|
|
|
||||||||
CONSCIOUSNESS, INFORMATION AND PANPSYCHISM
William Seager, University of Toronto, 1265 Military Trail, Scarborough,Ontario MIC 1A4, Canada. Email: seager@lake.scar.utoronto.ca
Abstract: The generation problem is to explain how material configurations or processes can produce conscious experience. David Chalmers urges that this is what makes the problem of consciousness really difficult. He proposes to side-step the generation problem by proposing that consciousness is an absolutely fundamental feature of the world. I am inclined to agree that the generation problem is real and believe that taking consciousness to be fundamental is promising. But I take issue with Chalmers about what it is to be a fundamental feature of the world. In fact, I argue that taking the idea seriously ought to lead to some form of panpsychism. Powerful objections have been advanced against panpsychism, but I attempt to outline a form of the doctrine which can evade them. In the end, I suspect that we will face a choice between panpsychism and rethinking the legitimacy of the generation problem itself.
I: The First Datum
The hard problem of consciousness, according to David Chalmers, is explaining why and how experience is generated by certain particular configurations of physical stuff. Let's call this the 'generation problem', bearing in mind that the term 'generates' might be misleadingly causal: the explanatory relation we seek might be identity, instantiation, realization or something else altogether.1 While the generation problem has the outward appearance of a genuine scientific problem, one might dispute whether it is useful, mandatory or, even, intelligible. Suppose one divided the problems of statistical thermodynamics into the easy and the hard. The easy problems would be ones like 'how/why do gases expand when heated', 'why/how does pressure increase with increasing temperature', etc. By contrast, the hard problem would be to account for the generation of thermodynamic properties by the 'thermodynamically blank' particles which form the subject of statistical mechanics. What a mystery! Not only does a collection of independent particles act like a gas with thermodynamic properties, the collection somehow generates these very properties.
It is easy to see through this sham mystery and many are the philosophers who would suggest that the case of consciousness is no different. Once you have explained the appropriate and, no doubt, exceedingly complex internal structures which, ultimately, generate behaviour there is simply nothing more to be explained. Don't mistake a task made impossible because it is utterly senseless for one that embodies a deep metaphysical mystery. Chalmers insists that because consciousness is not a functional property, someone asking for an explanation of how a 'behaviourally sufficient' functional organization generates experience is 'not making a conceptual mistake' (Chalmers, 1995, p. 203). Certainly, this is not as obvious a mistake as that which demands an independent, additional explanation of how heat arises apart from the account provided by statistical mechanics of the functional isomorphism between statistical and macrothermodynarnics. But how can one show that the case of consciousness is not fundamentally similar?
A straightforward reply is simply to point out the intelligibility of the classical problem of other minds. There is no a priori argumentation that can eliminate this problem; everyone who thinks about it can see that each of us is, in a fundamental sense, alone. Perhaps against this, Wittgenstein said: 'if I see someone writhing in pain with evident cause I do not think: all the same, his feelings are hidden from me' (1953/1968, p. 223). A tendentious reply is: of course not, for we operate on the assumption that other people do indeed have experience and this is no time to question basic assumptions. But what if it is a beetle writhing about as it is impaled on the specimen seeker's pin or a lobster squirming as it's dropped into the boiling water? There is no easy answer, let alone a philosophically innocent a priori answer, to the question of where in the chain of biological development experience emerges, although even at the levels of the beetle and lobster one certainly sees behaviour similar (at least) to that caused by pain.
Daniel Dennett's theory of consciousness (1991) can be seen as fixated on debunking the generation problem. His discussion of philosophical zombies (i.e. creatures that act just like us but who are entirely unconscious, entirely without experience) is reminiscent of Wittgenstein's remark, and is similarly perplexing. At one point Dennett says (pp. 405-6) that if zombies were possible, you wouldn't really be able to tell whether something was a zombie or not (every coin has two sides) so it would be immoral to treat a putative zombie as an entirely unconscious being. This is no argument against the possibility of zombies and so even less an argument undermining the intelligibility of the generation problem. Elsewhere, Dennett allows that animals do have experiences even though they do not have the fully developed consciousness of human beings (pp. 442 ff.). He intimates that many distinct functional architectures could underwrite the ascription of (I avoid saying 'generate'2) experience and with regard to bats in particular remarks that we can know something of the range of bat experience by finding out what the bat nervous system can represent and which representations actually function in the modulation of behaviour (p. 444). But this only tells us what the bat could be conscious of, it does not tell us whether the bat is conscious of these things, for there can be no question of eliminating the distinction between conscious and unconscious representations which 'modulate behaviour'. So here we find a very stark form of the generation problem located in a theory that was supposed to banish it: given the viability of the conscious/unconscious representation distinction (endorsed by Dennett even as he asserts that this distinction is not absolutely clear cut) and given the undeniable fact that some unconscious representations modulate behaviour, what makes the difference between conscious and unconscious behaviour modulating representations in nonverbal animals? How come the states that represent bodily injury in bats are conscious experiences, if they are, whereas those representing the details of wing positioning during the hunt are not, if they aren't? (Here I am just imagining that bats are like me: I feel the pain in my ankle but am not usually aware of the complex foot work involved in running over an uneven surface, yet both involve behaviour modulating representations.) We have no recourse to the usual behavioural test of consciousness here — verbal behaviour — since, of course, bats can't tell us what they are aware of, but Dennett generously, if puzzlingly, admits that animals have experiences despite this.
Or again, in his discussion of splitbrain cases Dennett denies that 'commissurotomy leaves in its wake organizations both distinct and robust enough to support . . . a separate self' (p. 426). But the issue should be, does the right hemisphere have experiences (is it like a nonverbal animal) whether or not it is a fullfledged self. Obviously, the right hemisphere deploys various representations and some of these modulate behaviour (as the mass of splitbrain research amply reveals). So what makes them, or some of them, into experiences? If it is not simply the behaviour modulating powers of a representation, is it a representation's having behaviour modulating power above degree n (on some scale of efficacy)? Obviously, this is the generation problem all over again: what makes n (or a vaguely defined region around n) the right sort of thing to enable consciousness?
