Philos Stud (2009) 142:371–386 DOI 10.1007/s11098-007-9192-y
Is knowledge a natural kind? Tuomas K. Pernu
Published online: 14 March 2008 Ó Springer Science+Business Media B.V. 2008
Abstract The project of treating knowledge as an empirical object of study has gained popularity in recent naturalistic epistemology. It is argued here that the assumption that such an object of study exists is in tension with other central elements of naturalistic philosophy. Two hypotheses are considered. In the first, ‘‘knowledge’’ is hypothesized to refer to mental states causally responsible for the behaviour of cognitive agents. Here, the relational character of truth creates a problem. In the second hypothesis ‘‘knowledge’’ is hypothesized to refer to mental states causally responsible for the evolutionarily successful behaviour of cognitive agents. Here, the problem lies in the fact that evolution by natural selection is not necessarily conducive to truth. The result does not necessarily amount to eliminativism, however, since the naturalist may consistently reject the condition of truth that lies behind these problems. Keywords Naturalism Naturalistic epistemology Causality Proximate/ultimate Cognitive ethology Truth Reference Eliminativism
1 Introduction The central concern of epistemology has typically been the theory of justification. Foundationalists (empiricists and rationalists) claim that only beliefs that are grounded on the proper beliefs or evidence, in a proper way, are justified. Coherentists, on the other hand, claim that only those beliefs that cohere with one’s other beliefs are justified. Naturalism can also provide its own theory of justification, namely reliabilism (e.g. Goldman 1986). But there is also a stronger T. K. Pernu (&) Department of Biological and Environmental Sciences, Physiology, University of Helsinki 00014, P.O. Box 65, Helsinki, Finland e-mail:
[email protected]
123
372
T. K. Pernu
form of naturalism in which the whole idea of the project of justification is questioned. Let us make a distinction between two types of naturalism. According to metaphysical naturalism, reality is one all-encompassing natural system whose structure science aims to reveal. According to methodological naturalism, science and philosophy should work together, sharing both the methods and the aims of research. And as a result, there are no philosophical questions outside of science; in particular, there is no ‘‘first philosophy’’, philosophy independent of science, which seeks to provide proper justification for all our beliefs. According to methodological naturalism, all attempts to justify science outside of itself have failed in the past and the project of justification should thus be abandoned. Although metaphysical naturalism and methodological naturalism are in principle two different doctrines they are obviously intimately connected. In particular, the idea to approach knowledge itself from the point of view of metaphysical naturalism is a natural consequence of adopting methodological naturalism. Since, according to methodological naturalism, the traditional epistemological project of justification is not viable, there is simply nothing else left than the task of providing a description of what knowledge factually is. According to Quine (1969a), after the rejection of the project of justification, ‘‘epistemology still goes on’’ simply by falling ‘‘into place as a chapter of psychology and hence of natural science’’ (p. 82). In this new situation epistemology ‘‘studies a natural phenomenon, viz., a physical human subject’’ (Quine 1969a, p. 82). Kornblith (1998, 1999a, 1999b, 2002) has since developed this project outlined by Quine (1969a) (cf. also Millikan 1984a). According to Kornblith (1999b), the proper ‘‘subject matter of epistemology is not our concept of knowledge, but knowledge itself’’; knowledge ‘‘should be viewed as a natural phenomenon’’ and it ‘‘should be investigated in just the same way in which we investigate other natural phenomena’’ (p. 327). In effect, Kornblith (1999b) suggests that ‘‘we should view knowledge as a natural kind’’ (p. 327). Let us take for granted that this project is sensible. Does it follow from this that it can also be carried out? Obviously not; it is only possible to view knowledge as a natural kind if there exists such a thing as the natural kind of knowledge. If knowledge was viewed simply as a natural phenomenon and investigated in just the same—empirical, no doubt—way as other natural phenomena, we might discover that there is not, in fact, any knowledge around. So the project is based on an assumption that there is such a natural phenomenon as knowledge. In what follows, it will be argued that this assumption is in fact unfounded.
2 The metaphysics of natural kinds The philosophical usefulness of the notion of natural kind is quite obvious. The homogeneous properties of a natural kind explain the efficiency of inductive inference (cf. Quine 1969b; Kornblith 1993): when a is known to be an instance of a natural kind A and when it is known that all As necessarily share the property P, it can be inferred that P(a). Furthermore, the notion of natural kind can be thought to
123
Is knowledge a natural kind?
