind." We will be concerned with a much shorter time-scale _ milliseconds, minutes or hours rather than eons. My main aim in this book is to give a new formulation of the hypothesis that mental process is itself a form of evolution. That when you think, remember or feel, the process going on in your head is actually a process of evolution by natural selection.

    This is not a new hypothesis _ it can be traced back to Darwin and several of his early followers, including Herbert Spencer (1873). For most of the twentieth century, however, evolutionary theories of mind and behavior have been out of fashion. Only in the past decade-and-a-half has a substantial community of scientists returned to the intuition underlying the speculations of the early Darwinists. For example:

    1)Gregory Bateson (1980) has analyzed mind and society in evolutionary terms, thus arriving at several novel insights into the nature of culture and mental illness.

    2)John Holland (1975) and others have designed computer programs called "genetic algorithms" which learn according to an evolutionary strategy. These algorithms have proved useful for many practical problems, such as controlling a camera eye or determining the optimal placement of gas pipelines (Goldberg, 1989).

    3) C. MacFarlane Burnet's (1976) theory of clonal selection, the basis of modern immunology, bears a very strong resemblance to Darwinian evolution. Inspired by the clonal selection theory, immunologist Gerald Edelman (1987) has proposed that the brain, in a certain sense, evolves by natural selection. Following this idea he and his colleagues have constructed "evolutionary" computer models of neural vision processing.

    4)Other neuroscientists, for instance Changeaux (1985), Young (1965) and Calvin (1987), have proposed alternate evolutionary theories of brain function.

    5)Jason Brown (1988), Larry Vandervert (1991), Paul MacLean(1991) and others have proposed psychological theories based on "microgenesis," the idea that mental process recapitulates evolutionary process.

    This body of work is diverse and extremely exciting, and it has led to a variety of important insights, several of which will be utilized in the pages to follow. These scientists have integrated ideas from all sorts of different disciplines (e.g. Holland is a computer scientist, Bateson a biologist/anthropologist/psychologist, Edelman an immunologist/ neuroscientist, Brown a psychiatrist). However, it seems to me that these very different thinkers from very different fields have all succumbed to a common error. They have attempted to construct theories of the mind or brain by imitating the standard "Neo-Darwinist synthesis" theory of evolution. But I believe that this theory _ to be called strict Darwinism from here on _ is, if not outright wrong, at best inadequate and highly misleading. Along with many leading biologists, such as Gould (1977, 1980, 1984), Salthe (1985) and Lima de Faria (1988), I believe that strict Darwinism must be replaced or at least supplemented with an evolutionary theory more cognizant of the self-organizing properties of complex systems. My strategy here is to begin by using some ideas from theoretical computer science to sketch the broad outlines of a self-organizational evolutionary theory. Only then do I move on to formulate an evolutionary theory of mind and brain, one modeled on the self-organizational theory of evolution rather than the strict Darwinist theory.     The psychological/neuroscientific theory thus obtained has many points in common with the ideas of Edelman, Bateson, Holland and Brown. Unlike these earlier theories, however, it presents a unified, comprehensive model of mentality, one which encompasses _ in an abstract way _ all generally recognized processes of reasoning, memory, perception, action and affect.

    This may seem like a lot of ground to cover in such a slim volume. Indeed, it is a lot of ground to cover. However, the burden is eased by periodic reliance on my previous book, The Structure of Intelligence (SIfrom here on). SI gives a unified theory of mental process in terms of the concepts of pattern and multilevel optimization. Here we shall call this theory the "dual network model," and use it to order the psychological aspects of our investigations (it plays particularly large roles in Chapters 1 and 6).

    The main conclusion which I reach here is that, while mind/brains are very different from ecosystems, there are two important dynamical processes which are common to both: namely, evolution by natural selection and form creation by sexual reproduction. These processes manifest themselves somewhat differently in mind/brains than in ecosystems, but this does not detract from the commonality. After all, the Navier-Stokes equations manifest themselves differently in the heart than in the ocean, but no one doubts that they apply to all fluids. The key point is that these two very different complex evolving systems share two essential processes.

A Note on side Hypotheses

So, the main theme of the book is that mind/brains and ecosystems share the processes of evolution by natural selection and form creation by sexual reproduction. However, I have not adhered to Edgar Allan Poe's dictum that every single word of a story should be directed toward the same unifying goal. Quite the opposite: a large number of "side hypotheses" are made, regarding the nature of mutation processes, the dynamics of immune systems, the origin of structural stability in organisms, the role of heterochrony in the evolution of mind, and so on. These hypotheses are related to the main theme, but they are not crucial to it.

    Some of these hypotheses are proposed with great confidence; others are more speculative. Some are highly unorthodox and will probably find few supporters. However, I do not expect or desire the reader to agree with everything I say. The important thing is that all of these hypotheses are plausibly falsifiable, yet none of them is proved false by current data.

    In a well-developed field _ like, say, solid-state physics or microbiology _ it makes sense to develop ideas in a deductive, Euclidean manner, without introducing side hypotheses and without musing about philosophical implications. But evolutionary theory, neuroscience and cognitive psychology are all in phases of rapid, tumultuous development -development regarding basic conceptual premises as well as technical details. Given this, it seems to me very appropriate to adopt an attitude

of openness to new ideas, even those which seem unlikely at first. At any rate, that is the spirit in which these side hypotheses are offered.

