THEORETICAL FOUNDATIONS FOR COGNITIVE SYSTEMS

AND ITS RELEVANCE TO PSYCHIATRY.

Henrique Schützer Del Nero, Alfredo Maranca, Lucia Maciel, José Roberto Piqueira

Institute for Advanced Study-University of São Paulo

Address: P.O. BOX 11135 ZIP: 05422-970 São Paulo BRAZIL

E.Mail: h.delnero@aevermind.brasil.net

ABSTRACT

The authors try to suggest that Cognitive Science cannot be helpful to Psychiatry and to Neropsychology, if the theoretical foundations of functional emergence from the physical to the mental are not revised. Both models, Traditional Artificial Intelligence, and Connectionism, nevertheless differing in the way entities are manipulated, i.e. their rules of connection, still maintain mental entities untouched. The authors call this position a mental-semantic-priority. Connectionism broke only one barrier towards reduction, that of rules of connection that may resemble brain computations. But it maintained entities at the mental level, interpreted according to ordinary language. If reduction of terms is impossible, due to intentionality, multiple instantiability, and so forth, the reduction of larger blocks --those that might interest to Psychiatry-- might be possible. This is what we call syndrome-type-reduction. The reduction might be done through a Dynamical Systems Analysis of the oscillatory mode of assemblies of neurons, trying to grasp bifurcations and chaos at the space of solutions. This would be a way to perform what we call topological computation, in a very broad sense of computation. Topological computation is suggested as a way the brain might use to build the sets of mental syndromes, being at this moment considered only four categorial instances of blocks: conscious-voluntary control, automatic control, dreams and psychoses.