373
ground scientific explanation: the fact that certain things reliably cause certain other things to happen can be explained by noting the homogeneous structure of the substances involved. And obviously this inference works also the other way around: the existence of a natural kind is inferred from the homogeneous causal power. So it is natural to think that the main purpose of science is to reveal the natural kind structure of reality (cf. Ellis 2001). One should not take the notion of natural kind too seriously, however. The classical view of natural kinds has incorporated such notions as objectivity, discreteness, intrinsicality, aspatio-temporality and essentiality. The idea is that the question whether a particular belongs to a class of certain natural kind should have a clear-cut and metaphysically determinate answer. Depending on one’s favourite flavour of empirical science, several of the classical conditions of natural kinds can be challenged (e.g. Griffiths 1999; Okasha 2002; Sober 1980). A better articulated recent development in the metaphysics of natural kinds is the introduction of the notion of homeostatic property cluster kinds (Boyd 1988, 1991, 1999; Keil 1989). According to this approach, natural kinds are property clusters maintained by homeostatic causal mechanisms. When the properties of a cluster kind are bound together by causal homeostasis, we can make reliable inductive inferences of a kind’s behaviour while denying that this inference would be grounded on a unique and well defined set of necessary and sufficient properties that constitutes the kind’s essence. So, in the case of gold (Au), we can say that particular substances are gold, not because they share a set of essential features, but because they share a cluster of properties maintained by causal homeostasis. By concentrating on causal homeostasis one rejects many of the conditions of the classical view of natural kinds (cf. Boyd 1999, pp. 143–144). Most importantly this view does not acknowledge any essential properties that all and only members of a certain kind necessarily share, since the set of properties in the homeostatic cluster is not fixed (and even the underlying homeostatic mechanisms may vary). From this it quite naturally follows that the boundaries between natural kinds may be vague: some entities will display only some properties of a cluster kind (and/or only some of the corresponding homeostatic mechanisms are present), and as a result it will sometimes be indeterminate whether an entity belongs to a certain kind or another. Naturally the properties in the cluster and its underlying homeostatic mechanisms may also vary in space and time. The notion of homeostatic property cluster kinds is no doubt much more appropriate for analyzing both scientific and vernacular classification. In particular, it allows the kinds to evolve which is important both scientifically and psychologically. But at the same time the view is philosophically problematic: it relies on the notion of causation which itself is notoriously problematic. It remains to be seen whether this problem is fatal. One could, as a first defence, note that the problem is not as serious as it seems at first glance since what is crucial to this view is that certain homeostasis is upheld by some causal mechanisms. The exact nature of these mechanisms can be left open and it is also possible that there are many different types of such mechanisms. From the point of view of the current topic, the issue of whether one should favour one view of natural kinds or the other is quite inessential. When one is
123
374
T. K. Pernu
claiming that knowledge is a natural kind one is referring to an idea that knowledge constitutes a legitimate object of empirical inquiry, that knowledge is as an objective part of nature as any other object of empirical sciences. Using this particular term is just a way phrasing the issue. And if the notion of natural kind begins to be a burden to the argumentation on the issue itself, the burden could easily be removed by rephrasing the issue in some other terms. But there could be a lesson one should learn from the debate about the correct metaphysics of natural kinds. Whatever the view, one is adhering to the objective homogeneousness of the natural kind. The homeostatic property cluster view simply traces the source of this objective homogeneousness to homeostatic causal mechanisms. If one follows this train of thought, the question of whether knowledge is a natural kind is translated into a question of whether there are specifically epistemic (or ‘‘epistemically homeostatic’’) causal mechanisms. If one is to make an enquiry into the nature of the putative natural kind of knowledge, where would one start to look for the causal mechanisms that uphold the kind? Knowledge presumably has something to do with the behaviour and mental life of cognitive agents. When the causal background of behaviour is taken into consideration the following distinction is in place (cf. Mayr 1961, 1982). One can either talk about the immediate causes of the behaviour or causal (evolutionary) history behind this type of behavioural dispositions. This distinction in mind, in what follows two hypotheses are taken into consideration: knowledge as a proximate natural kind and knowledge as an ultimate natural kind.
3 Why knowledge is not a proximate natural kind Let us look at the following standard expression schema: A knows that P. This is the typical way in which the verb ‘‘to know’’ is used. If we were to conduct an investigation into the natural phenomenon of knowing it would be natural to treat it as a propositional attitude along with other such attitudes as believing, remembering, desiring, etc. One could thus form a hypothesis according to which knowledge is an ethological or psychological natural kind: knowledge is something that is needed in explaining the behaviour and psyche of cognitive agents. Kornblith (1999b, 2002) approaches the issue of knowledge as a natural kind in precisely this way. He cites studies in cognitive ethology. For instance, when an intruder approaches the nest of a piping plover (Charadrius melodus) the bird may do a broken-wing display: the bird acts as injured, cripples away from the nest attracting the intruder after itself, and flies off (Ristau 1991). One might be tempted to characterize this behaviour by using epistemic notions: the bird knows that its nest is under a threat; the bird knows that by acting in a certain way it may distract the intruder and eliminate the danger; and the bird knows when this is accomplished and when it can safely fly away; etc. One could immediately object that in such locutions epistemic notions are used equivocally and should not therefore be taken
123
Is knowledge a natural kind?
375
seriously. But at the same time it is also clear that epistemic notions are often used in exactly this way (cf. Ristau 1991). (And this fact about the use of ‘‘knowledge’’ is crucial, for if the whole idea is to investigate knowledge empirically one has nowhere else to start the investigation but the actual cases of knowledge attribution.) One could also object that epistemic ascriptions are inappropriate when such organisms as birds are concerned simply because any intentional ascription in such cases will be inappropriate (given that the behaviour of birds is merely mechanistic and non-intentional—and hence non-epistemic). This is an important point, not necessarily due to its correctness, but because the question of what distinguishes intentional behaviour from non-intentional is essential to cognitive ethology and philosophy of mind. But this is a separate question that will not be dealt with here (however, see Sect. 5 below). So, rather than asking, do we need to postulate intentional mental states in general, the task is now to ask whether we need to postulate mental states characterized as knowledge in particular in explaining the behaviour of organisms with representational capacities. Can one give any general constraints on intentional explanations? Let us make the standard distinction between two different notions of mental content in the following way: A mental (representational) state is narrow =def. the content of the mental state is individuated opaquely. A mental (representational) state is wide =def. the content of the mental state is individuated transparently. Two mental states, M1 and M2, are thus narrowly (type-) unidentical if and only if M1 is a mental state about A and M2 is a mental state about B, when A = B or A = B but there is no mental state about this co-referentiality. So, for instance, ‘‘A believes that Clark Kent has an X-ray vision’’ is a different mental state from ‘‘A believes that Superman has an X-ray vision’’, even though Clark Kent in fact is Superman but A is ignorant of this. On the other hand, M1 and M2, are widely (type-) unidentical but narrowly (type-) identical if and only if M1 is a mental state about A and M2 is a mental state about B, when A = B but there is a mental state about A = B. So, for instance, ‘‘A believes that water is wet’’ is a widely different mental state from ‘‘A believes that water is wet’’ if in the first case water is H2O and in the second case water is XYZ but A is ignorant of this fact (Putnam 1975). On the basis of this distinction the following argument can be formulated (cf. Fodor 1980, 1987): (i)
Psychological explanations are causal explanations: the attribution of a particular mental state to an agent explains why the agent behaves in a certain way. (ii) A science that provides causal explanations individuates the phenomena of its interest on the basis of causal efficacy: if M1 and M2 have the same causal roles M1 and M2 are one and the same phenomenon (instances of the same type of a phenomenon). (iii) Agents who have widely unidentical mental states but narrowly identical mental states have identical causal powers.
