Dretske offers a teleological view of how thought, belief, rationality, and intelligence can develop naturally in dumb (language-less) animals, and in some respect he does this as a response to philosophers like Davidson who want to limit us to saying that language is necessary for thought along with rationality. I argue that the safe part of Dretske’s argument is that a simple animal that has the capacity to learn to form representations of things in the world can be said to have beliefs and desires in a certain sense. I argue that this is a safe claim against Davidson since it is no problem to say that an animal has a belief so long as it behaves as if it has a belief. This will require that we spell out how it is that a simple animal could behave as if it had a belief, which entails that we spell out how it is that an animal could learn to form representations. At the same time, I argue that while having beliefs in this very limited sense is not sufficient for rationality, we cannot then say that rationality is necessary for having thoughts and beliefs. This will require that we draw out the distinction between rational/irrational and ‘routinized’ belief. McGeer and Pettit’s ideas about non-routinized minds are certainly helpful.
The idea, then, is that we can use McGeer and Pettit to distinguish proto-thought and belief from full-fledged rationality, thereby striking a balance between Dretske and Davidson. My thesis is that when we make this distinction we discover four interrelated truths: proto-thought and belief can arise in dumb animals (a part of Dretske is sound), proto-thought and belief does not require rationality (a part of Davidson is unsound), proto-thought and belief is not sufficient for rationality (a part of Dretske is unsound), and complex behavior in the form of what McGeer and Pettit call ‘content-attention’ is necessary for rationality (a part of Davidson is sound, and McGeer and Pettit help draw out this out). In proving these points, we will learn how Davidson over-extends his requirements for having thoughts, how Dretske over-extends his qualifications for having rationality, and how McGeer and Pettit offer clarificatory language that would keep each philosopher from over-stepping into the other’s territory.
Dretske’s thesis is that dumb animals learn to form beliefs when indicative sense-events assume an intentionally representational aspect. First, we describe indication. Indication is part of the non-mental world: tree-rings indicate the age of the tree, smoke indicates fire, and the clock indicates the time so long as it works correctly. (Dretske, 212). Indication is a law-like correlation between one piece of information and another. Dumb animals experience indication: a bit of sense-data reliably indicates a bit of information, although at this point the indication does not relay information, to the animal, as information. Here, indication is that animals’ detection of salient features of the outside world according to its phylogenetic propensities (Dretske, 236). When a generic dumb animal Buster sees a furry worm (furm), the furm will indicate to Buster—not that ‘a furm is there’—but rather a sense experience of what we would know is a furm. Because of Buster’s phylogeny, this sensation always happens reliably, and so the furm indicates, in a law-like manner, a specific sense experience in Buster.
Indication occurs in non-mental machines as well as in animals, and from all indication, even non-mental, intentionality arises as a consequence. A clock that is “on time” indicates that its gears are functioning up to speed with the official U.S. time in a certain area code, and it indicates this alone—while it may coincidentally be that it is 9 a.m. when the hands point to 9 a.m. on the face, still the working clock still only indicates that the position of the hands matches with the official U.S. time. This means that the clock indicates an intensional (with an ‘s’) bit of information: it is “opaque” that the clock indicates, or refers to, a correlation between the hand-position and the official U.S. time, and nothing else (Dretske, 211). Then, from the fact that the clock indicates a bit of intensional information, it follows conceptually that this non-mental clock is an intentional system. Because it indicates an opaque bit of information, it is set up to ‘choose’ what it indicates, and so in a sense it has intentionally with respect to what it indicates.
But strictly indicative intentionality is not sufficient for representation or belief; while intentionality is necessary for representation, strict indication is incompatible with representational abilities, and so it is incompatible with having beliefs. A clock does not ‘represent’ or ‘believe’ that its hand-position correlates with the time because it cannot intentionally represent meaningful content, as meaningful content must be detachable from its causes (Dretske, 213-4). The clock is only intentional about time “under and aspect” (Dretske, 213): it is locked into indicating the functioning of its gears, and even though this may be coextensional with the time of day, still the indication of the position of the clock’s hands cannot be detached from its cause (the gears). Until we detach what the hands indicate from the cause and have it represent the time, the position of the hands only intensionally indicates that the gears are keeping the clock up-to-speed. Thus, the same factor that makes the clock intentional—its strictly indicative intensionality—is what makes it non-representational.
