Structure of Intelligence -- Copyright Springer-Verlag 1993

Back to Structure of Intelligence Contents

Chapter 11

Consciousness and Computation

11.0 Toward A Quantum Theory of Consciousness

    For sixty years physicists have struggled with the paradox of quantum measurement. However, despite a number of theoretical advances, rather little progress has been made toward resolution of the basic dilemma. The problem is one of physics versus phenomenology. According to quantum physics, no physical entity is ever in a definite state; the most one can ever say about a given entity is that it has certain probabilities of being in certain states. And yet, both in daily life and in the laboratory, things do sometimes appear to have definite states.

    For instance, the equations of quantum physics predict that, in many situations, an electron has a 50% "chance" of having a positive spin, and a 50% "chance" of having a negative spin. Yet when the physicist probes the electron in his laboratory, it appears to have either a positive spin or a negative spin. According to the equations of quantum physics -- the Heisenberg equation and the Schrodinger equation -- such a reduction to a definite state is impossible.     Of course, one may have various degrees of probabilistic knowledge. In certain situations, one might know an electron to have a 90% "chance" of having positive spin, and a 10% "chance" of having negative spin. But there can never be 100% definite knowledge. Heisenberg's indeterminacy principle says that one can never have complete knowledge of the state of any particle: the greater the accuracy with which one knows its position, the less the accuracy with which one can know its momentum; and vice versa. In order to predict what the particle will do in the future, one needs to know both the position and the momentum; but according to quantum physics, this is possible only probabilistically.

    This sort of indeterminacy is a proven scientific fact, inasmuch as quantum theory is the only known theory that correctly explains the behavior of microscopic particles, and it predicts only probabilities. Classical mechanics and electromagnetism gave definite answers about the behavior of microscopic particles, but these answers were experimentally wrong. Furthermore, it seems likely that, if quantum theory is someday superseded, the theory which followsit will build on the probabilistic nature of quantum theory, rather than regressing to classical ideas. In fact, it has been proved mathematically (Bell, 1964, 1987) that any physical theory satisfying certain simple requirements must necessarily have properties similar to those of quantum theory: it must deal only in probabilities.

BEYOND THE PROJECTION POSTULATE

    In his classic treatise on quantum theory, John von Neumann (1936) introduced the "projection postulate", an addition to the basic principles of quantum physics which states that, when an entity is measured, it reduces to a definite state. This approach appears to be adequate for most practical problems of quantum mechanics; and, although, many physicists find it unacceptable, there is no equally elegant alternative. The only trouble is that no one has ever given a satisfactory definition of "measurement".

    Originally it was thought that a microscopic event could be considered to be measured when it "registered" an effect on some macroscopic entity. The justification for this was the belief that, as entities become larger and larger, the probabilistic nature of quantum physics becomes less and less relevant to their behavior. For instance, according to quantum physics a baseball dropped from a window has an infinity of possible paths, but one of them, or one small class of them, is overwhelmingly more likely than the others.

    But this naive identification of measurement with macroscopic effect cannot stand up to criticism.     Spiller and Clark (1986) have constructed a Superconducting Quantum Interference Device (SQUID) which is about the size of a thumbnail and yet displays the same sort of uncertainty as an electron. One can never know both the intensity and the flux of its magnetic field with perfect accuracy; there is a finite limit beyond which further accuracy is impossible. Its state is fundamentally a probabilistic superposition.

    And it appears that the brain may display a similar form of quantum indeterminacy (Changeaux, 1985; Penrose, 1990). Recall that a neuron fires when its charge exceeds a certain threshold amount. It follows that, on occasion, highly unpredictable quantum phenomena may push the charge of a neuron over the threshold. And this neuron may then set other neurons off, and so on -- in this manner a tiny quantum indeterminacy may give rise to a huge neurophysiological uncertainty. If the extra charge has a fifty-fifty chance of being there, then the entire pattern of neuronal firing that ensues from its presence probably has about a fifty-fifty chance of being there. A pattern of neuronal firing might, for instance, represent a state of mind. And when you consider the fact that there are over a hundred billion neurons in the brain, the possibilities for interlocking quantum uncertainties are astounding. The exact numbers are difficult to estimate, but it appears that this may be a significant phenomenon.