123
376
T. K. Pernu
Ergo: Wide and narrow mental states are one and the same psychological phenomenon. How plausible is this argument? To premise (i) one might object that psychological explanations may also have quite different aims than providing simple causal explanations. This may be so, but surely the central aim of intentional explanation is to specify the intentional causes of behaviour and action. And it should be quite clear that causality matters: if mental states would be causally irrelevant they would also be scientifically irrelevant. On the basis of the metaphysics of natural kinds there should not be anything wrong with (ii). On the contrary, it fits well even into the homeostatic property cluster view of natural kinds: two things are instances of the same natural kind if the two things are maintained by identical causal homeostasis. (A metaphysical argument for individuating properties on the basis of their causal powers is also given by Kim (1992) and Shoemaker (1980).) (iii) is an important point, although its truth may not be as evident as the truth of the first two premises. Let us take it for granted that causality acts locally on particles; i.e., there is no action at a distance. (The issue of locality is of course notorious in the philosophy of physics (cf. Bell 1964, 1966; Einstein et al. 1935; Frisch 2005; Lange 2002; Maudlin 1994; etc.) but its import to the current discussion is not clear.) Only a magic theory of intentional behaviour would allow events occurring outside the agent to figure as causes of the agent’s behaviour without going through the agent’s inner-structure. The notion of narrow content tries to get a grip exactly on the fact that agents have internal states the occurrence of which are causally responsible for the agents’ intentional behaviour. That is, narrow content supervenes on the internal physical or functional structure of the agent and hence is causally responsible for the behaviour of the agent. This, in fact, is an immediate and trivial result of the definition of narrow mental state. As Quine (1960, 1961) has put it, intensional contexts are marked by referential opacity. For example, the inference BaP(b) & b = c ) BaP(c) is not valid; the conclusion can be deduced from the premises only if an additional premise is added, namely: Bab = c. The following conclusion can thus be reached: intentional explanation needs only to deal with narrow mental states. (This conclusion is quite widely accepted. In addition to Fodor (1980, 1987) it has been endorsed e.g. by Block (1986), Egan (1999), Kim (1982) and Stich (1978, 1983).) Now, truth is necessarily a relational property. A proposition or a belief is true if and only if it is properly correlated to the obtaining facts (if some variant of a correspondence theory of truth is assumed). Truth, on the other hand, is a necessary condition of knowledge. An agent cannot know that P, if it in fact is not the case that P. According to the standard definition of knowledge, knowledge is justified true belief. There is an ongoing debate in epistemology about the nature of justification (prompted largely by Gettier (1963)) but not about the condition of truth. If truth is a necessary condition for knowledge, and if truth is necessarily relational, it would seem to follow that, if there are epistemic mental states, those states are wide mental states. But since wide mental states are causally non-efficacious, epistemic mental states would lack causal efficacy. But obviously causally non-efficacious entities
123
Is knowledge a natural kind?
377
should not figure in the metaphysics of science as natural kinds: if a given thing does no causal work in the universe it simply does not exist from the point of view of science. Hence, knowledge is not a proximate natural kind. This has been a difficult way to put an obvious truism: we only need to appeal to beliefs, not knowledge, in explaining the behaviour of cognitive organisms. (Even Kornblith (1999b, 2002) readily admits this.) A behaves in a certain way because it believes that so-and-so is the case: there is no way in which A could step outside its beliefs and check whether or not they are in accordance with the facts. In intentional explanation ‘‘knowledge’’ is totally superfluous. (Fodor (1980), Kim (1982) and Stich (1978) have noted exactly the same thing in passing.) So even though the standard phrase ‘‘A knows that P’’ would lead us to treat knowledge as a proximate natural kind, the hypothesis has to be rejected.