A Note on Style

The ideas discussed in the following pages are intensely interdisciplinary. However, very few readers will possess an expertise in more than one or two of the relevant fields. This dictates a rather less formal style than would be appropriate for a specialist journal. Terminology is kept to a minimum; it is introduced only where absolutely essential for clarity.

    In addition, I have taken a relatively relaxed attitude toward the mathematics and computer science involved. The definition-theorem-proof format has been avoided, along with the implementation details of computer simulations. My aim here is to present some new ideas about the nature of evolving systems. Often mathematics is a necessary tool, and often the results of computer simulations are important pieces of evidence. However, the ideas are the main point, not the details of the mathematics or the computer work. This does not imply a lack of rigor _ all terms are precisely defined, either in the text or in standard references. And my own computer experiments (which mainly regard peripheral topics) are described in sufficient detail that the adequately expert reader might replicate them on his or her own. However, wherever possible I have given an intuitive discussion in place of a mathematical derivation or a table of computer results.

Synopsis

Even if these efforts at simplification are successful, there is no way to avoid the inherent complexity of the arguments which this book exists to present. In order help these difficult arguments go down more smoothly, let me begin by presenting a synopsis of the more important points. This synopsis is highly compressed, and it is a lot drier than the material which it compresses (I hope!). But if it gives the reader even a vague idea of what the following pages hold in store, it will have served its purpose.

    1) We begin with Gregory Bateson's Metapattern (borrowed from Charles S. Peirce), an ontological axiom which states that the stuff of the biological world is pattern. Concepts from the theory of computation are used to make this axiom mathematically and conceptually precise.

    The important concept of emergent pattern is defined: a pattern emerges between two entities if it is present in the combination of the twoentities, but not in either of the entities separately. And the structural complexity of an entity is defined as the "total amount" of pattern in it. If the Metapattern is accepted, then these two concepts become essential to any analysis of biological reality.

    2) We turn from these abstractions to a concrete biological example: the mammalian immune system. The theory of clonal selection states that immune systems evolve by natural selection; using the computer simulations of Alan Perelson, Rob deBoer and their colleagues as a guide, we inquire as to the exact nature of this evolution. We conclude that, in immune systems, survival is roughly proportional to emergence: those antibodies which generate a large amount of emergent pattern in conjunction with other antibodies, will tend to survive.

    3) It is argued that this correlation between emergence and survival also applies to the evolution of species: that an organism which generates a large amount of emergent pattern in conjunction with its environment will tend to survive. Since the environment of each organism consists partially of other organisms, this implies that an ecosystem is an enormously complex "system of pattern equations." This principle is illustrated with biological examples.

    Next, it is suggested that complex structures evolve by a process called systematic structural instability. Structural instability is when a small change in the input to a process leads to a large change in the structure of the output. A hierarchical process possesses systematic structural instability if: 1) each of its levels takes as input patterns in the levels below it, and 2) each of its levels contains some structurally unstable processes. Many biological examples of structural instability are known.

    This completes our general analysis of evolutionary theory, the theme of which is the use of concepts from the theory of computation to confront the self-organizing phenomena neglected by strict Darwinism.

    4) Attention is turned to the brain, which is modeled as a network of neurons. The general behavior of neural networks is discussed and then, following Edelman, the neural network model of the brain is refined into a model of the brain as a network of neuronal groups. A map is defined as a connected set of neuronal groups which tends to act as a unit; still following Edelman, we propose that the essence of brain dynamics and structure lies in the structure and dynamics of the network of interconnected maps. The first two of three proposals regarding the structure and dynamics of the network of neural maps are made: that neural maps are structural hierarchically and interact according to the multilevel methodology, and that there tends to be a large amount ofemergent pattern generated between nearby maps.

    5) The theory of genetic optimization is introduced. Using this theory as a guide, we present our third hypothesis about the structure and dynamics of the network of neural maps: that neural maps reproduce sexually and differentially. We describe a neural network model according to which this is indeed the case. Along the way to this model, several other points regarding genetic optimization are raised _ for example, the relation of sexual reproduction to epigenesis is analyzed in a novel way.

    These three hypotheses about the network of neural maps, taken jointly, imply that the neural maps in any given brain may be modeled as a population reproducing sexually and evolving by natural selection (and creating new form by systematic structural instability).

    6) Lacking any substantial biological evidence in favor of (or opposing) these three proposals about the network of neural maps, we instead attack the matter from a psychological angle. We introduce the dual network model of mind from SI and show that, if one takes this model as an assumption, one obtains the three proposals about the network of neural maps as consequences.

    This completes our analysis of mental/neural process as evolution by natural selection.

    7) Finally, we investigate the implications of this model of the brain for the evolution of mind over historical time. Brown's theory of microgenesis _ which states that individual thoughts, as they arise, pass from archaic into more modern stages _ is reviewed and critiqued in the context of the theory of the dual network. And the theory of random graphs is used to show that, as the size of a dual network gradually increases, its effectiveness may experience a sudden jump or "phase transition." This implies that the evolution of intelligence was quite possibly a case of punctuated equilibrium (Eldredge and Gould, XX).

    Microgenetic theory is combined with this concept of a phase transition, to yield the hypothesis that the development of an individual thought often involves punctuated equilibrium _ that, as an emerging thought passes from older to more modern forms, there often comes a certain point at which it suddenly becomes more effective and definite. This suggests that, in a sense, every sudden stroke of insight is a re-living of the historical emergence of intelligence.