    One intriguing alternative to the projection postulate is Everett's (1957)"many-worlds hypothesis", which assigns to each uncertain situation an array of universes, one corresponding to each possible outcome. For instance, according to the many-worlds hypothesis, every time a physicist observes an electron to have positive spin, there is an alternate universe which is exactly identical to this one, except that in the alternate universe the physicist observes the electron to have negative spin. This is an interesting possibility, but it is empirically indistinguishable from the projection postulate, since these alternate universes can never be observed.

THE QUANTUM THEORY OF CONSC IOUSNESS

    Another alternative, first proposed by Wigner (1962), is that "measurement" may be defined as "registration into consciousness." To see the motivation for this radical idea, let us turn to the infamous paradox of Schrodinger's cat (1948). Here the peculiarity of quantum theory is elevated to the level of absurdity. Put a cat in a soundproofed cage with a radioactive atom, a Geiger counter and a vial of poison gas. Suppose that the atom has a half-life of one hour. Then it has a fifty-fifty chance of decaying within the hour. According to the dynamical equations of quantum physics, this is all one can know about the atom: that it has a fifty-fifty chance of decaying. There is no possible way of gaining more definite information.

    Assume that, if the atom decays, the Geiger counter will tick; and if the Geiger counter ticks, the poison vial will be broken. This set-up is bizarre but not implausible; a clever engineer could arrange it or something similar. What is the state of the cat after the hour is up? According to quantum theory without the projection postulate, it is neither definitely alive nor definitely dead -- but half and half. Because the atom never either definitely decays or definitely doesn't decay: quantum physics deals only in probabilities. And if the atom never either definitely decays or definitely doesn't decay, then the cat never definitely dies or definitely doesn't die.

    One might argue that the cat is not in a state of superposition between life and death, but rather has a fifty percent chance of being alive and a fifty percent chance of being dead. But according to quantum theory without the projection postulate, the cat will never collapse into a definite state of being either alive or dead. What sense does it make to suggest that the cat has a fifty percent chance of entering into a state which it will never enter into? The function of the projection postulate is to change the statement that the cat is half dead and half alive into a statement about the probabilities of certain definite outcomes.

    Of course, the fact is that if we look in the box after the hour is up, we either see a dead cat or a living cat. Phenomenologically, by the time the observation is made, one of the two possibilities is selected -- definitely selected. But when, exactly, does this selection occur? Since measurement cannot be defined as macroscopic registration, this is a very serious problem.

     And the problem is resolved very neatly by the hypothesis that probabilisticoccurrences are replaced by definite occurrences when they enter consciousness.

    For instance, this implies that Schrodinger's cat is not half dead and half alive, but rather has a fifty percent chance of being dead and a fifty percent chance of being alive. The cat becomes definitely dead or definitely alive when a conscious being sees it. As Goswami put it,

        it is our consciousness whose observations of the cat resolves its dead-or-alive dichotomy. Coherent superpositions, the multifaceted quantum waves, exist in the transcendent order until consciousness brings them to the world of appearance with the act of observation. And, in the process, consciousness chooses one facet out of two, or many, that are permitted by the mathematics of quantum mechanics, the Schrodinger equation; it is a limited choice, to be sure, subject to the overall probability constraint of quantum mathematics (i.e. consciousness is lawful).... [C]onsciousness... is not about doing something to objects via observing, but consists of choosing among the alternative possibilities that the wave function presents and recognizing the result of choice. (1990, p. 142)

That is, the mind does not create the world in the sense of reaching out and physically modifying events. But it creates the world by selecting from among the wide yet limited variety of options presented to it by the probabilistic equations of physics.

11.1 Implications of the Quantum Theory of Consciousness

    The measurement paradox is not the only philosophically troublesome aspects of quantum physics. Bell's Theorem (1987), with its implication of instantaneous communication between distant events, is equally unsettling. The simplest example of this is the Einstein-Podolsky-Rosen (EPR) thought experiment. Two electrons, initially coupled, are shot off in different directions. It is assumed that each one flies for millions of miles without hitting anything. Each one, according to quantum physics, has a fifty-fifty chance of spinning to the right or to the left -- there is no way to make a more accurate prediction. However, the rules of quantum physics do imply that the two are spinning in opposite directions: if one is spinning to the right, then the other one is spinning to the left; and vice versa.