Accordingly, a sense-event could become a meaningful representation if what it represents detaches from what it indicates, thereby losing its strictly indicative intensionality. We see this in Buster, who learns to represent a furm as something beyond what a furm experience indicates. Through experience, his furm-experiences become representations of furm-information; these experiences become eligible for representations by assuming an aspect beyond the mere sense-data that causes them. Thereby Buster distinguishes himself from strictly indicative intentional-systems like clocks. When this happens, we see that the distinction between non-mental intentional systems and ontogenetic animals is not only plausible but that it is essential for understanding how animals could have representational beliefs.
Here’s how Buster forms representational beliefs: Buster is an ontogenetic animal-system dependent on learning to do A (avoid) when FSW (something dangerous, i.e. a furry stingy worm) is present. He is phylogenetically established so that his detector-systems reliably indicate something that is actually dangerous to him, although they do not yet represent this information as dangerous information. After experience with a stingy furm causes pain and avoidance, Buster learns to exhibit avoidance because he experiences a furm, and so a natural learning process confers on the FSW-indicator the function of carrying information about FSW-ness (specifically the stingy S-part). What used to be an indication of furms—the actual sense-experience—becomes the representation of FSW—that there is furm present (Dretske, 223). Buster has developed a “capacity to reorganize control-circuits so as to exploit information in coordinating behavior”; he now responds to the world according to representational aspects of his experiences (ibid). Because furms no longer merely indicate an FSW-experience but, through learning, assume the job of indicating FWS-ness, therefore the indication is separated from its cause and takes on the representational meaning of “there is a furm” (Dretske, 238).
This has major implications for Buster’s status as a believer. Buster no longer responds to his furm-experiences qua indicative furm-experiences; now he responds to furm-experiences qua furm-representations. At this point it is undeniable that Buster displays “proto-thought” and “belief-like features” (Dretske, 222-3). Buster responds to information about experience and not just experiences themselves, and, according to Dretske, this is enough to say that Buster is a believer because it is clear that believing is seeing as (Dretske, 236). Buster’s behavior can only be explained by his having a kind of belief about furms: he avoids furms because he learns to see them as stingy, and not just because he sees them. We say he has beliefs because he behaves as if he has access to beliefs about the full-fledged information concerning a furm’s stinginess. We know this because he is not pre-programmed to see furms as stingy; he only learns this.
We also know that Buster has beliefs because he can be mistaken in his representational attitudes; thus it really is that a furm experience causes avoidance behavior because of the information it carries. For instance, Buster easily misrepresents E-worms as furms when he avoids E-worms after learning about the stinginess of furms. E-worms have done nothing wrong to him, so it must be that he avoids them because he learns to overextend his belief about furms with his belief about these E-worms that look a lot like furms. If an E-worm that closely resembles an FSW-furm shows up, Buster will likely avoid. Nothing inherent to the E-worm could have causes this behavior: it must be that Buster has used (false) information that he has learned about furms in order to respond to E-worms as furms. He responds to the E-worm based on the (proto) concept and belief that that E-worm “is a furm.”
We have established that Buster has proto-thoughts and beliefs, or at least undeniably “belief-like features” (Dretske, 223). Buster behaves according to the way the world is represented as being, and thus certain aspects of the world affect his behavior because they are represented in as having certain significant properties. Thus we have established two of our four truths: beliefs can arise in dumb animals, and rationality is not necessary for having beliefs. We rest assured that it is not “clear that a very complex pattern of behavior [language] must be observed to justify the attribution of a single thought” to dumb animals, as Davidson suggests (Davidson, 476). Now it is time to explore the implications. It seems intuitively very hard to swallow the idea that a dumb animal could have a belief. This intuition crops up because of our prejudice towards limiting belief to rational belief. Of course, Buster does not have that, but still, we have nailed down the type of belief he does have. What, then, is it about these proto-beliefs that make them, not rational or even irrational, but something else?