A theory of consciousness ought to tell us what consciousness is, what things in the world possess it, how to tell whether something possesses it and how it arises in the physical world (both synchronically from physical conditions and diachronically as an evolutionary development). The hard problem of consciousness is evidenced by the very real 'zombie problem' we have with animals. The honey bee, for one more example, acts like a creature that has experiences (visual, olfactory, as well as painful and pleasurable). Its behaviour, we have reason to suppose, is modulated by a complex system of internal representations generated, maintained and updated by a sophisticated neural parallel processor (rather like our own, if much less complex) — representations coordinated, for all I know, by the famous 40 hz oscillations. Now, on which side of the fuzzy line between sentience and nonsentience does the bee reside, or in the fuzzy zone itself?3 More important, for whatever answer, why? Suppose we made a robot bee that fitted well into bee life (beginnings are being made, see Kirchner and Towne (1994)). Suppose also we were sure the robot could not have experiences (it was truly an 'apian zombie'). Would that show that bees do not have experiences? Why? On the other hand, suppose we think that bees most certainly do experience things. Would that show that the robot also experiences (it certainly passes a bee-level Turing Test)? Why?4
As Chalmers claims, no extant theory of consciousness really addresses this range of questions, even as they admit that questions about bees' experiences are perfectly intelligible. Forgive me for harping on this, but the existence of the generation problem is absolutely crucial. Without it, there is no hard problem of consciousness. With it, the problem looks very hard indeed.
So hard, in fact, that Chalmers looks to a radical solution to bridge the so-called explanatory gap between physical system and conscious system. He suggests that consciousness is a fundamental feature of the universe, which must be simply accepted as the First Datum in the study of the mind. This is a neat way to finesse the generation question since there can be, by definition, no explanation of why fundamental features of the world arise whenever they do arise. For example, there is no explanation of why fundamental particles come in their observed mass ratios.5 Perhaps, once we accept the reality of the generation problem, there is no other way to proceed. And since I am strongly inclined to see the generation problem as a real problem, I am naturally attracted to Chalmers' solution. Yet I confess to find some disturbing elements in Chalmers' account, which I will argue suggest that a yet more radical view of the problem of consciousness might be dictated by his assumption that consciousness is a fundamental feature of the universe.
II: Organizational Invariance and Explanatory Exclusion
Begin with Chalmers' idea of the conditions under which consciousness arises, what he calls the principle of organizational invariance. Strictly speaking, this principle asserts only that 'any two systems with the same fine grained functional organization will have qualitatively identical experiences' (p. 214). But it follows from this that whether or not a system, S, is conscious depends upon its fulfilling some functional description. For suppose not: then there is some other, nonfunctional feature of S, call it Q, that generates consciousness. We could then build a system functionally isomorphic to S that lacks Q which will, by hypothesis, not be conscious, which is impossible by the organizational principle.6 It is disturbing that consciousness can be an absolutely fundamental feature of nature while being dependent upon particular systems satisfying purely functional descriptions, with the relevant similarity among these description being behavioural capacities. No other fundamental feature of the world has this character, or a character even remotely like it. It is rather as if one declared that 'being a telephone' was a fundamental feature of the world, generated by a variety of physical systems agreeing only in fulfilling the relevant, highly abstract, behaviourally defined functional descriptions.
Also, since Chalmers is adamant that consciousness presents a hard problem because it is not itself a functional feature it is very odd that consciousness should depend solely upon whether a system meets a certain abstract functional description. Of course, seeing that consciousness is a truly fundamental feature we cannot ask how it is that all and only systems meeting certain functional descriptions are conscious, yet this idea does seem to deepen rather than mitigate the mystery of the generation problem.
We also face here a variant of the generation problem which grows out of the inherent vagueness in the phrase 'fine grained functional organization'. Chalmers provides the example of a brain suffering a gradual substitution of its neurons, one by one, by 'silicon isomorphs' as an instance of his principle of organizational invariance (ibid.). But how do we know that this is the appropriate level of specificity of functional description? The silicon isomorphs, we may suppose, are not internally functionally identical to the neurons they replace, yet the internal workings of a neuron certainly fulfils some functional description. Or, from the other direction, why couldn't we replace large groups of neurons with a single silicon device that mimics the input/output relations of the whole neural group it replaces? Here is a new generation problem: how come just a particular level of functional description generates consciousness, and exactly which level does it? The problem is that functional duplicates of a system on various levels could, in principle, duplicate the whole system's behaviour. Would any system that acts like a conscious system be judged conscious by the principle? If not, suppose we have two systems which are functionally isomorphic at level n but are not functionally isomorphic at level n-1 (as in Chalmers' own example as I take it). Whether they share the same states of consciousness apparently depends upon which level is the appropriate level of description, but who decides? What does the universe know about levels of functional description?
A pernicious problem of explanatory exclusion7 arises from the aligning of consciousness with functional description. Any functionally described system must be actually instantiated by some assemblage of physical parts, if it is to take any part in the workings of the world. The causal efficacy of the system depends entirely upon the causal efficacy of its physical instantiation. Thus when we say such things as 'the thermostat turned on the furnace' the efficacy of the thermostat is entirely explained by the particular physical instantiation of this thermostat (say by the physical details of its thermocouple, or whatever else lets it serve its function). Perhaps a better example would the power of water to dissolve salt: this is entirely explained by the interactions of individual H20 molecules with the NaCl molecules that constitute salt, and these interactions are in turn entirely explained by the ultimately quantum mechanical properties of hydrogen, oxygen, sodium and chlorine. There is no room for water to have any causal powers, save those grounded in its constituents. The principle of causal grounding states that the causal efficacy of any complex, whether functionally or mereologically described, is entirely dependent upon the causal efficacy of the basic constituents of its physical instantiation. The problem is now worrisomely clear. Does consciousness have any causal power in the world? If the causal powers of conscious systems obey the principle of causal grounding so that the causal powers of any conscious system are entirely dependent upon the powers of its instantiation then, seeing as consciousness is a fundamental feature of the universe which cannot be reduced to its instantiations, consciousness has no efficacy in the world — consciousness turns out to be completely epiphenomenal. On the other hand, if this conclusion is resisted and some independent causal power is granted to consciousness, then some assemblages of physical parts have causal powers that don't depend entirely upon the causal powers of those parts. This is what philosophers call radical emergentism.8 Only here we have an ultra radical form for it is not the mere assembling of physical parts into particular molar combinations that yields the emergent properties, but rather it's the assemblage managing to fulfil a certain abstract functional description that produces the miracle (we might call this the doctrine of radical functional emergentism). Neither horn of this dilemma is very attractive.