    Now suppose someone measures one of the electrons, and that it all of a sudden assumes a definite value. Then the other electron will immediately also assume a definite value -- because it is known that the two are spinning in opposite directions. If one is measured to be spinning to the right, then the other is instantaneously known to be spinning to the left. When Einstein conceived this example, he thought he had disproved quantum mechanics -- because nothing so absurd could possibly be true. After all, he asked how doesthe one electron tell the other one which way to spin? Special relativity forbids information to travel faster than the speed of light; so it would seem that if the particles were sufficiently distant, the value of the spin of one particle could take eons to reach the other particle.

    But, absurd as this may be, it is an experimentally proven fact (Aspect and Grangier, 1985). Scenarios very similar to the original EPR thought experiment have been tested in the lab. It turns out that, mathematically speaking, this peculiar "nonlocality" does not contradict special relativity, because no information is being transmitted, only a correlation. But this is very little consolation: it is a violation against the spirit, if not the letter, of special relativity.

    Reality does not consist of pairs of electrons, coupled and then shot out into space a million miles in opposite directions. Consider the following thought experiment. Split apart two coupled physical systems, say A and B. Suppose that, from the state of A, one could infer the state of B, and vice versa. Leave A alone but let B interact with C for a while, and then separate B from C. Finally, measure A. A is collapsed into some definite state. If B had not interacted with C, one could say that the state of B would also, immediately, collapse into some definite state. But the state of B now depends also upon the state of C, which according to quantum physics has no definite value but is rather an array of possibilities. So the measurement of A does not collapse B to a definite state. But it does, however, decrease the uncertainty involved in the state of B. It increases the "closeness" of B to a definite state.

    Technically speaking, assume that p=(p1,p2,...,pn) denotes the probabilities of the various possible states in which B might be. Then one may show that, in the situation described above, the measurement of A necessarily changes p into a new set of probabilities p%=(p1%,...,pn%) so that H(p1,...,pn) < H(p1%,...,pn%), where H is the entropy function

    H(p1,...,pn) = -[p1logp1 + ... + pnlogpn]

A similar statement may be made when the possible states of B form a continuum rather than a discrete set. Recall that the entropy of a probability distribution is a measure of its uncertainty, or its distance from the most certain distribution.

This thought experiment may be generalized. What if the state of B cannot be completely determined from the state of A? If the state of A yields any information at all about the state of B, then it is plain that the same result holds. If A and B were ever coupled, no matter how loosely, no matter what they have done since, measurement of A reduces the uncertainty of the probability distribution characterizing the states of B. Bell's Theorem implies that this sort of propagation of certainty is a necessary aspect of any physical theory that is mathematically similar to quantum theory.

     In terms of the quantum theory of consciousness, what does this mean? A little consciousness can go a long way! If two sets of possibilities have been coupled in the past, and are then separated, then whenever consciousness makes one of them definite, the other one becomes definite automatically,instantaneously, without any physical causation involved.

QUANTUM CONSCIOUSNESS, BIOLOGICAL CON SCIOUSNESS

    By introducing consciousness, one obtains a philosophically elegant resolution of the paradox of quantum measurement. But in a way we are abusing the word "consciousness". What qualities does this abstract entropy-decreasing consciousness share with the common-sense understanding of consciousness? We have reconciled the physics of measurement with the phenomenology of measurement only by separating the physics of consciousness from the phenomenology of consciousness.

    Mandler has proposed that

        ... [C]onscious constructions represent the most general interpretation that is appropriate to the current scene in keeping with both the intentions of the individual and the demands of the environment. ...Thus, we are aware of looking at a landscape when viewing the land from a mountaintop, but we become aware of a particular road when asked how we might get down or of an approaching storm when some dark clouds "demand" inclusion in the current construction. In a problem-solving task, we are conscious of those current mental products that are closest to the task at hand, i.e. the likely solution to the problem. (1985, p.81)

Whether or not this particular formulation is exactly correct, it seems plain that some similar characterization must hold true. Consciousness seems to have a role in planning and decision-making, but it is rarely involved in the minute details of everyday life: walking, turning the pages of a book, choosing words in conversation, doing arithmetic with small numbers, etc. In the language of the previous chapters, this means that -- as already stated -- consciousness has contact with only a certain restricted range of the perceptual hierarchy.