4 Why knowledge is not an ultimate natural kind Let us now consider another hypothesis. One could construct the following kind of transcendental argument: (i)
Since we are now here, our behaviour in the past has been reproductively successful. (ii) It is a necessary condition of reproductively successful behaviour that the system that guides behaviour is in accordance with the facts. Ergo: Our cognition is in accordance with the facts. After rejecting the ethological hypothesis Kornblith (1999b, 2002) makes exactly this move. According to him we need knowledge to explain ‘‘the possibility of successful behaviour’’: ‘‘knowledge explains the possibility of successful behaviour in an environment, which in turn explains species fitness’’ (1999b, p. 331). This is not a totally new way to approach epistemology naturalistically. Quine (1969b) already noted that ‘‘creatures inveterately wrong in their inductions have a pathetic but praiseworthy tendency to die before reproducing their kind’’ (p. 126). (i) is quite self-evident, although the exact trajectory of our evolutionary history is under debate (e.g., how much, and in what way, natural selection has among other evolutionary forces influenced our evolution). (ii) on the other hand is deeply problematic. It could be noted from the outset that one possible reason why the ultimate hypothesis should be rejected is to be found in the preceding discussion. Evolution by natural selection is interested solely in the behaviour of organisms: successful behaviour is rewarded with survival and with the possibility to reproduce; unsuccessful behaviour is punished with demise and extinction. But if the content of epistemic states is necessarily wide, and if the wide content of mental states is causally non-efficacious, it follows that natural selection is not interested in the epistemic states of the behaving organisms. So, according to this argument, if knowledge is not needed to explain the behaviour of an individual it is not needed to explain the successful behaviour of an individual.
123
378
T. K. Pernu
But things get more complicated when the evolutionary history of cognitive capacities is taken into account. Relational properties do make an evolutionary difference after all. In fact, fitness itself is a relational property. Fitness is relative adaptedness: A is more fit (better adapted) than B in an environment E if and only if A has a propensity to leave more descendants in E than B has (cf. Brandon 1990). Naturally, if E is varied, so is the fitness of A. So it appears that from the evolutionary point of view the situation is in fact turned upside-down: natural selection is not interested in what goes on within an organism; all it cares about is that the organism behaves in accordance with the facts. On the basis of this it might be possible to argue that it makes all the causal difference in the world whether an agent has true or false beliefs: agents acting on false beliefs die out. It would seem intuitively clear that in the course of evolution organisms having true beliefs are better off than organisms having false beliefs. But this intuition is in fact false. It mixes together two separate issues. True, all natural selection cares about is that the organism behaves in accordance with the facts. But ‘‘behaves in accordance with the facts’’ should not be equated with ‘‘behaves on grounds of true beliefs’’. An organism might simply behave in accordance with the relevant facts while having a large amount of false beliefs. The question is now whether there is any reason to think that instead of just tuning behaviour natural selection drives specifically true beliefs into fixation. It is clear that natural selection is not a force of evolutionary innovation but a force that fixes the phenotypical traits of a given population. (Population genetics acknowledges only mutation and recombination as the forces of evolutionary innovation.) Natural selection selects from the available alternative traits the ones that are best adapted to the given environment. In other words: natural selection tries to make the best out of what it has. On the basis of this there are many reasons why natural selection might fail to drive true beliefs into fixation. First of all, there might not be any true beliefs for natural selection to work on. In that case natural selection works purely on behavioural dispositions. But why would natural selection care for true beliefs if it already has all that it needs? That is, if a true belief would pop out, what difference would it make if the organisms were already tuned to behave appropriately enough? If, from the point of view of natural selection, behaviour is all that matters, then truth would again turn out to be superfluous. But more importantly, natural selection might even work against the fixation of true beliefs. There is no reason why the epistemically most optimal solution and the biologically most optimal solution should coincide. In fact, the opposite is more probable. Truth is expensive: a system that concerns itself with what is the case instead of behaving appropriately will no doubt spend more time on contemplating than on reproducing. (What we have here is the frame problem in ecological setting.) What natural selection favours is quick and dirty heuristics: solutions that work effectively and reliably rather than are truth conducive. One can convince oneself of the fact that epistemic and biological needs are two different things not necessarily in sync by considering the perceptual illusions, for instance. So, for example, in the case of Mu¨ller-Lyer illusion the perceptual system provides us with the information that one vertical line is longer than another when in fact they are of
123
Is knowledge a natural kind?
379
equal length—and the illusion persists even after we come to know (after measuring them) that the lines are in fact equal. Or consider the case of simple conditional learning. Take an experiment by Garcia et al. (1972), for example, where rats develop a strong aversion to a food of distinct flavour if right after being fed they are exposed to such doses of radiation that it makes them nauseous (some refer to this as Garcia effect (Stein 1996; Stich 1985, 1990)). The course of events could be put in intentional terms thus: the rats developed a false belief that a certain type of food is not suitable for eating. What again was favoured was appropriate behaviour, not true beliefs. No doubt this is exactly the way things usually go in the course of evolution. As Stich (1990) puts it: ‘‘from the point of view of reproductive success it’s often better to be safe (and wrong) than sorry’’ (p. 62). The cost of false positives is quite insignificant in comparison to the cost of a false negative. It would be tempting to think that if behaviour is successful it must be an outcome of a system of true beliefs. But biological facts speak against this. Appropriate behaviour may result from false beliefs and an effort to form true beliefs may, in fact, be a hindrance. This may be an uncomfortable conclusion since it leaves us in uncertainty about our epistemic status. In a similar situation Descartes (1637) appealed to God: even if he could not be directly sure about the truth of his most clear and distinct ideas, their truth is guaranteed by God itself since God (being benevolent and omnipotent) would not deceive him. As Cummins (1998) has acutely pointed out, ‘‘some contemporary theorists are tempted to replace God by evolution’’ (p. 122). It should thus be concluded that if a naturalist sees the appeal to God as a desperate ad hoc attempt to save our transcendent epistemic status, she should see the appeal to evolution in a similar light and reject the hypothesis that knowledge is an ultimate natural kind.