This problem of explanatory exclusion can also be seen to arise from another of Chalmers' principles: that equating the phenomenal character of conscious experience with 'information states'. Now, every physical state is an information state relative to some possible information receiver, and the causal differences which correspond to differences in the information encoded into any physical state are normally thought to obey the principle of causal grounding (this fact, of course, is what underlies our ability to exploit physical processes to transmit information). So again we have our dilemma: if conscious experience is isomorphic to information load then the causal powers of conscious experience are either (1) entirely dependent upon the physical properties of the information bearer or (2) some information bearers violate the principle of causal grounding. If (1) we have explanatory exclusion and conscious experience is epiphenomenal. If (2) we have another form of radical emergentism, now somehow dependent upon the information carried by the physical state in question. Again, neither horn is attractive.
It is also worth mentioning a familiar ambiguity in the notion of information, which may mean nothing more than the 'bit capacity' of a physical state or it can mean some semantically significant content carried by the transmitted bits. It is not clear to which sort of information Chalmers means to assign the phenomenal qualities of conscious experience, though what he says inclines me to the former interpretation. The bit capacity of the brain is no doubt gigantic, but it is obviously doubled by considering two brains as a single system, yet it is doubtful that there is a kind of third consciousness associated with the interaction of two human beings, even though these two brains then form a causally interacting system. So we have yet another generation problem: which information states actually yield consciousness, and why/how just those?
III: The Nature of Information
Chalmers conjectures that perhaps information is itself a fundamental feature of the world, which makes it a 'natural associate' of consciousness. But consciousness and information connect at the level of semantic significance, not at the level of bit capacity. Insofar as the classical theory of information is situated at the level of bit capacity it would seem unable to provide the proper (or any, for that matter) connection to consciousness. Furthermore, the classical theory treats information as a feature of certain causal processes (albeit very abstractly conceived) which is to say that information is a functional notion: information is embodied in causal processes that can be variously instantiated, and obeys the principle of causal grounding (so leading to the problems of explanatory exclusion discussed above).
We can begin to move towards a more radical view of the fundamental nature of consciousness with a move towards a more radical view of information. This view of information sees causal processes as one species of information transfer but does not expect that all information 'connections' will be restricted to such processes. The natural place to seek a notion of information like this is in Quantum Mechanics.
The role of 'pure information' can be illustrated by a simple discussion of the famous two-slit experiment. A beam of photons, electrons, atoms or whatever is directed towards an appropriately separated pair of slits in an otherwise opaque surface. A detector screen is set up behind the slits. QM predicts, and less ideal but more practical experiments amply verify, that the 'hits' on the screen will form an interference pattern, which results in some way from the interaction of the two possible paths an element of the test beam can take to the screen.
More particularly, the QM formalism demands that the state, @, of the particles in the beam be represented as a superposition of the states associated with each spatial path:
(1) @ = @1 + @2
(for clarity here, I leave aside the nicety of normalization — for details see the appendix). @1 represents the particle taking the left slit — call this path 1. @2 represents the particle taking the right slit — path 2. Now it turns out that when we calculate the probability of a particle hitting a certain region of the detector screen both possible paths contribute to this probability in a complex way; there is no way to regard the particle as in reality taking just one of these paths (though we are ignorant of which one) so that our calculation could proceed simply by adding up the probabilities generated by each path independently. This is revealed by the 'interference pattern' found on the detector after a sufficient number of particles have traversed the apparatus, which is strikingly different from the pattern expected if the particles were passing, one by one, through just one of the slits.
This feature of the two-slit experiment is well known and, as is equally well known, the interference pattern disappears if we have some way of determining which path the particles are taking. This is sometimes explained in terms of the disturbance of the particle's state which such a measurement will involve, and sometimes it is said that such a disturbance is unavoidable and is the proper account of this aspect of the phenomenon. But this is not what the theory says. There is no need to posit disturbance in order to explain the loss of the interference pattern; mere information about which path the particles take will suffice. For suppose that there was a perfect detector that could determine which path a particle has taken without altering the particle's state. Such a detector would be capable of only two output states, let's say L, R (for left slit and right slit respectively), and the output of the detector would be perfectly correlated with the components of the state, @1 and @2. There is nothing mathematically wrong with the idea of a perfect detector and the effect of the detector's presence will be to eliminate the interference pattern despite having no effect on the state of the particles. This is of some interest to those who need to be reminded that complementarity is not the result of the clumsiness of measurement, but is rather an intrinsic and ineradicable feature of QM. The theory maintains that the mere fact that our detectors carry the relevant information is sufficient to destroy the interference effects, whether or not the detector in some way 'disturbs' the system under measurement.
The kind of information at issue here is not bit capacity but the semantically significant correlation of 'distinct' physical systems, where there is no requirement that the correlation be maintained by some causal process connecting the two systems. This remarkable feature of QM is made more apparent by a device hinted at by the notion of a perfect detector. What about the possibility of retrieving the original interference patterns simply by erasing the information within the detector? Since the particle states have not been altered by the initial operation of the detectors, this would appear to be at least theoretically feasible. To speak figuratively: the particles, now far along on their way towards the screen upon which their position will eventually be recorded, have no idea whether their paths have been registered or not. Such an interference retrieval device is called a quantum eraser (see Scully and Drühl, 1982; Scully et al., 1991; also Englert et al., 1994; Seager, forthcoming). Quantum erasers are somewhat delicate creatures, for they must obey the stricture that no 'bit channel' be established between the systems in question — in this case the operation of the detectors and the target screen (for this could violate a fundamental stricture of the theory of relativity that no information be transmitted faster than the velocity of light). Nonetheless, they can be, at least theoretically, realized. Let us imagine that we have added to our two-slit plus perfect detector apparatus some unspecified method of erasing the information in the detector (this is actually the delicate part of constructing an eraser).