    The decision-making aspect of consciousness is intuitively harmonious with quantum theory: in making a decision, one is reducing an array of possibilities to one definite state. There is a sense in which making a decision corresponds to selecting one of many possible universes. But the quantum theory of consciousness gives us no indication of why certain decisions are submitted to consciousness, but others are not.

    One of the main problems here is that it is not clear what function the quantum theory of consciousness is supposed to serve. In Wigner (1962) or Goswami (1990), consciousness is essentially defined as the reduction to a definite state, or more generally as the decrease of the entropy of an array of possible states. This interpretation gives a transcendentalist resolution of the mind-body problem, made explicit by Goswami when he suggests that, as a heuristic tool, we consider the mind to be a coupling of two computers, a classical computer and a quantum computer. The quantum computer behaves in a way which transcends ordinary biophysics, and it is this transcendencewhich is responsible for consciousness.

    But there is another, more radical, way of interpreting the quantum theory of consciousness. One may begin with the assertion that consciousness is a process which is part of the dynamics of certain physical systems, e.g. human brains. This means that consciousness has some direct physical effect: that, for instance, when a pattern of neural firings enters consciousness, consciousness changes it in a certain characteristic way. The biochemical nature of this process is of course largely unknown. However, Edelman (1989) has made some very interesting hypotheses. In his notation, consciousness may be described as the continual interaction between C(W) and C[C(W). C(I)], where C(I) is the neural basis for categorization of I, the interoceptive input  -- autonomic, hypothalamic, endocrine. It is evolutionarily earlier, driven by inner events, mediated by limbic and brain-stem circuits coupled to biochemical circuits, and it shows slow phasic activity. C[W] is the neural basis for perceptual categorization of W, the exteroceptive input -- peripheral, voluntary motor, proprioceptive and polymodal sensory signals -- and is mediated by the thalmus and cortical areas. It is driven largely by outer events, is fast, and handles many more signals in parallel. C(W).C(I) represents the neural basis of interaction and comparison of two categorical systems that occurs, for example, at the hippocampus, septum, and cingulate gyri. C[C(W).C(I)] is the neural basis of conceptual recategorization of this comparison, which takes place in the cingulate gyri, temporal lobes, and parietal and frontal cortex. (The boldface C indicates conceptual categorization.)

Less technically, what Edelman proposes is that consciousness is the interaction between two processes: 1) the recognition of patterns in perceptions, and 2) the interaction between the recognition of patterns of perception and the recognition of patterns in internal, emotional, chemical stimuli.

    Given this biological characterization of consciousness, one may then hypothesize that the entropy reduction of arrays of possible states is correlated with those changes the states of conscious systems which correspond to conscious acts. This point of view -- which I will call the strong interaction postulate -- places less responsibility on quantum theory than the interpretation of Wigner and Goswami: it does not require quantum theory to explain psychological facts. Rather, it portrays consciousness as the point of connection between psycho-biological dynamics and physical dynamics; the bridge between the mind and the world.

    The quantum theory of consciousness, as presented by Wigner or Goswami, implies a transcendentalist resolution of the mind-body problem. But though it is useful for intuitively understanding quantum theory, it is not at all adequate for understanding consciousness. The strong interaction postulate is not merely a reinterpretation of quantum theory: it states that consciousness, in some sense,plays an active role in forming the physical world.

    In terms of the many-worlds interpretation, strong interaction implies that the brain-states of conscious entities put a special bias on the possible universes of the future. Everything in the universe figures into the array of probabilities of possible future universes -- but conscious systems are involved in an additional feedback process with this array.

    The idea of strong interaction may be worked out in much more detail, but that would lead us too far astray. It may be that future developments in physics will render this entire discussion nonsensical. However, as Penrose (1989) has pointed out, it is also possible that the relation between mind and body will be essential to the next revolution in physics.

SELF-AWARENESS

    Finally, I would like to point out that the quantum view of consciousness yields an interesting interpretation of that intangible feeling of self-awareness that accompanies consciousness of external objects or definite ideas. Consider the following scenario. P and Q are closely coupled algorithms, each one continually modifying the other. Simultaneously, consciousness greatly reduces the uncertainty of both the distribution of possible states of P and the distribution of possible states of Q. The reduction of the uncertainty P then reduces the uncertainty of Q yet further; and vice versa. The result is that the combined entity P%Q has, in effect, looked at itself and reduced its own entropy.