INTRODUCTION

This brief article tries to state that there is certain incompatibility between the modern Theory of Dynamical Systems (TDS) and Cognitive Science (CS). The "semantical priority" that governs symbolic and sub-symbolic paradigms is not compatible with a "syntactical" view based upon the possible way brains codify information. Multiple instantiability arguments, like those that maintain functionalistic positions are incompatible with a type-reduction of mental objects. Token-identity is a weak position, easy to be adopted, but insufficient if one deals with crucial problems in Neuropsychology and in Psychiatry. From the theoretical point of view, we think that "emergent" entities might be present in the transitions from the brain level to the mental level. However, from another point of view, reduction must be sought in a certain amount, compatible with disciplines that deal with both levels: with the mental and with the physical. Neuropsychology relating brain lesions and mental disabilities states relations that cannot be omitted. Psychiatry deals with symptoms at the mind level and drugs that act at the brain level. Quantitative changes promoted by drugs change qualitatively symptoms at the mind level. Cognitive Science can have two different attitudes regarding this problem: either accepting it, as if it didn't disturb the cognitive paradigm, or rejecting it. We think that both reactions may happen. This article tries to put things in place, regarding their theoretical foundations and suggesting that a "semantical" approach to the brain domain might enlighten certain types of reduction, what we call "syndrome-type reduction" based upon topological computation. I. Type-reduction and emergence When reduction from the mental to the physical was still pursued, there remained a problem: there is a difference between reducing entities and reducing theories (Nagel 1961). Reducing a theory of mind to a theory of the brain is totally different from reducing mental terms to physical terms. Intentionality and consciousness play a vital role regarding the reduction of mental states to brain states. The impossibility of having a radical translation (reduction) from a mental idiom (mentalism) to a physical one (physicalism) (Fodor 1975) due to cross-classification, multiple instantiability, etc. lies in the evidence that there are non-countable physical facts that can stand for intentional operators. (Fodor accepts that there can be cross-classification among mental and physical predicates, but this only scaffolds token-physicalism and never type-reduction). "I believe that p" subsumes a non-countable number of ps (Searle 1984) , thence any generalization of the mental sciences will not be radically translatable unto physical generalizations. Mainly, due to intentionality, mental terms have a strong relation to consciousness (Flanagan 1986) which precludes twice the radical translation from mental states to brain states. We assume that all these problems hold for the reduction of terms. Intentionality, and thence consciousness, are tied to the reductions of terms. Reducing theories is foremost complicated because neither Psychology has had a good candidate to be reduced, nor the Neurosciences have a whole theory that could reduce the mental one. We think that in the Cognitive Science community, since its beginnings, nobody tried to talk about the reduction of theories, in the strong sense of theory that comes from the Philosophy and Logic of Science, due to the motives stated above. Debates about reduction and emergence held always about entities, be they terms (intentional), be they "faculties" like memory, attention, etc. Functionalistic positions (Putnam 1986, Marr 1982) are based upon the same argument: the same program can be run in different architectures, henceforth the dissociation between the level of algorithm/computation and that of implementation is radical. Studying the human brain is peripheral to a general theory of mind, because the mental blocks and the rules that govern their conjunctions, forming strings, inferences, etc. are independent from the physical laws governing the implemental substratum. Nevertheless, there is an ontological identity (token-physicalism) between the levels, which doesn't imply that there is a map from one to the other (predicate emergence with monistic essence). This is the ground where the foundations of Cognitive Science lie. In the very beginning, Traditional Artificial Intelligence (TAI) denied radically the benefits that could come from brain- knowledge. There were mental entities (intentional terms, representations, schemes, etc.) , mental rules (particularly based upon Logic- Predicate Calculus), brain entities (e.g. synapses) and brain rules (e.g. the action potential). Reduction of entities and rules were impossible. The mind would be software and the brain hardware. A general science of the mind, including symbol manipulation capacities, was to be dissociated from the brain. The mental entities had what we call "semantical priority". The meaning and existence of mental entities were supposed to be reliably reported and were supposed to be almost well described by ordinary language. If the entities were "faculties", there would have to be "theories" that endorse their existence. However, neither the brain laws, nor the brain entities could enlighten the way rules and entities are formed at the mental level. Brain substrata might carry them, but artificial architectures might do it differently (at the implemental level), despite being equivalent at the "mental" level. Connectionist Artificial Intelligence (CAI) partially changed these postulates. Doubting that there was a logical rule that connects entities, CAI proposed models based upon a mechanism of connection that might resemble the brain mechanism. However, the entities remained untouched, in spite of being slightly different in CAI from those of TAI: they can be distributed, they can be the interpretation that arises from a certain pattern of the