The analysis of the quantum eraser does indeed reveal that the interference patterns can be retrieved by simply erasing the information in the, detectors, and this without the establishment of a bit channel (again, see the appendix for details). The operation of the eraser strongly suggests that each particle is responsive to the state of the eraser. But there can be no question of any causal process between the particle and the eraser operation if we arrange our experiment properly (we could, for example, make the distance between detector screen and the eraser so great and delay the operation of the eraser so long that a light signal could not reach the particle from the eraser before it got to the screen).
The natural interpretation of both the quantum eraser and the simpler, basic two-slit experiment is that there is a noncausal, but information laden connection amongst the elements of a quantum system. And this connection is not a bit channel or any sort of causal process (which shows once again, incidentally, that we are dealing here with a semantic sense of information). Here, perhaps, we find a new, nontrivial and highly significant sense in which information is truly a fundamental feature of the world (maybe the fundamental feature).
IV: Panpsychism
It seems to me possible to use this more robust sense of the fundamental nature of information to mold a theory which takes consciousness to be itself a fundamental feature of the world, where I mean by fundamental something elemental, not dependent upon the satisfaction of any functional description by any physical system, and not subservient to the principle of causal grounding. Chalmers himself makes a gesture towards such a theory in his remarks on information and notes that such a theory is 'not as implausible as it is often thought to be' (p. 217). We might as well be blunt about it: the theory at issue is panpsychism, which is the doctrine that 'all matter, or all nature, is itself psychical, or has a psychical aspect' (this from the OED), and it is indeed thought to be implausible. I offer a defence of it only with great diffidence. The generation problem seems real to me and sufficiently difficult to warrant fairly untrammelled speculation. Several strands of thought, some in defence of and some attacking panpsychism also come together in a curiously satisfying way once we unite the ideas that consciousness is a foundational feature of the world with our new notion of information and its significance.
I said above that on Chalmers' account, consciousness is a radically emergent phenomenon and hence is fundamental only in the sense that it cannot be explained in terms of the properties of the relevant complex systems that exhibit it. Chalmers is also adamant that consciousness cannot be reduced to these subordinate properties. It was noted some time ago, by Thomas Nagel (1979), that the denial of radical emergentism coupled with nonreductionism seems to entail panpsychism. The argument is straightforward: if consciousness is not reducible then we cannot explain its appearance at a certain level of physical complexity merely in terms of that complexity and so, if it does not emerge at these levels of complexity, it must have been already present at the lower levels.9 Thus, if we are to reject a radical emergentism and yet respect the generation problem we will be naturally driven to panpsychism.
Panpsychism has seen better times. Perhaps it was the favoured doctrine of our forebears, echoed in the animism of many prescientific cultures. The polymath philosopher Leibniz endorsed a form of panpsychism, essentially for the reasons given by Nagel. But panpsychism was always at the fringe of scientific/philosophical respectability and tended to lose whatever respectability it possessed as the scientific understanding of the world expanded. So it is somewhat ironic that the revolution in biology wrought by Darwin occasioned a rekindling of interest in panpsychism. In a paper which still retains interest W. K. Clifford (1874)10 presented an argument that was evidently in the air: the theory of evolution's application to the mind requires that some element of consciousness be present in all matter. Clifford says of consciousness, in recognition of the generation problem, that
we cannot suppose that so enormous a jump from one creature to another should have occurred at any point in the process of evolution as the introduction of a fact entirely different and absolutely separate from the physical fact. It is impossible for anybody to point out the particular place in the line of descent where that event can be supposed to have taken place. The only thing that we can come to, if we accept the doctrine of evolution at all, is that even in the very lowest organism, even in the Amoeba which swims about in our own blood, there is something or other, inconceivably simple to us, which is of the same nature with our own consciousness ... (1874, p. 266).
Is this not Nagel's argument in a nutshell? Emergence is impossible, reduction is absurd — so elements of consciousness must be found in the basic construction materials of the universe (in Clifford's restriction of his argument to organisms we see a vitalistic error, for the generation problem will arise no less for the gap between organism and nonorganism than for any gap in the intraorganism hierarchy). The addition of the theory of evolution which gives, or at least, at the time of Clifford, postulates, a palpable mechanism by which the simple is differentially compounded into the complex adds impetus to the slide towards a true panpsychism.
On the other hand, one can raise potent objections against panpsychism. Perhaps the single most concentrated and insightful attack on panpsychisin is found in William James's Principles of Psychology (1890). James vigorously scourges the view he derisively terms the 'minddust' theory and presents what I think is the most difficult problem facing any panpsychist theory of consciousness. I will label this (1) the combination problem, which is the problem of explaining how the myriad elements of 'atomic consciousness' can be combined into a new, complex and rich consciousness such as that we possess. Isn't this just the generation problem all over again? James is very good on this:
Take a sentence of a dozen words, and take twelve men and tell to each one word. Then stand the men in a row or jam them in a bunch, and let each think of his word as intently as he will; nowhere will there be a consciousness of the whole sentence. We talk of the 'spirit of the age' . . . but we know this to be symbolic speech and never dream that the spirit . . . constitute[s] a consciousness other than, and additional to, that of the several individuals whom the word 'age' denote[s] (p. 160)
Or again,
Where the elemental units are supposed to be feelings, the case is in no wise altered. Take a hundred of them, shuffle them and pack them as close together as you can (whatever that might mean); still each remains the same feeling it always was, shut in its own skin, windowless, ignorant of what the other feelings are and mean. There would be a hundred-and-first feeling there, if, when a group or series of such feeling were set up, a consciousness belonging to the group as such should emerge. And this 101st feeling would be a totally new fact; the 100 original feelings might, by a curious physical law, be a signal for its creation, when they came together; but they would have no substantial identity with it, nor it with them, and one could never deduce the one from the others, or (in any intelligible sense) say that they evolved it (p. 160, original emphasis).
In sum, James thinks that the second fundamental posit of panpsychism — that units of experience can merge into higher forms of experience — without which panpsychism offers no escape to those enthralled by the generation problem is 'logically unintelligible' (p. 158).
If James is right then the combination problem points to a distinctive generation problem in panpsychism which is formally analogous to the problem of generating consciousness out of matter. Panpsychism will have no advantage over physicalism if essentially the same problem lurks at its heart, and of course, it faces the intrinsic implausibility of asserting that atoms are conscious (in whatever degree you like — it remains undeniably implausible). If James is right then nothing whatever is gained by the first postulate of panpsychism, and hence the utility of making it in the first place is entirely undercut.