5 Epistemic eliminativism and the representational theory of mind Now, one might object that the argumentation above has proved too much, that it leads into total rejection of the representational theory of mind. For those who take the representational theory for granted this would constitute a reductio ad absurdum of the argument. The initial response might be that it is not advisable to take representational theory of mind for granted within the naturalistic framework. There are many reasons why one should be careful about interpreting the representational theory realistically (e.g. Kim 1998). So if the current argument gives more evidence for the possibility that the mentalistic paradigm is misguided it might be more prudent for a naturalist to accept this consequence than cling on to its denial. But could it also be possible to isolate the question of whether we should postulate epistemic mental states in particular from the question of whether we should postulate representational mental states in general? Would it be possible, in theory at least, to answer ‘‘no’’ to the previous question and ‘‘yes’’ to the latter? Let us sketch such a possibility.
123
380
T. K. Pernu
The problem boils down to the fact that representation is a relational notion: it is a relationship between the representation and the object represented. Together with the requirement that in psychological explanation one should rely only on narrowly individuated mental states, this has prompted some to embrace totally syntactical theory of mind (e.g. Stich 1983). Others have found it too difficult to give up on semantical theory of mind and have in consequence tried to construct internalistic semantics for mental representations (Fodor 1987; Jacob 1997) (the term ‘‘narrow content’’ was coined for referring to exactly such internalistically individuated mental representations). As already noted above, it is quite clear to people in this tradition that ‘‘knowledge’’ is a term that should not have a role in psychological explanations (Fodor 1980; Kim 1982; Stich 1978). But some have found it too difficult to give up on the essentially relational character of mental representations. In consequence, they have embraced an externalistic theory of mind that grants causal efficacy to widely individuated mental states (Dretske 1988, 1995; Millikan 1984b, 1989, 1993; Price 2001). What provides the causal efficacy to wide mental states, it is argued, is their selectionhistory: the states are selected for because they match the demands of the environment, and hence it is this relation between the state and environment that constitutes the metaphysical identity of these states. This theory has the somewhat paradoxical consequence that internalistically identical agents can nevertheless be in different mental states (Davidson 1987). And as result of this, even if this theory is adequate when type-individuated mental states are concerned, it is less credible when mental states are token-individuated: unless one gives up on the locality requirement of causal interaction one has to admit that in each individual case of mental causation, it is the internal properties of the mental state that do the final causal work. So it is questionable whether this theory is able to provide what we would expect from a theory of mind (and if not, what is left of the representational theory of mind). But we can set this highly problematic issue aside in this context. The question that is still left open is whether it is possible to accept the externalistic theory of mind and reject the idea of epistemic mental states at the same time. The main point in the argument against the ultimate hypothesis above is that it may often be biologically more prudent to act on false beliefs rather than on true ones. There is no theoretical reason why a proponent of the externalist view could not admit this. The externalistic theory of mind is concerned with the issue of assigning contents to mental states. According to it the contents of mental states should be individuated by their selection-history. This selection process might quite well involve mismatches between the content of the state and the environment, in the vein explained above. It must also be kept in mind that these theories do not operate solely with the notion of natural selection. It is essential that there is some causal process connecting the content of a mental state to its referent (establishing the ‘‘aboutness’’ relation). This process involves also various forms of subjective and intersubjective learning—and no doubt these are the more common ways of fixing contents. Now, it would not be a wise move to claim that all our intentional states are ‘‘about’’ the actual world (much less ‘‘true’’) even if the contents of our mental states are causally anchored to the actual world. One has to admit that we have false beliefs and that we have mental states whose contents do not have
123
Is knowledge a natural kind?
381
referents in the actual world—after all, one of the essential features of intentionality is that its objects can be nonexistent. So there is no theoretical reason why a proponent of this sort of an externalist theory of mind could not claim, on the one hand, that the intentional contents of mental states are fixed by selection processes, and that these processes might very well favour contents that have no referents in the actual world, on the other. A lot in this comes down to what we are prepared to regard ‘‘causally efficacious’’ in this context. Those with internalistic intuitions would admit that the elimination of epistemic mental states is a trivial consequence of the rejection of all externalistically individuated mental states. Internalism is based on a view that causal interactions are local interactions and that in attributing causal powers to entities we should only refer to the internal properties of these entities. Whether this assumption is ultimately valid or not is out of the scope of the current topic. But how problematic it may be, it is important to realize that externalist do not totally reject this train of thought. The claim is rather that externalistically individuated mental states are explanatorily or causally relevant: the relational features of mental states are relevant because ultimately they are responsible for the internalistically individuated mental states having the role they have. So even though in every token case of causal interaction the internalistically individuated features of the mental state do the concrete causal work, the externalistically individuated features are relevant because they explain why these internal features have the function they have. As some phrase it, behaviour can be ‘‘proximately mechanistic while being ultimately intentional’’ (Allen and Bekoff 1997). So the final question is whether this commits the externalist to argue that epistemic mental states are causally efficacious. Not necessarily; to an externalist externalistic individuation is essential in explaining behaviour but all relational features of mental states need not count as causally efficacious. There has to be something that sets ‘‘A believes that P’’ apart from ‘‘A knows that P’’ (intuitively it is the fact that the latter entails the truth of P). An externalist has to make a difference between these two types of states as well. So once the externalist story about propositional attitudes is ready, a further story would have to be told about the more specifically epistemic propositional attitudes. But one should be able to end the story with the first part and claim that whether the contents of the attitudes are true or not is irrelevant from the psychological point of view. After all, an externalist too has to give a story about how it is sometimes necessary to count mere beliefs as explanatorily relevant (i.e. in cases where the subject clearly has beliefs that are contrary to facts, and acts accordingly). Whether an externalist is able to give a credible account about this, is essentially the problem of misrepresentation and representational indeterminacy that the externalist is bound to face. But once such an account is given, the externalist can stop there and claim that epistemic mental states are superfluous as far as the causal powers of the subject are concerned. So, by showing how misrepresenting is possible and delimiting the scope of psychological explanation to mental states that potentially misrepresent, one could consistently claim that while externalistically individuated mental states are causally efficacious, epistemic mental states lack such power.