network (e.g. an attractor), shortly they are "subsymbols" (Smolensky 1988). If rules are no longer those that spring from Logic connectives, but dynamical principles that have arisen from the dynamics of artificial synapses, the entities remained untouched. The "semantical priority" still held. If one wants to model anxious disorders, "anxiety" will be in both models, TAI and CAI, assumed to exist, being the interpretation that stands for a file, a sequence o 0s and 1s, a "neuron", an attractor, etc. Neither TAI, nor CAI tried to suggest that at the level of entities there could be something at the implemental level that allows and constrains the entities that can be formed. Shortly, we suppose that, although from the syntactical point of view the importance of brain knowledge changed with CAI, the "semantical priority" remained the same in both paradigms. Even in CAI terms like "gestalt" and "emergence" still have room, which renders Neuroscience half-important in the cognitive endeavor. In CAI models the rules of connecting entities are closer to the supposed brain rules, but this doesn't mean these rules can enlighten the nature of the mental entities. We suggest that these facts underlie the proliferation of different groups with slight different underpinnings that use the term "cognitive", despite being different from the fundamental point of view. Neuropsychology and Psychiatry, among others, are disciplines that adopt slightly different approaches--i.e. they suppose that the knowledge of the brain might give more than only a half-job based upon the way entities relate among themselves. Of course, among the scientific community it is common for happen research to take place without a severe discussion about crucial principles. What we are trying to say is that the above phenomena occur in those disciplines, because their methodologies and their aims are more compatible with a radical translation from one level to the other. Ask a biological psychiatrist if he/she expects, in principle, for a radical solution from the anatomical and physiological point of view, or if he/she thinks there will always be a set of disorders that cannot be explained by any brain theory, that cannot be treated with any drug and that there will always be room for psychotherapy as a medical tool. We suggest that, apart from eliminative materialism (Stich 1983) that is a philosophical position, the methodologies and beliefs of many scientists, in the areas cited above, are that the future will render radical reduction without the elimination of ordinary language. The reader might say that one thing is to discuss the philosophical foundations of a problem and other the scientists' beliefs about the discipline where this problem remains hidden in the obscure place of epistemological and ontological foundations. Scientists don't usually think of philosophy of science and philosophers sometimes don't have much experience regarding what kind of beliefs and commitments a scientist has (if they pertain to the sociology of science it is still science). That's why certain crucial concepts in any field, sometimes are not considered crucial in the other. For example, among philosophers the reduction of mental terms is an obstacle, be the term -- pain or fear --; on the contrary, among scientists pain is more objectively accessible than fear, mainly if there is denial, repression, social pressure, etc. preventing the subject to have smooth access to his/her own inner feelings, thoughts and will (the three elements that constitute the mind according to Ryle-1949). Cognitive Science can handle the above problems twofold: it can deny that there is a third alternative to TAI and CAI, assuming that "semantical priority" is unquestionable, no matter what several scientists believe, or it can face the problem of finding a theory that inverts semantical priority to syntactical priority. Syntactical priority would be a way to define the classes of possible beings, entities and meanings that might arise from the brain computations (in a broad, not necessarily algorithmic, sense of computation). Of course, CS could say that this philosophical inversion doesn't hold to entities but to larger cognitive sets. Mental terms would still be non-reducible, but groups of mental entities -- that might form mental syndromes, not yet mental theories -- could be candidates to be totally reduced. This is a position that might accommodate the dissent among the various cognitive disciplines, mainly because we think that the kind of reduction Psychiatry and Neuropsychology pursue is not so tied to "intentional" idioms, but closer to clusters of mind reports-- which in medical terms is called symptom, and that together with signs constitutes syndromes. Syndrome-type-reduction (STR) might be an intermediate level between entities and theories, that could promote a reversal in the semantical priority that holds in TAI and in CAI. STR might be based upon the way the brain encodes and manipulates information, positing that this way of coding and interpreting information is what renders possible a class of mental entities to appear. Instead of explaining mental contents that constitute the semantics of mental entities, STR would propose, maybe in a kantian style, a brain syntax that is a form that renders possible relatively invariant clusters of mental entities to exist. The mental as a form might be the qualitative result of quantitative changes at the firing rate of neuron assemblies. This is compatible with today's Psychiatry regarding the use of drugs in mental diseases. Let us, briefly, examine the profile of STR.II. Syndrome-Type-Reduction, "syntactical-brain-priority" and Topological Computation STR departs from the problems stated above and from a more practical one: what is the relevance of anatomical, physiological and environmental factors in mental diseases? Traditional approaches, typical from the XIX century imposed a dissociation between mind and brain. Psychopathology and Psychotherapy (here including