    It is not justifiable to say that P%Q did not really look at itself, that what really happened was that P and Q looked at each other. Because according to quantum physics, if we observed P % Q to see what was really happening, this would change the probability distributions. P and Q are quantum coupled, and this means they are effectively one entity. Clearly, this situation is not rare: feedback between different prominent structures is probably not the exception but the rule.

    According to this analysis, the feeling of self-awareness is not logically inherent to consciousness; it is rather an extremely common by-product of consciousness. This accounts for the fact that we are not continually absorbed with the sensation of self-awareness: it flits in and out of consciousness. Self-awareness is not quite the same as consciousness, but the two are inextricably interlinked.

INTELLIGENCE AND CONSCIOUSNESS

    Clearly, the quantum theory of consciousness is in a very early stage of development. However, none of the details are really essential here. The primary point of our excursion through quantum theory was to arrive at one simple hypothesis: that whereas Turing machines cannot possess consciousness, quantum computers can.

    This hypothesis has profound implications for the relation between consciousness and intelligence. To see this, we must consider a certain crucial but vastly under appreciated shortcoming of the theory of Turing machines. Mathematically, it is easy to deal with Turing machines of arbitrarily large processing capacity. But in physical reality, it is impossible to build an arbitrarily powerful Turing machine.

    If the parts of a machine are very small or very closely packed, then they are susceptible to quantum effects, and the machine is a quantum computer, not strictly a Turing machine: its behavior depends crucially on the peculiar properties of indeterminacy and nonlocality. But if the parts of a machine are not very small, and not very closely packed, then they must spread over a large expanse of space. However, according to the theory of special relativity, information cannot travel any faster than the speed of light. Therefore, there is a limit to the speed of a machine made of large and/or sparse parts.

    From these considerations it follows that, for any given time period T, there is a certain limit to the amount of computation that a physical Turing machine can do in time T. Even without estimating the specific numbers, it is clear that this limit is considerably smaller than the total amount of computation which a quantum computer can do in time T. Deutsch has shown that an abstract quantum computer cannot compute any functions which an abstract Turing machine cannot also compute. However, within any specified period of time, there is some physical quantum computer which can compute functions that no physical Turing machine can.     

    Now, intelligence depends not only on absolute computing power but also on speed. Therefore it follows from our assumptions that there is a certain degree of intelligence which quantum computers can attain but Turing machines cannot. Coupling this with the hypothesis that quantum computers but not Turing machines possess consciousness, one obtains the following intriguing conclusion: there may be a certain level of intelligence which can be attained only by conscious entities.

11.2 Consciousness and Emotion

    One often hears comments to the effect that "even if a computer could somehow think, it could never feel." And Dreyfus (1978), among others, has argued that this imposes strict limitations on the potential power of computer thought. After all, what is intuition but a sense of what "feels right"?

    The weakest point of such arguments is that they do not refer to any particular definition of emotion. Without a definition of emotion broad enough to apply, at least potentially, to entities substantially different from human beings, how can one make a fair judgement as to the emotional capacity of computers?

    One might argue that no such general definition is possible; that the only wayto understand human emotions is through human biology, which is inherently applicable only to entities substantially similar to human beings. This argument is bolstered by the numerous vagaries and self-contradictions which plague psychoanalysis and other classical theories of emotion, and also by the many impressive achievements of molecular psychology. However, it is nonetheless not implausible that there is a general structure of emotion.

    In his 1887 classic Laws of Feeling, Paulhan made an intriguing suggestion as to what this structure might be. And more recently, Mandler (1985) has outlined a theory very similar to Paulhan's, and gathered together a great deal of data in favor of it. These theories are preliminary and incomplete, and they are not essential to the main ideas of this book. However, they do indicate how one might develop a theory of emotion compatible with the ideas of the previous chapters.

    MacCurdy, a psychoanalyst, expressed Paulhan's core idea excellently in his 1925 Psychology of Emotion: it is precisely when instinctive reactions are stimulated that do not gain expression, that affect is most intense. It is the prevention of the expression of instinct either in behavior or conscious thought that leads to intense affect. In other words, the energy of the organism, activating an instinct process, must be blocked by repression before poignant feeling is excited.