Another objection flows from this one11 which might be called (2) the unconscious mentality problem. One might be inclined to avoid the implausibility of the first posit by accepting the mentality of the elemental units of mind while denying that they are actually conscious experiences. But this would of course leave the generation problem unsolved and might even be thought to exacerbate it, for how are we to account for the generation of conscious experience from the combination of nonconscious entities, even if they are in some sense mental entities. In this case, panpsychism faces a problem which is strictly analogous to the generation problem facing physicalists.
Yet another serious problem arises upon considering the role of mentality in the workings of the world. One might expect that a fundamental feature as significant as consciousness should take some part in the world's causal commerce. But if it does play such a role, then we should expect it to turn up in our investigation of the physical world; we should expect, that is, to see physically indistinguishable systems at least occasionally diverge in their behaviour because of the lurking causal powers of their mental dimension. In that case, our physical picture of the world is radically incomplete and many would find this extremely implausible. I often have to worry about whether my car will start, but I thankfully don't have the additional worry about its failing to start even when there is absolutely nothing mechanically wrong with it but just because it 'feels like' staying in the garage today! Let's call this (3) the completeness problem. I will reserve my replies to these objections until later, but one unsatisfying reply to (3) should be discussed here. A panpsychist could urge that physically identical systems will have, in virtue of their physical identity, identical mental features and so physically identical systems will always behave in exactly similar ways even if the mental aspect is providing some of the driving force. This is unsatisfying because it immediately raises the explanatory exclusion problem: what ground for positing any mental influence at all if the physical properties of the system can account for all its behaviour? The mental then becomes, at the very least, explanatorily epiphenomenal and threatens to be a truly superfluous appendage. So the problem is that either panpsychism asserts that our physical picture of the world is incomplete or that mentality is explanatorily epiphenomenal. The first horn is implausible and the second undercuts much of the point of the panpsychist enterprise.
Finally, there are the two simplest objections. We have (4) the no sign problem: there is no evidence whatsoever of a nonphysical dimension to the elemental units of nature and, (5) the not-mental problem: if there was some feature of these units we chose to label as 'mental', what possible ground could one provide to justify this label. Surely we would like to see some 'sign' of mentality, as such, in the basic features of the world before we could think there was any real content to the doctrine of panpsychism.
V: The Quantum Panacea
There is a coherent view of panpsychism that can go some way towards answering all of these objections. I want to examine them back to front since it seems to me that by and large they were presented in order of decreasing difficulty. As to (5): if one takes consciousness to be a truly fundamental feature of the world then it will not seem odd that it might manifest itself in regions remote from our normal encounters with it. There is no apparent sign of any gravitation between subatomic particles but since we take gravitation to be fundamental we are willing to accept that the gravitation force between two electrons really does exist. But we must always remember that those philosophers who deny that there is any generation problem for consciousness will be likely to regard the ascription of consciousness to anything that gives no behavioural sign of consciousness as more than implausible but utterly unintelligible.12 I have tried to argue above that the generation problem is a real problem and this means that one can postulate with at least bare intelligibility that consciousness is a fundamental feature of the universe.
And this provides something of a reply to (4) as well. For if the analogy with gravitation is acceptable, then we would expect that the effects of the 'degree' of consciousness associated with the elemental units of physical nature would be entirely undetectable. There is no requirement that fundamental features provide operationally observable effects at every possible scale. This reply may be sufficient, but it also may not be necessary for, significantly, it is not entirely clear that the elemental units present absolutely no evidence of their postulated nonphysical (and indeed mental) aspect.
To explain what I mean is to address (3). I think it is reasonable to expect that a truly fundamental feature of the world should take a distinctive causal role in the world. And so we would expect that a picture of the world that is expressed in purely physical terms, without making any reference to this fundamental feature, would be incomplete. Occasionally, that is, the world should act in ways that are inexplicable from the purely physical viewpoint. No one really knows whether human thoughts and actions are entirely determined by physical features, so no one really knows whether human behaviour is purely physically determined either. But let us regard the elemental units of physical nature and see if they ever act in a way that is inexplicable from a purely physical standpoint. Of course they do — the Quantum theory insists upon this. As a physical theory, QM asserts that there is no explanation of certain processes since these involve an entirely random 'choice' amongst alternative possibilities. The world's behaviour does leave room for an additional fundamental feature with its own distinctive role.
There are various proofs that QM cannot be extended into a fully deterministic physical theory.13 As I understand it, these proofs all depend on disputable assumptions. But we might nonetheless take them as at least a sign of the ineradicable incompleteness of a purely physical picture of the world.
It will be urged, along the lines of (5), that the incompleteness appealed to here has absolutely no relation to consciousness, but this is not entirely clear. If we ask what features of the world our elemental units seem to respond to, one major influence is information. In the two-slit experiment, we might say that the particles are informed about the results of the perfect detector; in the quantum eraser the particles are informed whether information has been erased or not, in the demonstrable absence of any causal connection between them and the eraser. Responsiveness to information is hardly foreign to the realm of mentality although here it applies in an admittedly very circumscribed and impoverished sense, but this is to be expected of a fundamental feature manifesting itself at an elemental level. It may be worth repeating here that the kind of information at issue is not just the bit capacity of classical information theory but something more like semantically significant information and this is a notion of information more akin to mentality.
On this view, the elemental units of physical nature possess a mental aspect which enjoys a distinctive causal role in the behaviour of those units. Thus it grasps the first horn of the dilemma: the physical viewpoint is incomplete. But, I urge, at the level of the elemental units, the physical picture of the world does indeed look incomplete. And it may be that this incompleteness extends upward through the complex hierarchy of physical composition.
Reflecting upon the composition of more complex physical entities brings us naturally to the most difficult problem facing panpsychism: the combination problem. For while it is manifest that the basic physical elements combine in a multitude of ways to produce molecules, proteins and people, it is far from clear that it even makes sense to speak of the combination of basic mental elements, even granting they are in some sense conscious,14 into distinct and more complex conscious experiences.