Psychoanalysis), for different reasons, assumed a certain independence between mind and brain. Karl Jaspers in his classic "Psychopathology" (1920), in spite of stating the relevance of physical factors in the genesis of mental diseases, uses a conceptual framework based upon Brentano and in the first Husserl (1900) that is known by "Descriptive Psychology" or Phenomenology (the term leads to mistakes since Husserl used it in his second phase with a different conceptual framework underneath); Freud stated many times the provisory character of his works, since one day everything would be explained in a brain language. However, as we said above, there is a difference regarding a scientist working with a certain methodological approach and the same scientist willing to believe in certain things: the conceptual framework of Psychoanalysis, despite the author, are strongly tied ton non-reducible doctrines concerning the mind. Freud is a functionalist in this sense (Flanagan 1986). Psychiatry after the fifties became strongly motivated by advances in Psychopharmacology and in the Neurosciences. Today, in spite of many wholes, the discipline may dream on a virtual day where mental illness might be explained without certain mentalistic doctrines. Does this mean that there will be no realm for talk-therapies, "empathy", social pressures, etc.? Aren't they encoded as physical entities like everything else? What is the difference between a drug that alters the frequency of action potentials at the synaptic level, via a modification in an enzyme and in the receptor, and of hermeneutics and language that imposes new strategies, plans and interpretations for a person? Strangely, if one accepts a narrow Psychiatry, based upon only in drugs and in Neuroscience, one is denying the monistic essence of the Central Nervous System (CNS): if everything is converted into signals, what is the difference between altering the end product, e.g. the firing rate, or of altering a disposition of certain circuits? The strong reductionistic psychiatrist doesn't see that stating such a position implies: either a future of eliminative materialism without language (the language that carries the symptoms the same psychiatrist uses to make a diagnosis) or an acceptance of functionalism regarding the mind with the following problem language still exists to endorse symptoms, but it is totally dissociated from the level where drugs will act, i.e. the brain that interests this or another psychiatrist. If there is still trouble with the mind this will constitute an environmental disturbance tied to Sociology and not to Medicine. If one doesn't want to commit the above inconsistency, one has to accept that monism imposes that everything is a signal in the CNS. Of course, there is a topography that imposes a certain modularity in the execution of functions, but the way the CNS codifies information, via action potentials, is the only common predicate among bottom-up (Neurosciences and Psychopharmacology) and top-down (Psychodynamical theories) strategies to study mental illness. If everything is a signal in the CNS, the very essence of the mind might be scaffolded by the character of this signal (of course, if one doesn't want to accept token-physicalism). This signal, conjoined with the topographical character of the CNS (in humans there happens a great evolution regarding the cortex, mainly the frontal lobes) might have something interesting that allows quantities to change at the synaptic level and abrupt qualitative changes in the inner mental domain to occur (as in the external behavioral domain). Then, information is coded via a frequency of action potentials and quantitative changes at this level (for example during drug administration, eletroconvulsotherapy, etc.) promotes qualitative changes in the mental screen. There is a strong invariance between these quantitative/qualitative sets, be it related to the onset of mental diseases (type, evolution, etc.), be it related to the response to drug theraphy (time to respond to a certain drug, withdrawal syndromes, etc.). Syndrome-type-reduction departs from these two problems: inverting the semantical priority that holds in TAI and in CAI (in spite of the fact that it doesn't deny the irreducibility of intentional blocks, but proposing a syndrome-reduction that is tied to the forms of the mental and not to its contents), and given a way to Psychiatry to be still a broad top-down and bottom-up discipline, but with a strong scientific (here meant formal) character. If signals underlie the mental, and if one sees quantitative changes in the implemental level, with qualitative changes at the mental level, and if there is a strong invariance among these changes, whose theory could embrace and maybe explain these three elements- quantities, qualities and strong invariance, having a signal domain underneath, scaffolding the basis for the mental to appear? The Theory of Dynamical Systems (TDS) is a good candidate, because it explains via the concept of bifurcation in the parameters space, qualitative changes in the space of states of a system. The formal concept that seduces when one deals with slight variations in a certain variable, leading or not to dramatic changes in the solution is that of structural stability (Guckenheimer & Holmes 1983). Considering a pendulum, the parameter that describes the friction coefficient may be positive, negative or zero. In the domain of positive and negative values of this parameter every slight perturbation e doesn't lead to topological variations in the state space: there is no topological variation, the qualitative solution is the same, the system shows structural stability. For zero values in this parameter, a slight perturbation will change dramatically the systems's behavior from the topological point of view. This is structural non-stability and this value is called bifurcation value. Bifurcations are topological variations in the system's behavior when parameters are in critical values (called bifurcation values), that when