McGeer and Pettit would agree with Dretske so far as Dretske describes Buster as a “well-behaved intentional system” i.e. a system that acts in ways that fulfill desires in light of representational beliefs (McGeer and Pettit, 282). Buster has beliefs as a constraint-conforming organism: he “conforms to evidence-related and action-related constraints satisfactorily” (ibid). In our example, Buster adjusts to incoming furm-evidence faithfully, thereby forming beliefs that determine his satisfactory performance within his needs as an organism. (McGeer and Pettit, 283). He adjusts to the furm-indication so that it represents furry-stingy-worm presence, and this causes the appropriate behavior according to the constraint that he must avoid pain. It might then seem that this satisfactory behavior implies Buster’s rationality, and in fact Dretske makes this leap. There must, then, be a factor that we can nail down that distinguishes Buster’s satisfactory belief-behavior with more complicated rational behavior that he does not have.
One way we can clue in to the difference is to consider that now Buster cannot help experiencing furms (or E-worms) as FSW’s. In this sense, a new kind of strict intensionality, different from that of clocks, emerges: now furm and sufficiently furm-like sensations are strictly intensional with respect to representing furm presence. By becoming strictly intensionally representational, these sensations take on this aspect forever: they automatically play the dual role of creating responses to furm-evidence and directing avoidance action (McGeer and Pettit, 284). As McGeer and Pettit say, once Buster learns about his furm-experiences, his beliefs about furms “dictate how things are taken to be” for him (McGeer and Pettit, 285). What occurs in Buster when he encounters furms has become forever detached from furm-indications: when he sees furms he now only behaves according to certain meaningful information about them.
Buster is locked into representational furm-beliefs; consequently he is locked into having furm-beliefs about E-worms. Now Buster’s satisfactory behavior does not seem so rational: if people ran away from houseflies that resembled hornets, we would not say that they were acting rationally even though they were doing something that loosely satisfies their needs. Buster is locked in like this because the range of his intentional activity is limited to “the domain in which [he] is capable of exercising discrimination,” i.e. pain-avoidance and seeking comfort (McGeer and Pettit, 285). He routinely believes “there is a furm” because a pre-set design forces him to ‘put furm-information to work’ and avoid pain. This arises from the fact that Buster’s limited mind does not allow him form beliefs about beliefs. He can believe “there is a furm,” but he cannot “believe that there is a furm”: he is blind to the contents of his beliefs. While he does use furm-information as furm-belief, he cannot feel that he believes. Consequently, his furm beliefs, while they function qua beliefs, do not have a representational content beyond their function.
This means that Buster’s belief really is a primitive, atomic proto-thought: no content about “furms” per se come before his awareness. Furms cause avoidance strictly by crude representations of a stingy property mixed with crude furry and wormy representations. When furm-belief and furm-avoidance happens, there is no awareness of the contents of the belief; there is no awareness of the furry, the stingy, or the wormy qua F, S, or W. The reason is that there is no awareness in Buster of having beliefs, which awareness alone would allow him to internally reflect on his belief in order to distinguish its contents. There is no content-attention because Buster does not have a language that would enable him to attend to his belief-contents (ibid). Speech represents that things are represented as being a certain way (McGeer and Pettit, 286), and so the speaker can attend to the fact that things are being represented; automatically there is belief about belief, and so there is attention to belief-contents.
The practical implications of content-awareness are manifold. First, when we draw belief-contents within the scope of our attention, this makes it possible for us to break out of routine and form new beliefs and desires based on our awareness of belief-contents (ibid). We become non-routinized believers; we unlock ourselves from Buster’s crude way of responding to information. Also, when we attend to beliefs, we can believe that the contents of those beliefs are true or false, and by discerning truth-related properties in propositions, we learn to “identify constraints on what beliefs [we] should form and act on as well-behaved intentional systems” (McGeer and Pettit, 287). We can identify that we should make sure that we give assent only to true beliefs, and we thereby identify constraints on belief formation. From constraint-identification, humans can practice constraint implementation (McGeer and Pettit, 288). We can intentionally act in order to gain a richer exposure to constraints (education) and we can enhance responsiveness to constraints; we can conscientiously move beyond automatic belief by “intentionally monitoring and regulating inferential transitions,” and by taking “intentional initiatives to counteract resistance to reason” (McGeer and Pettit, 289). So, the non-routinized ability to attend to belief-contents qua belief-contents opens a deep behavioral well.