I doubt that the difficulty of the combination problem can be completely overcome, but I think that a fairly natural response to it springs from a little deeper look at the metaphysical presuppositions underlying James's position. According to James, the combination problem stems from a very general consideration:
no possible number of entities (call them as you like, whether forces, material particles, or mental elements) can sum themselves together. Each remains, in the sum, what it always was; and the sum itself exists only for a bystander who happens to overlook the units and to apprehend the sum as such; or else it exists in the shape of some other effect on an entity external to the sum itself. Let it not be objected that H2 and O combine of themselves into 'water', and thenceforward exhibit new properties. They do not. The 'water' is just the old atoms in the new position H-O-H; the 'new properties' are just their combined effects ... (pp. 158-9.)
Or again:
Just so, in the parallelogram of forces, the 'forces' themselves do not combine into the diagonal resultant; a body is needed on which they may impinge, to exhibit their resultant effect. No more do musical sounds combine per se into concords or discords. Concord and discord are names for their combined effects on that external medium, the ear. (p. 159.)
I won't dispute that such a view has a certain attractiveness; it seems no more than a reasonable generalization of the mereological reductionism of which the world provides so much evidence. But we know it to be false. The most startling revelations of its error spring, as the reader knows or guesses, from QM. Consider again the two slit experiment. It is the most natural assumption in the world to regard the particles as they pass through the two slits to form a mixture which contains one-half the particles in state @1 (a state representing, recall, the particle as having passed through the left slit) and one-half the particles in state @2 (particle having passed through right slit). But they do not. They instead are in the superposition of the two possible states, @1 + @2, and the superposition is a 'combination' of states which itself forms a genuinely new state with properties observably different from the properties of the mixture. Quantum wholes are not just the sum of their parts.
Yet the ability to enter into superpositions like @1 + @2 is a reflection of the properties of the elements that enter into it, so the notion of mereological reductionism is not to be expunged from our philosophy altogether, which is surely a good thing for this sort of reductionism ties at the heart of our notion of scientific explanation itself. However, we cannot accept the principle of mereological composition espoused by James and thus there is no argument from general principles against the panpsychist's combinations of elemental mental units into distinctive mental wholes.
VI: Further Speculation
Thus can the philosophical objections against panpsychism be answered. The kind of panpsychism I have envisaged states that the physical world-view is incomplete, as evidenced by the fact that physically identical systems can nonetheless act in different ways. The 'hidden variable' is not physical but a form of elementary consciousness — but, as Clifford remarks, the kind of consciousness 'which goes along with the motion of every particle of matter is of such inconceivable simplicity, as compared with our own mental fact, with our consciousness, as the motion of a molecule of matter is of inconc6vable simplicity when compared with the motion of the brain' (p. 267). This is the 'psychist' part of the picture. The 'pan' part of the picture comes to the assertion that consciousness is an utterly fundamental feature of the world: not one element of physical reality is lacking its associated mental aspect. These mental elements combine according to some principle by which the summing together of parts yields more than just the assemblage of parts in causal interaction, just as do the physical elements.
We might speculate that there is a connection between the summation principles, so that in cases of superposition of states of physical elements we have mental combination as well. If we extend this idea to the case of multiparticle systems immersed in a non-ideal environment, which in truth we must, we arrive at the notion that quantum coherence might underlie more complex states of consciousness, for only coherent multiparticle systems will preserve the peculiar quantum mechanical properties that underlie the appropriate 'summation rules'. However, just how large systems could maintain coherence in the face of a highly energetic environment is quite unclear. Still, this idea has been espoused by a surprisingly large number of authors (see Michael Lockwood's, 1989, discussion of this issue; more recently Roger Penrose, 1994, has adopted it) but they fail to see the rather natural connection between panpsychism and their views. I mean that quantum coherence cannot solve the generation problem satisfactorily, but it might solve the combination problem.
In any event, a series of yet more speculative ideas suggest themselves if we entertain this approach, which I would like to sketch here in conclusion. The first idea is that only systems that can maintain quantum coherence will permit 'psychic combination' so that complex states of consciousness will be associated only with such systems. Given that the brain supports animal and human consciousness, the brain (or some significant part of it) is such a quantum system (this is almost the hypothesis of Penrose in Shadows of the Mind (1994); it differs from Penrose in that it denies that there is any new physics behind the phenomena). On the other hand, we may suppose that envisagable computers will not sustain quantum coherence; let's say that they are devices that deamplify quantum effects, and so they will not support complex, unified states of consciousness. Of course, an intentionally designed quantum computer, if such can be constructed, would not necessarily suffer from this weakness.15 Here we have a modern reincarnation of an old idea, which goes back at least to Leibniz, distinguishing unified entities, or what Leibniz called organisms, from mere aggregates.
It might be objected that most quantum coherent systems could hardly be thought to sustain any kind of complex life of consciousness, as for example a pot of liquid helium. This seems a cogent objection fostering further speculation: could we not imagine that the nature of the combinatory consciousness is connected to the informational structures of the physical system at issue? The essential simplicity of the structure of liquid helium is informationally impoverished as compared to the complex structures of the brain (even though, we are assuming, both maintain coherence). Thus while our speculative panpsychist ought to (in fact, has to) admit that the liquid helium does indeed have an associated unified state of consciousness, it would remain an extremely primitive state of consciousness, perhaps not so different from the state of consciousness associated with the single lowly helium atom.
Looking at this point from another direction, modem computers have an informationally rich internal structure, which is what permits the complex range of behaviour in which they can engage. Yet since they are quantum deamplifiers, panpsychism, (at least of the stripe we are discussing here) denies that they have any conscious mental life. Thus the panpsychist might support Searle's contention that computers have no understanding, which he draws from his famous 'Chinese Room' thought experiment (Searle 1980), and at the same time explain it, in terms a little clearer than Searle's brute insistence that the brain 'secretes intentionality'.
I must reiterate my diffidence in presenting such speculations. To borrow a phrase from Nagel, they all too obviously reek of the 'faintly sickening odour of something put together in the metaphysical laboratory'. The panpsychism offered here is a purely philosophical theory; as it stands, it has no distinctive empirical consequences. Still, I find it remarkable that a number of issues involved in the question of consciousness get a surprisingly unified treatment under panpsychism. It does seem to me that the acceptance of the reality of the generation problem and the subsequent perception of its extreme difficulty leads quite naturally, as Chalmers notes, to the idea that consciousness is a fundamental feature of the world. I would like to urge that panpsychism is the most natural way to incorporate consciousness as truly fundamental. But I would actually expect the argument to lead many back towards the difficult task of denying the reality of the generation problem.