123
382
T. K. Pernu
So what is the further story that an externalist might have for the causal relevance of epistemic mental states? It is, for example, the story told in the argument above: knowledge would explain the possibility of successful behaviour. As noted, it is quite evident in this argumentation that if we are only interested in explaining behaviour, it is sufficient to appeal to mere beliefs (as Kornblith (1999b, 2002) admits). So the thing that sets the attribution of ‘‘A believes that P’’ apart from ‘‘A knows that P’’ is that the latter is needed when successful behaviour, rather than behaviour in general, is encountered. But, as it is argued above, this is not necessarily so (if ‘‘A knows that P’’ is taken to entail the truth of P). But whatever one’s attitude towards these arguments are, it is clear that explaining behaviour (on the grounds of ‘‘A believes that P’’) and explaining successful behaviour (on the grounds of ‘‘A knows that P’’) are two separate things. One should be able to buy an externalist story about the first thing and refrain from going along with the story offered for the second one. It is clear that the arguments in the two separate hypothesis considered here have parallels in the arguments concerning the nature of mental representations. Those who are more attracted by internalistic arguments are no doubt more attracted by the eliminativistic argumentation here than those attracted by externalistic arguments. But these issues should be kept separate. The jury is still out with the case of internalism versus externalism in philosophy of mind. But whatever the verdict, it will not oblige anyone to reject epistemic eliminativism.
6 The semantics of ‘‘knowledge’’ A case has been made for the rejection of the hypothesis that knowledge is a natural kind. Before establishing this as a conclusion a caveat must be taken into account. Let us think of science as a set of terms (predicates rather than singular terms). If the main purpose of science is to reveal the natural kind structure of reality it is natural to think that science itself is (at least) a collection of natural kind terms. Let K now stand for ‘‘knowledge’’. The conclusion of the preceding discussion can then be expressed as follows: K fails to refer, and should hence not be included in the set of terms that constitutes science. In other words, K should be eliminated. But on what grounds can it be asserted that K fails to refer? Let us formulate the standard two-dimensional analysis of meaning of a given term T in the following way. Extension of the term T is the set of actually existing entities to which the term applies. Intension of T is the set of features that constitutes the concept of T. Typically extension of a term is determined by its intension: to be a member of E an entity must have the features given in I. This analysis accords well with our actual linguistic practice. For instance, the term ‘‘dog’’ is used when and only when a set of certain features is present (namely, four legs, a tail, a fur, etc.). Similarly, the term ‘‘unicorn’’ fails to refer (has an empty extension), if there is no entity that would have the set of features that constitutes the term’s intension. So, according to this classical approach, there is a set of necessary and sufficient conditions that an entity has to satisfy in order to belong to the extension of a given term. According to another approach (inspired by Wittgenstein (1953)), there is no
123
Is knowledge a natural kind?
383
fixed set of necessary and sufficient conditions that the entities belonging to the extension of the term have to satisfy, but a more or less indeterminate cluster of such conditions. But it is common to both these approaches that the extension is determined by intension. (These approaches can be called descriptive.) A third approach is different in this respect. According to the causal theories of reference (Kripke 1972; Putnam 1975), the extension of a term is determined by a causal relation, which obtains between the term and its referent. So, if the reference does not go through the intension, the intension may in fact be revised by extension. In other words, we can be wrong about the intensions of our terms. It has been noted (cf. Lycan 1988; Stich 1996) that eliminative approaches seem to tacitly rely on descriptivist theories of reference. So the folk psychological terms, for instance, are to be eliminated since nothing neurophysiological in fact matches their intensions. But why not apply the causal theory to these cases? Why not conclude on the basis of this that the folk psychological terms actually refer to neurophysiological phenomena? Why not conclude that as science has moved on, we have come to know, not that there are not any beliefs, but that beliefs are neurophysiological states? That is, why not conclude that we were wrong, not about the extensions of the folk psychological terms, but about their intensions? The argument presented above, which led to the rejection of knowledge as natural kind, relied heavily on the assumption that truth is a necessary condition for knowledge. It was shown that ‘‘true belief’’ does not refer to any entity that would do the causal work that is required for it to figure in scientific explanations. But has it thus been shown that there is no such thing as knowledge? Why not conclude that there is a flaw in the intension of K rather than in its extension? (It should be stressed that on the basis of homeostatic property cluster view of natural kinds there is no theoretical reason why truth could not be dropped out of the cluster of properties that form the kind: there are no privileged, essential properties that all the members of the kind must share.) There is no obvious reply to these questions. One can only appeal to the fact that when the concept of knowledge is actually learned, it is immediately thought to imply the truth of what is known (cf. Matthews 1980). It is quite clear that knowledge implies the truth of what is known, and it could be pointed out that if one talks about knowledge without truth one has changed the subject. At the very least it should be demanded that if knowledge does not in fact have anything to do with truth, it should be rigorously proven rather than merely suggested. It is also clear that there is no reason in principle why science could not eliminate terms that fail to refer. After all, it did so with ‘‘phlogiston’’. Obviously, at times, it is natural to conclude that we have used a term that, on the basis of its intension, does not have an extension. It would be a really strange form of naturalism that would allow us to claim that there surely is such a thing as phlogiston, but we just do not know yet what it is. One could note that there is an obvious reason that explains why there has been no problem to eliminate ‘‘phlogiston’’. ‘‘Phlogiston’’ is a theoretical term that was introduced for a specific theoretical purpose. ‘‘Knowledge’’, on the other hand, is a term that is in a constant everyday use. This will no doubt influence our willingness to eliminate K.