Now we can focus attention on Davidson’s claims to see how they apply to Buster’s case. Davidson says that in Buster-cases, we can continue to describe Buster’s behavior as belief behavior while recognizing that Buster does not actually have beliefs: in this case we are “applying a pattern of explanation that is far stronger than the observed behavior requires” (Davidson, 478). Davidson thinks this because he mistakes belief-in-general with belief-about-belief; he says explicitly that belief requires the concept of belief (Davidson, 476). For him it is clear that “semantic opacity distinguishes talk about belief” (Davidson, 475); but semantic opacity in this sense is not the kind of semantic opacity that we have seen works as proto-belief in Buster. For Davidson, semantic opacity means that a belief must be about something under some description (ibid). It is not enough that a furm opaquely represent “furm-presence” beyond mere furm-indication: Davidson requires that for a furm-event to be a furm-belief, Buster must believe of the furm some description of the furm. Davidson requires Buster to be able to attend to belief-contents in order that he can even think.
Clearly this is going too far for belief requirements, as we have shown that Buster really does have furm-beliefs. On the other hand, it is not going too far in terms of rationality requirements. For Buster to be able to respond rationally and intelligently to his furm belief, he clearly must be able to attend to the contents of that belief. Otherwise, he is only locked into routinely responding to furms. Buster is not a rational animal; he cannot form propositional attitudes about his beliefs because he cannot identify with them (Davidson, 475). Buster’s beliefs have no particular propositional content, and so he cannot constrain himself to acting only on belief-contents that he marks as ‘true’ (Davidson, 476). But again, we do not require rationality for belief or thought per se; it is not clear that belief requires the concept of belief.
If Dretske had stopped there, there might have been no ground for a Davidsonian attack. But Dretske over-compensates in his defense against Davidson by conflating proto-belief with rationality. From the successful argument that dumb animals can have thoughts, he moves to the controversial claim that “a system that acquires the power…to represent objects…will also automatically be an intelligent system…capable of believing in a rational way” (Davidson, 224, my emphasis). This immediately strikes us as a problem: given that Buster cannot even attend to his furm-beliefs, there is no way that he can have a rational furm-thought. The subtle problem responsible for this over-extension is that Dretske conflates ‘intelligence’ with ‘producing need-satisfying outputs’ (Dretske, 225). Dretske confuses rational belief with belief that merely ‘does its job,’ which is just to “produce intelligent (i.e. need-satisfying) output” (Dretske, 225, my emphasis). It is clear that just because Buster uses information to satisfy his needs, in no way does this automatically entail that he is behaving rationally. Need-satisfaction is not rationality because rationality requires, among other things, the ability to attend to the contents of his beliefs in order to do rational things with that attention. And also we saw that with Buster, mere need-satisfaction leads to the irrevocable misrepresentation of E-worms as furms.
Throughout this argument we have traced a solid defense against the skeptic who thinks that a dumb animal, because it is not rational, could not have any sort of belief. At the same time, we rebuke the idea that simple thought is sufficient for or automatically entails rationality; we untangle rationality and proto-belief so that we know exactly what we are talking about when we say that Buster could have crude belief. By drawing this distinction with the help of McGeer and Pettit, the claim that an animal could have crude beliefs becomes much more credible than Dretske would make it seem. We say that there are crucial aspects of rationality—namely content-attention—that a dumb animal does not possess. By holding this line, we respect the full power of rationality and do not confuse it with mere need-satisfaction. We prevent attacks from those who would claim that an animal has no thoughts because it is not rational. Since we are not conflating the two, we do not help out our opponents by overextending our claims.
posted by Michael Kleinman | 17.12.09