Appendix: The Two-slit Experiment and the Quantum Eraser
The state of a particle after passing through the two-slit experiment is:
(1) @ = 1/Ö2 (@1 + @2)
The probability of a region of the detector screen, r, being hit be a particle is given by the inner product:
(2) <@|Pr@>,
where Pr is an operator which projects onto the subspace representing those states in which the particle is found in r. The full expansion of (2) reveals the interference terms:
(3) @d = 1/Ö2[<@1|Pr@1> + <@2|Pr@2> + <@1|Pr@2> + <@2|Pr@1>].
The first two terms respectively represent the probability of the particle being in region r if it takes path 1 or if it takes path 2. The final two terms are the unavoidable cross terms which account for the interference pattern.
Now we introduce our perfect detectors. The particle plus detector state, after passing through the apparatus and being detected, would be written as a superposition of tensor products so:
(4) @d = 1/Ö2[(@1 Ä L) + (@2 Ä R)]
Now if we wish to compute the probability of finding a particle in region r, we require an operator that works on the tensor product space of the particle plus detector; this operator is Pr Ä I, where I is the identity operator. The probability of finding the particle in region r is now:
<@d|(Pr Ä I)@d>
Written out in full this gets rather messy:
(6) <1/Ö2[(@1 Ä L) + (@2 Ä R)]|(Pr Ä I)1/Ö2[(@1 Ä L) + (@2 Ä R)]>,
but if we abbreviate (@1 Ä L) to X, (@2 Ä R) to Y and the operator (Pr Ä I) to O, the fundamental form becomes apparent:
(7) <1/Ö2(X+Y)|O1/Ö2(X+Y)>
which is analogous to (2) above. However, when (7) is expanded the cross terms take on a distinct form; the first step gives us:
(8) 1/2[<X|OX> + <Y|OY> + <X|OY> + < Y|OX>].
The expansion of just the first and last term of (8) should be enough to reveal what will happen to the probabilities in this case. The noncross term case goes as follows:
| (9) <X|OX> | = <@1 Ä L|(Pr Ä I)(@1 Ä L)> |
| = <(@1 Ä L)|(Pr@1 Ä L)> | |
| = <@1|Pr@2> × <L|L> |
Since all our state vectors are normalised, <L|L> = 1 and (9) is simply the probability of the particle being in region r if it took the first path. As we would expect, the detector state has no effect on this probability.
Consider now a cross term of (8), say <Y|OX>:
| (10) <Y|OX> | = <@2 Ä R|(Pr Ä I)(@1 Ä L)> |
| = <@2 Ä R|(Pr@1 Ä L)> | |
| = <@2|Pr@1> × <R|L> |
This cross term is accompanied by the factor <R|L> (the other cross term of (8) win of course be accompanied by <L|R>). But in a perfect detector, distinct indicator states are orthogonal, so these inner products have the value 0 and the interference terms thus disappear. The probability that the particle will be found in region r is now just the surn of the probability of its being in r if it takes the path I and the probability of its being in r if it takes path 2.
The introduction of the quantum eraser is somewhat complex and depends upon a 'mathematical trick' (see Scully, Engleret and Walther (1991)). Four new states are defined as follows:
@+ º 1/Ö2(@1 + @2)@+ º 1/Ö2(@1 - @2)
G+ º 1/Ö2(R + L)
G- º 1/Ö2(R - L)
The states G+ and G- are to be thought of as states the detector can enter through the operation of the eraser. The original state, @d, can be written in terms of our new states:
(11) @d = 1/Ö2[(@+ Ä G+) + (@- Ä G-)>,
as can be verified from the properties of the tensor product. So this state exhibits no interference since the cross terms contain the vanishing <G+|G-> and <G-|G+>.
The action of the eraser is revealed by considering the probability of the particle being in region r given that the detector is in the state G+. On the assumption that the detector is in G+ the second term of (11)'s left side must vanish and the probability will be calculated from the state @+ Ä G+. The probability of the particle being in region r given that the detector is in state G+ is:
(12) <@+ Ä G+|(Pr Ä I)(@+ Ä G+)>,
which reduces to
(13) <@+|Pr@+> × <G+|G+>.
Since <G+|G+> = 1 the probability we seek is simply <@+|Pr@+>. But, given the definition of @+, this probability expression expands into (3) above. We have recovered the original two-slit configuration with its interference effects despite the operation of the detector and we have done so via the operation of the eraser!
But no bit channel is set up. For consider the probability on the alternative assumption that after the operation of the eraser the detector goes into state G-, which will be equal to <@-|Pr@->. This state generates interference effects too, but they are precisely opposite to those of (13). So unless we know which of G+ or G- the detector system goes into we cannot 'see' any interference pattern, and there is no way to transmit information about the detectors save by ordinary means.
Footnotes
1 The generation problem has been around for along time; a very clear formulation is given by John Tyndall (as quoted by William James): 'The passage from the physics of the brain to the corresponding facts of consciousness is unthinkable. Granted that a definite thought and a definite molecular action in the brain occur simultaneously; we do not possess the intellectual organ, nor apparently any rudiment of the organ, which would enable us to pass, by a process of reasoning, from one to the other' (quoted in James, 1890, p. 147, from Tyndall, 1879).
2 But I note that the reviewer for the New York Times, George Johnson (1991), takes Dennett to be providing a generation theory of consciousness: '. . . from the collective behaviour of all these neurological devices consciousness emerges — a qualitative leap no more magical than the one that occurs when wetness arises from the jostling of hydrogen and oxygen atoms.' I very much doubt that Dennett would accept this view of his theory, for he patently never attempts to solve anything like the generation problem in his book (Dennett 1991).
3 For more on the mental lives of honey bees, see Griffin (1992).
4 Imagine we try to answer by, to take a current example, noting that the bee brain deploys the 40 hz oscillation binding system whereas the robot's processor does not. Then: how come only the 40 hz BS generates consciousness? Aren't other binding systems possible? Of course, this worry holds for any putative purely physical correlate of consciousness.