123
384
T. K. Pernu
On the basis of this, let us finally formulate the following conjecture: The more vernacular a given term is the more likely it is that its reference will be determined by its extension at the expense of its intension. K is as vernacular a term as it gets, probably as old as humanity, and certainly not constructed for any specific theoretical purpose. If the above conjecture is valid, it is likely that we will keep looking for the extension of ‘‘knowledge’’ even though it would seem clear at the outset that no such thing could in principle be found. Perhaps one-day epistemologists will tell their students that in the past people used to think that knowledge had something to do with truth, but that it is now known that this is not the case.
7 Conclusion It has been argued that knowledge is causally non-efficacious; and that since causal homeostasis at the minimum is a necessary condition for being a natural kind, knowledge is not a natural kind; and that hence the term ‘‘knowledge’’ fails to refer and should consequently be eliminated from the language of science. What, then, could explain the conclusion that knowledge is not a natural kind? Why is ‘‘knowledge’’ in constant use in philosophy, science and everyday life if it is so obvious that it does not refer to a natural kind? It has been argued that the term does not belong to the object language of science. Nothing has been said about the metalanguage. Obviously, of these two, it would be natural to place ‘‘knowledge’’ to the latter. ‘‘Knowledge’’ is a judgemental term. Its purpose is to evaluate and prescribe our cognition, not to describe it. This would explain why ‘‘knowledge’’ still has an indispensable function even if it does not have a use in the object language of science. The question ‘‘What is knowledge?’’ could therefore still be a meaningful question. It is just a question that is of no interest to natural science and cognitive psychology. It is a question that interests those who are interested in evaluating our cognition—namely: philosophers. This, of course, is the way it has always been, before naturalistic philosophy took over and changed the subject. Acknowledgements I would like to thank Prof. Kristian Donner, MA Antti Kuusela, Dr. Markus Lammenranta, MA Elina Nurmi, Prof. William Rottschaefer, Prof. Gabriel Sandu, Dr. Kari Vepsa¨la¨inen and an anonymous referee of Philosophical Studies for helpful criticism, comments and discussions. This work was financially supported by the Department of Philosophy of the University of Helsinki and Emil Aaltonen foundation.
References Allen, C., & Bekoff, M. (1997). Species of mind: The philosophy and biology of cognitive ethology. Cambridge, MA: MIT Press. Bell, J. S. (1964). On the Einstein-Podolsky-Rosen paradox. Physics, 1, 195–200. Bell, J. S. (1966). On the problem of hidden variables in quantum mechanics. Reviews of Modern Physics, 38, 447–452.
123
Is knowledge a natural kind?
385
Block, N. (1986). Advertisement for a semantics for psychology. In P. French, T. Uehling Jr., & H. Wettstein (Eds.), Midwest studies in philosophy vol. 10: Philosophy of mind. Minneapolis: University of Minnesota Press. Boyd, R. (1988). How to be a moral realist. In G. S. McCord (Ed.), Moral realism. Ithaca, NY: Cornell University Press. Boyd, R. (1991). Realism, anti-foundationalism and the enthusiasm for natural kinds. Philosophical Studies, 61, 127–148. Boyd, R. (1999). Homeostasis, species and higher taxa. In R. Wilson (Ed.), Species: New interdisciplinary essays. Cambridge, MA: MIT Press. Brandon, R. N. (1990). Adaptation and environment. Princeton: Princeton University Press. Cummins, R. (1998). Reflection on reflective equilibrium. In M. DePaul & W. Ramsey (Eds.), Rethinking intuition: The psychology of intuition and its role in philosophical inquiry. Lanham: Rowman & Littlefield. Davidson, D. (1987). Knowing one’s own mind. Proceedings and Addresses of the American Philosophical Association, 60, 441–458. Descartes, R. (1637). Discours de la Me´thode pour Bien Conduire sa Raison & Chercher la Verite´ dans les Sciences, Leiden. Dretske, F. I. (1988). Explaining behavior: Reasons in a world of causes. Cambridge, MA: MIT Press. Dretske, F. I. (1995). Naturalizing the mind. Cambridge, MA: MIT Press. Egan, F. (1999). In defence of narrow mindedness. Mind and Language, 14, 177–194. Einstein, A., Podolsky, B., & Rosen, N. (1935). Can quantum-mechanical description of physical reality be considered complete? Physical Review, 47, 777–780. Ellis, B. D. (2001). Scientific essentialism. Cambridge: Cambridge University Press. Fodor, J. A. (1980). Methodological solipsism considered as a research strategy in cognitive psychology. In Representations: Philosophical essays on the foundations of cognitive science. Brighton: The Harvester Press. Fodor, J. A. (1987). Psychosemantics: The problem of meaning in the philosophy of mind. Cambridge, MA: MIT Press. Frisch, M. (2005). Inconsistency, asymmetry, and non-locality: A philosophical investigation of classical electrodynamics. Oxford: Oxford University Press. Garcia, J., McGowan, B. J., & Green, K. F. (1972). Biological