In his own words, Paulhan's general law of feeling is simply thatdesires ... only give rise to affective phenomena when the tendency awakened undergoes inhibition.

Throughout Laws of Feeling, Paulhan implicitly assumes that "tendencies" are the stuff of mind. Since he never actually defines the word "tendency", I see no problem with reading "tendency" as "behavioral pattern".

    In the language of the preceding chapter, a "desire" is an instruction passed down the motor control hierarchy. Very low-level instructions probably do not deserve the label "desire", but there is no rigid cut-off point: the higher the level of an instruction, the more it is a "desire". Paulhan's hypothesis is that emotions occur only when such an instruction is not obeyed. This disobeyal may be due to mental incapacity or to the uncooperativeness of external reality. Emotion would never occur in an all-powerful, all-knowing, perfectly-running machine, because all of its internal instructions would invariably be fulfilled. Livesey, in the first volume of his 1986 Learning and Emotion, has sketched out similar ideas, although his analysis is less specific and hence less controversial.

HAPPINESS

    Paulhan apparently did not try very hard to apply his theory to particular emotions. He considered this to be an elementary exercise. Unfortunately, Icannot agree with him on this point: I have found this "exercise" to be formidably difficult. However, Paulhan did make two clear definitions, so let us consider these: happiness is the feeling of increasing order; unhappiness is the feeling of decreasing order.

    Paulhan did not define order; in the present context, it seems most straightforward to define the order of a set of patterns X as the sum over all x in X of the average, over all neighbors (y,z) of x in the mind's STRAM, of IN[x;(y,z)]. This implies that the "feeling of increasing order" is the "feeling of increasingly simple representation of the contents of one's subjective world."     To put it rather pedantically, this means that happiness is the feeling of recognizing patterns more effectively than in the immediate past; and, on the other hand, unhappiness is the feeling of recognizing patterns less effectively than in the immediate past. Or, more intuitively: happiness is the feeling of increasing unity.

    The only puzzling thing about this is that, according to Paulhan's definition, all emotion derives from inhibition; and therefore the "feeling of increasing simplicity" must mean the inhibition of those patterns which are rendered unnecessary or impossible by the advent of increasing simplicity. Is happiness, then, the feeling of stifling all the fruitless attempts to order the world which are rendered irrelevant by success? And is unhappiness, the feeling of stifling the habits instilled by a previously successful simplifying order of the world, in favor of further laborious attempts?

    This may seem a little bit bizarre. I would argue that, at any rate, this is one important meaning of the word "happiness." For instance, it explains, in a very rough way, why young children (who are continually presented with highly novel stimuli) obtain such pleasure from exploring and understanding. And, conversely, it also explains the human tendency toward closed-mindedness: the intrusion of novel patterns into a mental world of familiar ideas and routines will usually, at first anyhow, cause a decrease in the simplicity of one's representation of the world.

BEAUTY

    Next, let us consider the experience of aesthetic appreciation, of "beauty." One may define the beauty of X to Y as the amount of happiness gained by Y from the patterns which Y perceives in X, and this is not unsatisfactory, but it would be nice to have a definition which provided insight into the internal structure of the beautiful object. This we shall draw not from Paulhan, but from a loose interpretation of the work of Georg Simmel.

    In his essay on "The Face", Simmel (1959) proposed that "the closer the interrelation of the parts of a complex, and the livelier their interaction (which transforms their separateness into mutual dependence)," the greater the aesthetic significance of that complex. It seems to me that this "unity out of and above diversity," this "interaction" and "interrelation of the parts" is very well summedup by the concept of structural complexity. After all, an entity is not a priori divided into "parts"; the mind divides it into parts as part of the process of perception and comprehension -- the "parts" are patterns. And the "degree of interrelation" of the various patterns is, I suggest, simply the amount of pattern tying together the various patterns -- in other words, the structural complexity. Thus, it seems reasonable that the beauty of x to y is the happiness associated with that portion of St(x) which is perceived by y. Simmel's conception of beauty as emergent order coincides perfectly with Paulhan's idea of happiness as increasing order.