5 The set of 'brute facts' changes with the advance of science however. The velocity of light appeared to be a brute fact until Maxwell deduced it from independently measurable magnetic and electric parameters (and since 1983 the velocity of light has been a matter of definition). But there is little prospect of science altogether eliminating brute facts.
6 Thus Chalmers' position is a generalization of a view expressed by Richard Boyd (1980, p. 96): 'there are certain configurations such that whenever they are realized by a physical system, whatever substances compose it, the qualitative feeling of pain is manifested'.
7 I take this characterization and the general form of the problem from Jaegwon Kim (1994).
8 A doctrine interesting in its own right and especially popular earlier in this century. See ch. 8 of Kim (1994) for more on emergentism and its connection with the problem of explanatory exclusion.
9 I think that Nagel's argument is invalid, as it stands, because of an equivocation on the notion of 'reduction', which can be taken in either an episternic or an ontological sense. Chalmers is pretty clear that his notion of reduction is an ontological one (but see his remarks on p. 211) and this clarity rescues Nagel's argument (at the cost of making the 'no reduction' premise less secure). An alternative to panpsychism is, then, the view that while there is no explanatory relation between matter and consciousness — no solution to the generation problem that is — consciousness is, at the bottom of its being so to speak, a physical phenomenon. Such a view has been derisively labelled the New Mysterianism (by Owen Flanagan, 1992, whose attachment to neural correlates of consciousness I fear does not even begin to address the generation problem). In fact mysterianism is quite attractive if one accepts the seriousness of the generation problem while retaining an attachment to physicalism.
10 This is the same Clifford whose early speculations about the curvature of space prefigured General Relativity and the more radical programme that reduces matter to 'knots' of tightly curved space-time.
11 In these objections, I am indebted to Colin McGinn's critical remarks on David Griffin's manuscript (1994) defending a Whiteheadian style panpsychism as presented to a conference on Consciousness in Humans and Animals held at Claremont School of Theology last year. I would like to express my gratitude to Prof. Griffin for organizing this conference and inviting me, and to the other participants for their vigorous debate. I cannot deal with the intricacies of Process Philosophy paripsychism here but I thank Prof. Griffin for stimulating my interest in the doctrine.
12 As in Wittgenstein's example: 'You surely know what 'it is 5 o'clock here' means; so you also know what 'it's 5 o'clock on the sun' means. It means simply that it is just the same time there as it is here when it is 5 o'clock' (§ 350). If the metric of time was a really fundamental feature of the universe, as Newton seems to have believed, then there would actually be a sensible interpretation of '5 o'clock on the sun' even if we in fact did not have access to the universal temporal metric. But of course, the nature of time is such that it makes no sense to wonder when 5 o'clock on the sun really is. See Nagel's discussion of this point (1986, pp. 23 ff.).
13 See Jammer (1974) or Hughes (1989) for discussion.
14 This is the only answer to problem (2). The panpsychist must proclaim that it is consciousness itself that divides down to the elemental units. Otherwise the generation problem returns with its full force. But given the considerations adduced above which ameliorate its implausibility, there is no reason why the panpsychist cannot make this basic postulate.
15 See Deutch (1985) and Lockwood (1991,pp. 246-52) for discussions of this remarkable device.
References
Boyd, R. (1980), 'Materialism without reductionism: what physicalism does not entail', in Readings in the Philosophy of Psychology, vol. 1, ed. Ned Block (Cambridge, MA: Harvard University Press).
Chalmers, D.J. (1995), 'Facing up to the problem of consciousness', Journal of Consciousness Studies, 2 (3), pp. 200-19.
Clifford, W.K. (1874), 'Body and mind', Fortnightly Review. Reprinted in Lectures and Essays, ed. L. Stephen and F. Pollock (London: Macmillan, 1879), pp. 31-70.
Dennett, D.C. (1991), Consciousness Explained (Boston: Little, Brown and Co.).
Deutsch, D. (1985), 'Quantum theory, the Church-Turing principle and the universal quantum computer', in Proceedings of the Royal Society of London, A400, pp. 97-117.
Englert, B., Scully, M. and Walther, H. (1994), 'The duality in matter and light', Scientific American, 271, pp. 86-92.
Flanagan, O. (1992), Consciousness Reconsidered (Cambridge, MA: MIT Press).
Griffin, David (1994), Unsnarling the World Knot: Consciousness, Freedom and the Mind Body Problem, unpublished manuscript.
Griffin, Donald (1992) Animal Minds (Chicago: University of Chicago Press).
Hughes, R.I.G. (1989), The Interpretation of Quantum Mechanics (Cambridge, MA: Harvard University Press).
James, W. (1890), The Principles of Psychology, vol. 1, (New York: Henry Holt and Co. Reprinted by Dover Books, 1950).
Jammer, M. (1974), The Philosophy of Quantum Mechanics (New York: Wiley).
Johnson, George (1991), 'What really gores on in there?', New York Times, 10 November. Kim, J. (1994), Supervenience and Mind (Cambridge: Cambridge University Press).
Kirchner, W. and Towne, W. (1994), 'The sensory basis of the honeybee's dance language', Scientific American, 270, pp. 74-80.
Lockwood, M (1989/91), Mind, Brain and the Quantum (Oxford: Blackwell).
Nagel, T. (1979), 'Panpsychism', in Mortal Questions (Cambridge: Cambridge University Press).
Nagel, T. (1986) The View From Nowhere (Oxford: Oxford University Press).
Penrose, R. (1994) Shadows of the Mind (Oxford: Oxford University Press).
Seager, W. (forthcoming). 'A note on the quantum eraser', in Philosophy of Science.
Searle, J.R. (1980), 'Minds, brains and programs', Behavioral and Brain Sciences, 3, pp. 417-58.
Scully, M. and Drühl, K. (1982), 'Quantum eraser: a proposed photon correlation experiment concerning observation and 'delayed choice' in quantum mechanics', Physical Review A, 25, pp.2208-13.
Scully, M., Engleret, B. and Walther, H. (1991), 'Quantum optical tests of complementarity', Nature, 351, pp. II 1-1 6.
Tyndall, J. (1879), Fragments of Science: A Series of Detached Essays, Addresses and Reviews (London: Longmans).
Wittgenstein, L. (1953/1968), Philosophical Investigations (Oxford: Blackwell). My page reference is to the 3rd English edition, 196-8.