One of the central obstacles to a complete, unified understanding of the world is the mind-body problem, which in recent years has been generalized to the problem of naturalizing intentionality: how can meaning or aboutness be seen to be part of the natural world? One traditional approach to solving this problem, one that is implicit in much work in cognitive science, is to close the gap by making the mind more like the physical, the mechanical. Thus, computers have played an important role. An alternative hope is this: perhaps we can make the problem easier, not by seeing the mind as physical, as mechanical; but by seeing the physical as having mental aspects, even (or especially) at the lowest (quantum) level.
This kind of idea has been advocated by Bohm [Bohm, 1990], and has been furthered by Pylkkänen [Pylkkänen, 1992]. Both are concerned with the idea of active information, the kind of information about its environment that a quantum particle purportedly carries. On Bohm's ontological interpretation, there is no traditional wave-particle duality; rather, there is a particle, with a determinate position and momentum, accompanied by a new type of field which gives rise to a new potential, the quantum potential. The interference patterns are caused by subtle variations in the quantum potential through which the particle moves, which in turn arises from the quantum field which is influenced by the experimental configuration (e.g., number and position of slits, etc.) It is claimed that the trajectory of the particle is such that the particle can be said to carry information about its environmental configuration. It is argued that the particle must know what the slit configuration is in order to avoid hitting points on the plate for which the wave mechanics indicate a zero probability of a hit. The particle hitting the plate at a place that has a zero probability of a hit under a particular slit configuration means that the slits are not in that configuration. And, the idea goes, if even a lowly particle can be seen to involve such mental phenomena as `` meaning'', ``carrying information'' and ``knowing'', then perhaps the physical/mental chasm can be crossed.
However, this idea has a fundamental difficulty: causal or statistical correlation is not the same as knowledge or having meaning. Otherwise, we would have to say that a broken window means something about the stone that smashed into it. And if we have to say that, then it looks like meaning is everywhere, in every physical interaction. But to say that is to water down the notion of ``meaning'' to the point where the chasm opens up again (to mix metaphors a little). For now it will be difficult to explain how the special case of meaning in cognitive systems arises out of the more inclusive, almost trivial notion of meaning as any kind of causal interaction.
However, both Bohm and Pylkkänen attempt to establish a difference between the effect of the quantum field and that of a classical field by claiming that the former, and not the latter, is dependent upon the field's form, rather than merely its intensity. Nevertheless, I do not see how this alone can save the notion of active information. Their explication still ignores the principal difference between our notion of the physical, and our notion of the mental: the mental is normative. Thoughts can be correct or incorrect, right or wrong, true or false. But the kind of information that quantum systems seem to have is of the everday, non-normative kind; quantum systems can't be incorrect. A quantum state ``means'' just whatever caused it; there is no room for falsity or error. Contrast this with thoughts: I might have a thought ``That is a cow'', that was caused by me seeing something across a field. As it happens, the thing that caused my thought was in fact a horse. This does not mean that my thought means that there was a horse there. Rather, it continues to mean that there was a cow there, and is therefore false. Unless we can make sense of such notions in a quantum system, then we will still have the large dualistic gap between mind and world. We will still wonder how a non-normative physical system can be the same as a normative mind.
Pylkkänen replies to this objection [Pylkkänen, 1992, p 96,]:
It would, however, seem that it is not possible to speak about misrepresentation (an important aspect of human intentionality) in connection with the quantum field as long as we have not been able to discover a sub-quantum level. But if there were such a sub-quantum level, it would then be in principle possible that processes in this level could sometimes interfere with the functioning of the quantum field, and ``fool'' the electron in[to] believing, say, that the second slit is open even if in fact it isn't. The electron would then ``mistakenly'' act as if both slits were open. This would require that there be some other way of giving the necessary form to the quantum field than what we now know to be possible (i.e. we can presently give the quantum field the form required for the two- slit behaviour only by having both slits open).
Even if the (seemingly desperate) request for belief in the existence of a sub-quantum level - a level for which we have no evidence - is granted, the above response must face the disjunction problem [Fodor, 1991]. Pylkkänen wants to say that the meaning of the quantum field is that there are two slits open, even though there is only one slit open, because some sub-quantum process p affects the quantum field in such a way as to give it the shape that it usually has when two slits are open. But on what grounds do we single out this as the meaning of the quantum field? One could just as well say that the meaning of the field is a disjunction: either there are two slits open, or there is one slit open and p is occurring. This meaning, which is just as valid as the one Pylkkänen favours, is not false. Thus we have no reason to believe that active information can misrepresent, and thus we have no reason to believe that it can help explain human intentionality.