FREE WILL

    Free will and consciousness are often considered identical. Conscious decisions are considered freely willed. However, this point of view is unjustified. The first argument against free will is that, physiologically and psychologically, it is clear that conscious decisions are far from unpredictable. They are influenced very strongly by unconscious memories and biases, i.e. by parts of the brain which have no direct role in consciousness. This argument might be contradicted as follows: perhaps other influences bias consciousness, but they do not determine its behavior completely. They influence the likelihood of consciousness making one decision or another, but this only permits us to predict the outcome of consciousness in a rough probabilistic sense.

    But, if this is the case, then how does consciousness actually make a choice? Empirically, there is no way of distinguishing between the hypothesis that a choice is made by free will, and the hypothesis that it is made at random subject to the probability distribution induced by outside influences.

    So the existence of free will is essentially a moot point. I suspect that, in the future, it will be more fruitful to analyze free will as an emotion. To see how this might be done, consider Nietszche's analysis of "freedom of the will" as

         the expression for the complex state of delight of the person exercising volition, who commands and at the same time identifies himself with the executor of the order -- who, as such, enjoys also the triumph over obstacles, but thinks within himself that it was really his will itself that overcame them. In this way the person exercising volition adds the feelings of delight of his successful executive instruments, the useful 'underwills' or undersouls -- indeed, our body is but a social structure composed of many souls -- to his feelings of delight as commander. L'effet c'est moi: what happens here is what happens in every well-constructed and happy commonwealth; namely, the governing class identifies itself with the successes of the commonwealth. (1968, p.216)

The feeling of free will, according to Nietszche, involves 1) the feeling that there is indeed an entity called a "self", and 2) the assignation to this "self" of "responsibility" for one's acts. It is easy to see how such a feeling would fallunder the category of happiness, because it certainly does serve to impose a simple "order" on the confusing interplay of patterns underlying mental action.     But what is this pattern called the "self", which the mind recognizes in its own operation? Given a definition of "self", free will could be defined as the emotion resulting from the "belief" or "recognition of the pattern" that, in the absence of the self, effective pattern recognition (i.e. happiness) would not be possible. But even then the question why this belief would emerge would not be answered. Clearly there is a great deal of subtlety involved here, and we do not yet possess the tools with which to probe it.

EMOTION, COMPUTATION AND CONSCIOUSNESS

    Regarding the emotional capacity of computers, Paulhan's theory yields an ambiguous verdict. Emotion is analyzed to involve a certain characteristic structure. One may say that this characteristic structure only becomes true emotion when it enters consciousness, in which case it might well be that a quantum computer but not a Turing machine can experience emotion. Or, on the other hand, one may say that this structure is always emotion, whether or not it is consciously experienced. Essentially this is a matter of semantics.

    Mandler (1975) has made a similar point, observing that emotions have a "hot" aspect and a "cold" aspect. The cold aspect is the abstract structure of nonfulfillment of expectation. The hot aspect has to do with the presence of certain chemical factors which cause the vivid, visceral experience of emotion. One might say also that the cold aspect has to do with mind, the hot aspect with body. It may be that consciousness is a prerequisite for "hotness". The hot aspect of emotion is the bodily effect of the abstract mental nonfulfillment of expectation. The means by which this effecting takes place -- by which structure affects chemical levels -- is essentially unknown.

    And, if consciousness is a prerequisite for emotion, is it perhaps also true that emotion is a necessary part of consciousness? It is well known that, when a person summons something from long-term memory into consciousness, the limbic system is activated. The exact reason for this is a mystery, but it is also well known that the limbic system is the center of emotion. This reflects a psychological concept that goes back at least to Freud, who suggested that it may be impossible to remember something unless that something possesses some emotional content.

    This "Freudian" hypothesis coincides well with the present model of mind. We have hypothesized that consciousness contains the most "prominent" patterns in the mind, where a "prominent" pattern is both intense as a pattern, and the object of a great deal of activity on high levels of the motor control hierarchy. Is it not reasonable that a great deal of activity will center around those instructions which are not obeyed the first time around? Merely by virtue of their failure, they will receive more attention -- they have to be tried again, or alternatives have to be found.

    In conclusion: all that can be said with certainty is that consciousness and emotion are closely related. The nature of this relation is not yet clear. It appears that emotion may be understood to consist of two aspects: "hot" and "cold", or perhaps "conscious" and "structural". Perhaps the structural aspect of emotion may exist independently of the conscious, hot aspect; but in practice the two seem to usually occur together.