slightly perturbed lead to dramatic changes in the space of states. If one is dealing with values in a distribution of course each perturbation of a parameter in a branch of structural stability will lead to homeomorphic solutions regarding the perturbed system, i.e. both solutions will have common traits from the qualitative-topological point of view. Chaos, a phenomenon that may happen in deterministic non-linear dynamical systems that are characterized by sensitivity to the initial conditions, is structurally stable, in spite of having an enormous quantity of states in the state-space, all unpredictable. Chaos can occur after successive bifurcations (Guckenheimer & Holmes 1983). Freeman (1992) and other researchers have proposed that there is chaotic behavior in the CNS and that this might be the source of richness and variability some systems present. If one adopts a signal-monistic referent to underlie the forms of the mental, the quantitative-dynamical description of large populations of neurons (assemblies) might be described by differential ordinary equations. Variations in the parameter space, when in bifurcation values lead to dramatic-qualitative changes in the systems flow (the temporal function that describes the systems dynamics). The mind might be seen as a qualitative description of topological changes. Whenever the topology is the same the systems performs automatic computation, however if there is a bifurcation and a sudden change in the topology, other brain areas are recruited, another brain-mode (the conscious-mode we suggest) assumes the problem. Consciousness, the very substratum of the mental, might be represented by one of its predicates: voluntary control. Computations over dynamical non-linear systems (trains of action potentials that come from neurons that are described as oscillators in the Hodkin-Huxley -1952- model) perform the switch from voluntary-frontal modes and automatic-cerebellar modes, and vice-versa, (throughout a pre-analysis that might happen at the hippocampus and thalamus), might analyze homeomorphisms and topological similarities (thence we call topological computation the syntax of STR). Consciousness, the very essence of the mental might be analized via the hypothesis of the synchronization among certain units (the 40 Hz hypothesis, cf. Koch and Davis 1994). A model of the mental based upon the oscillatory mode of neurons and their description via dynamical non-linear systems might ensure three kinds of topological behavior for a system: with bifurcation, without bifurcation and with successive bifurcations. We define, based upon these three possible behaviors, four syndrome-candidates to be mapped unto the signal-brain level: voluntary-conscious mode (one bifurcation), automatic-non-conscious mode (without bifurcation), dreams (successive bifurcations with non-robust chaos) and psychoses (successive bifurcations with robust chaos). STR departs from these four elements-- voluntary control, automatic control, dreams and psychoses-- trying to build a type reduction where the mental forms are defined via partitions in the topological domain that describes the oscillatory mode of assemblies of neurons. Shortly, trying to be an alternative to semantical-priority models and a basis for a strong Psychiatry, STR, here described very briefly, might have the following assumptions:a) a bottom-up strategy that doesn't consider mental phenomena as primitives but as results of the partitions in the topological space that can be made by complex systems.b) a redefinition of the class of mental contents, particularly consciousness, as a class of functions that might have equivalence with syndromes, being the first four types voluntary control, automatic control, dreams and psychoses.c) an analysis of objects that can be subject of voluntary and of automatic control as a class of oscillations that can be modeled through differential equations.d) use of the notion of structural stability, bifurcation parameter and chaos that allows to analize an oscillation topologically, the way the system might use to switch from the voluntary to the automatic mode, and vice versa. Conclusion Consciousness, in STR, is rescued as a function that scaffolds the mental and that is firmly tied to the brain style of computation, particularly when it is able to detect topological changes, recruiting the frontal lobes. Consciousness in this model might be considered a function that is the result of a topological analysis that happens at the preprocessing levels, being consciousness, and henceforth the qualitative-mental, a function or a form that demodulates information whenever it presents structural instability, i.e. a bifurcation in its parameters space.ReferencesFlanagan, O. (1986): The Science of the Mind. MIT PressFodor, J. (1975): The Language of Thought. Harvard University PressFreeman, W. (1992): "Tutorial on Neurobiology: from single neurons to brain chaos" in International Journal of Bifurcation and Chaos, vol.2 , number 3 Guckenheimer, J. & Holmes, P. (1983): Nonlinear Oscillations, Dynamical Systems, and Bifurcations of Vector Fields. Springer-Verlag. Berlin.Hodgkin, A & Huxley, F.(1952): "A Quantitative Descriptions of Membrane Current and Its Application to Conduction and Excitation in Nerve" in Journal of Physiology 117, 500- 544Husserl, E. (1900): Logische Untersuchungen. Max Niemeyr Verlag.Jaspers, K. (1920): Allgemeine Psychopathologie. Verlag von Julius Springer. BerlinKoch, C. & Davis, J. (ed) (1994): Large-Scale Neuronal Theories of the Brain. MIT PressMarr, D. (1982): Vision. W.H.Freeman and Company.Nagel, E. (1961): The Structure of Science. Harcourt, Brace and World, Inc.Putnam, H. (1986); Mind, Language and Reality. Cambridge University PressRyle, G. (1949): The Concept of Mind. Hutchinson.Searle, J. (1984): Minds, Brains and Science. Harvard University Press.Stich, S. (1983): From Folk Psychology to Cognitive Science. MIT Press Smolensky, P. (1988) "On the proper treatment of connectionism in Behavioral and Brain Sciences 11, 1-74