constraints on conditioning. In A. H. Black & W. F. Prokasky (Eds.), Classical conditioning II: Current theory and research. New York: Appleton-Century Crofts. Gettier, E. (1963). Is justified true belief knowledge? Analysis, 23, 121–123. Goldman, A. I. (1986). Epistemology and cognition. Cambridge, MA: Harvard University Press. Griffiths, P. E. (1999). Squaring the circle: Natural kinds with historical essences. In R. Wilson (Ed.), Species: New interdisciplinary essays. Cambridge, MA: MIT Press. Jacob, P. (1997). What minds can do: Intentionality in a non-intentional world. Cambridge: Cambridge University Press. Keil, F. C. (1989). Concepts, kinds and cognitive development. Cambridge, MA: MIT Press. Kim, J. (1982). Psychophysical supervenience. Philosophical Studies, 41, 51–70. Kim, J. (1992). Multiple realization and the metaphysics of reduction. Philosophy and Phenomenological Research, 52, 1–26. Kim, J. (1998). Mind in a physical world: An essay on the mind-body problem and mental causation. Cambridge, MA: MIT Press. Kornblith, H. (1993). Inductive inference and its natural ground: An essay in naturalistic epistemology. Cambridge, MA: MIT Press. Kornblith, H. (1998). The role of intuition in philosophical inquiry: An account with no unnatural ingredients. In M. DePaul & W. Ramsey (Eds.), Rethinking intuition: The psychology of intuition and its role in philosophical inquiry. Lanham: Rowman & Littlefield. Kornblith, H. (1999a). In defense of naturalized epistemology. In J. Greco & E. Sosa (Eds.), A blackwell guide to epistemology. Malden, MA: Blackwell. Kornblith, H. (1999b). Knowledge in humans and other animals. Philosophical Perspectives, 13, 327–346. Kornblith, H. (2002). Knowledge and its place in nature. Oxford: Oxford University Press. Kripke, S. A. (1972). Naming and necessity. In D. Davidson & G. Harman (Eds.), Semantics of natural language. Dordrecht: D. Reidel.
123
386
T. K. Pernu
Lange, M. (2002). An introduction to the philosophy of physics: Locality, fields, energy, and mass. Oxford: Blackwell. Lycan, W. G. (1988). Judgement and justification. Cambridge: Cambridge University Press. Matthews, G. B. (1980). Philosophy and the young child. Cambridge, MA: Harvard University Press. Maudlin, T. (1994). Quantum non-locality and relativity: Metaphysical intimations of modern physics. Oxford: Blackwell. Mayr, E. (1961). Cause and effect in biology. Science, 131, 1501–1506. Mayr, E. (1982). The growth of biological thought: Diversity, evolution, and inheritance. Cambridge, MA: Harvard University Press. Millikan, R. G. (1984a). Naturalist reflections on knowledge. In White queen psychology and other essays for Alice. Cambridge, MA: MIT Press, 1993. Millikan, R. G. (1984b). Language, thought and other biological categories. Cambridge, MA: MIT Press. Millikan, R. G. (1989). Biosemantics. Journal of Philosophy, 86, 281–297. Millikan, R. G. (1993). White queen psychology and other essays for Alice. Cambridge, MA: MIT Press. Okasha, S. (2002). Darwinian metaphysics: Species and the question of essentialism. Synthese, 131, 191–213. Price, C. (2001). Functions in mind: A theory of intentional content. Oxford: Oxford University Press. Putnam, H. (1975). The meaning of meaning. In K. Gunderson (Ed.), Language, mind and knowledge. Minneapolis: University of Minnesota Press. Quine, W. V. O. (1960). Word and object. New York: The Technology Press of The Massachusetts Institute of Technology & John Wiley & Sons. Quine, W. V. O. (1961). Reference and modality. In From a logical point of view: 9 logico-philosophical essays (2nd ed.). Cambridge, MA: Harvard University Press. Quine, W. V. O. (1969a). Epistemology naturalized. In Ontological relativity and other essays. New York & London: Columbia University Press. Quine, W. V. O. (1969b). Natural kinds. In Ontological relativity and other essays. New York & London: Columbia University Press. Ristau, C. A. (1991). Aspects of the cognitive ethology of an injury-feigning birt, the piping plover. In M. Bekoff & D. Jamieson (Eds.), Readings in animal cognition. Cambridge, MA: MIT Press. Shoemaker, S. (1980). Causality and properties. In P. van Inwagen (Ed.), Time and cause. Dordrecht: D. Reidel. Sober, E. (1980). Evolution, population thinking and essentialism. In From a biological point of view: Essays in evolutionary philosophy. Cambridge: Cambridge University Press. Stein, E. (1996). Without good reason: The rationality debate in philosophy and cognitive science. Oxford: Oxford University Press. Stich, S. P. (1978). Autonomous psychology and the belief-desire thesis. The Monist, 61, 573–591. Stich, S. P. (1983). From folk psychology to cognitive science: A case against belief. Cambridge, MA: MIT Press. Stich, S. P. (1985). Could man be an irrational animal. In H. Kornblith (Ed.), Naturalizing epistemology (2nd ed.). Cambridge, MA: MIT Press, 1994. Stich, S. P. (1990). The fragmentation of reason: Preface to a pragmatic theory of cognitive evaluation. Cambridge, MA: MIT Press. Stich, S. P. (1996). Deconstructing the mind. New York: Oxford University Press. Wittgenstein, L. (1953). Philosophical investigations. Oxford: Basil Blackwell.
123