This is where quantum learning might be able to help. Dretske [Dretske, 1986, p 35-6,] has attempted to naturalize intentionality with the notion of learning. As said before, we can't get a notion of falsity going for a state if we just equate its meaning with whatever causes it. But suppose that we equate the meaning of a state s with some x which comes to cause s during a learning situation, a situation in which a system could possibly learn a relationship between x and some relevant behaviour. Then even if, later on, after the learning sitaution, some y different from x causes the state, the state will not mean y, because y was not a cause of the state in a learning situation. A biological example: on this account, a rat's brain state B, typically caused by a bell, means, for a rat that has undergone conditioning, that there is food present because food caused (or shared a common cause with) the bell during the learning situation. So if, after the learning situation, a bell rings because someone hit it accidentally, and causes state B (as is likely), then that state still means there is food present; it does not mean that someone hit it accidentally, even though that was the cause. Since there need not be food present in this case, we have the possibility of falsity. Therefore, a quantum learning system might similarly acquire some form of intentionality, and begin the bridging of the physical/mental gap.
A similar response to the disjunction problem might be given, that avoids all talk of learning. That is, all one has to do is give a principled way of determining which situations are meaning- fixing situations, and one can then solve the disjunction problem. My above solution, following Dretske, proposes that the meaning-fixing situations are learning situations. But another suggestion is this: the meaning-fixing situations for a quantum system are the ones in which the quantum level gives a complete, correct characterization of the system. This would mean that the quantum field in Pylkkänen's example does indeed mean that both slits are open, since in the purely quantum cases, cases in which the sub-quantum does not interfere to yield a violation of quantum-mechanical laws, the quantum field in question is indeed the result of only one configuration: two slits being open. The situations in which the sub-quantum process p, together with one slit being open, yields that very same quantum field, is not a meaning-fixing situation, so the disjunctive meaning cannot be ascribed to the field. Thus, in the latter situations, the field is indeed false: it means something other than what is actually the case.
Although I have no direct rebuttal to this learning-free way of grounding quantum meaning, I do find it to be less satisfactory than my proposal involving learning. First, distinguishing meaning-fixing situations solely on the basis of whether quantum mechanics correctly describes the situation seems arbitrary and ad hoc. Learning has a non-arbitrary connection with intentionality and meaning; but why should respecting some quantum mechanical regularity be a source of normativity? We are left with a puzzle at least as great as the one with which we began. Furthermore, this principle could be applied at any level, resulting in an unacceptable ubiquity of meaning and intentionality. For if the quantum/sub-quantum reply is correct, then there would be no reason to reject a similar reply, one that appeals to a classical/quantum dichotomy. That is, one could say that all macroscopic events have meaning and intentionality as well: the meaning-fixing situations are ones in which classical physics gives a complete, correct account, and the possibility of misrepresentation is provided for by the possibility of situations involving macroscopic events that are influenced by the quantum level. It seems impossible to reject this proposal, and still support the quantum/sub-quantum proposal for grounding intentionality. A response based on learning, however, would not have such a problem, since it restricts meaning-provision to only those situations that involve learning.
Other approaches to naturalizing intentionality may suggest
other forms of quantum computation for those who wish to see
quantum systems as intentional. For example, the evolutionary
approach, e.g. [Millikan, 1984], proposes that a state s means that
p if s is the product of a process of natural selection, and
the explanation for why s was selected for was because it
was present when p was true. Thus, a quantum
implementation of genetic algorithms, with their
computational version of natural selection, might be another
way to get intentionality in at the most fundamental
level.
Once one has a notion of quantum information, one might use this show how information can be implicit, yet causally potent. The information that a particle has about its environment is not explicitly represented in the particle (there is no structure to the particle to provide an explicit representation), but it does have causal effect: it causes the particle to move in a particular way. Now one might think that one does not have to talk of information in this context at all; rather, we have action at a distance. The represented environment itself directly causes the action, so there is no need to invoke implicit information as a causal agent. But if we can make sense of a system being correct or incorrect, then we will not be able to say that what is represented is the direct cause, because the world might not be the way the particle's information takes it to be; the represented might not even exist. This idea of implicit but causally potent information has strong resonances with Bohm's idea of the implicate order [Bohm, 1980].