Beliefs, Desires, and Actions

Consider actions in which the agent uses a piece of technology (in a broad sense) to achieve a desired goal: using a car to get to a certain place, using a cell phone to communicate, taking an umbrella out in the rain, using a hammer to knock in a nail. It is natural to analyze the psychological background of such actions as involving (a) a certain desire and (b) a belief about the means to achieving that desire. For instance, the agent wants to get to a certain place and believes that driving a car is the best means to achieve this goal. There is a combination of belief and desire at work, which leads to the performance of the action. Theorists of action have seen in this kind of case a general analysis of action: any action is caused and explained by the presence of an antecedent pair of a desire and a belief, where the belief specifies the means to be adopted to satisfy the desire. The desire alone will not prompt or predict the specific action performed—an accompanying belief is also required; nor will the belief alone suffice to elicit the action—an accompanying desire is required. A desire to stay dry will not prompt you to take an umbrella unless you believe that umbrellas keep you dry, and the belief that umbrellas keep you dry will not prompt you to take one unless you desire to stay dry (you might feel like getting soaked today). Thus there arises a certain doctrine about the psychological basis of action: any action requires the presence of a desire anda belief—both are necessary conditions (and together possibly sufficient) for an action to occur. This is supposed to be an interesting and unique fact about action and its explanation, suggesting a certain kind of “holism” about psychological explanation. Let’s call it the “Combination Thesis” (or “CT” for short): the thesis that all actions are consequences of a belief-desire pair acting in concert. Thus we arrive at something called “belief-desire psychology”, taken to be a general format for psychological explanation: any explanation of action requires the specification of a desire and an instrumental belief concerning how best to satisfy that desire. Action is the result of something cognitive and something appetitive operating together.

I propose to question this widely held picture. I think it arises from overgeneralizing one sort of example of action—the kind that involves means-end reasoning. In particular, I think that some actions occur without any instrumental belief and some occur without any desire (in a natural sense of “desire”). That is, there are one-factoractions—just a belief or a desire operating on its own. The Combination Thesis is false, except in a limited range of cases; it is not fundamental to action as such. Let us consider a very simple kind of action: flexing one’s index finger. Suppose an agent has a desire to flex his index finger and does so: what means-end belief accompanies this action? On the face of it, none: he didn’t believe that a certain device was a good means to bring about the desired result. The agent didn’t think that contracting certain muscles in his forearm would be a good means to flex his finger—he simply did it. And even if such a thought had entered his head (maybe he specializes in hand physiology), there would be the question of whether thataction needed a belief about means for it to be performed (say, initiating efferent nerve impulses from the brain). No, the action simply resulted from a desire, with no means-end belief involved (no technology is being exploited). It is the same for swallowing, breathing, walking, swimming, etc. You don’t believe that swallowing is a good way to get food into your stomach when you eat, or that putting one foot in front of the other is a good way to walk—any more than animals do. You just do it. It is absurdly intellectualist to suppose otherwise. It is not as if you have an array of options available for getting food inside you, such as inserting it through your navel, and you choose one of them as the best means in the circumstances. Chewing and swallowing are just nature’s way of getting fed, not a cunning plan you have devised for achieving satiation.[1]You have a desire for food, so you do what comes naturally, as you do when inhaling oxygen into your lungs. In general, movements of the body (often called “basic actions”) are not preceded by means-end reasoning: a person never moves her body by reasoning about the best means of doing so. That is, basic actions are not subject to CT. Raising my arm is not something I need an instrumental belief to perform. If I want to raise it, I raise it, without consulting possible means. Thus the explanation of a basic action has the form, “Xdid Abecause he desired to do A”. It is the same for adult humans, infant humans, and most animals: basic movement is caused by a desire not by a desire-belief combination. There is simply no need for belief: the mind and body are just wired (innately or by learning) to translate desire into action. The right thing to say is that where there is means-end reasoning there is a belief corresponding to the means that combines with a desire, but in more basic cases there is no such belief, since there are no such means. A specific feature one kind of action has been overgeneralized to apply to all action. Indeed, whenever CT does hold there is always a basic action for which it does not hold, i.e. a movement of the body. This is even clearer for purely mental action such as calculating in the head: if I add two numbers in my head, I don’t have an instrumental belief about how best to get the result I want. I want the answer and I’m wired to get it (I know arithmetic); I don’t need any superintending belief about the best means of getting it. Another way to put the point: in tactical practical reasoning agents have instrumental beliefs, but most action is not of a tactical kind. What psychologists call sensory-motor activity is generally not tactical or instrumental or belief-driven; it is automatic, programmed, not thought out.

But is desire always necessary for action? It depends what you mean by “desire”, which tends to be a philosopher’s term of art (sometimes glossed as “pro-attitude”). It seems right to say that the agent needs to view the action favorably (certainly not unfavorably), but there are ways of doing that that are not really cases of desire. Suppose you believe that you have a certain desire but you don’t really have it (you have been brainwashed into the belief): won’t you still be inclined to do the thing in question? Then your action will be motivated by a belief abouta desire, but not by that desire—you apply to medical school, say, because you have been brought up to believe that medicine is your calling (in fact, it’s opera). This would be a case in which you act on a belief without the corresponding desire, though you can be said to view the action favorably. And isn’t it generally true that desires influence actions only if they are recognizedin some way? How could desires of which you are completely ignorant figure as causes of action (even unconscious desires need to be recognized by the unconscious executor)? Their existence has to be registered or acknowledged. So something like belief has to be added to them to produce action; and then we have the question how much the belief contributes to motivational force. This gets pretty messy, psychologically. The neat picture of the pristine desire and its helpful belief companion starts to seem too simplistic. Motivation has (or can have) a more complex and variable structure.  You can do something simply because you have a sudden urge to do it, and you can also do something because you believe you desire to (though you don’t)—in either case you lack one of the components postulated by CT. So there are now three types of case to consider: (i) belief and desire in combination, (ii) desire alone, and (iii) belief alone. Some theorists have argued that moral motivation consists of nothing but a moral belief; we need not take a stand on that issue to accept that beliefs aboutdesire can play a motivational role. Being under the impression that you have a certain desire can act as a prompt to action, whether you have that desire or not. Maybe the only general thing we can say is that the action has to look desirable to you—there is something to be said in favor of doing it. This can take the form either of desire plus belief, or simply desire, or believed desire (or maybe just belief that the action would be morally good). Animal action will largely consist of the second category; agents with advanced practical reasoning will do a good deal of the first kind; the third kind will be restricted to those individuals deluded or confused about what it is they really want. There is no psychological structure common to all cases. The psychology of action is not monolithic.

Suppose someone suggests that action is what is caused by need, so that to explain an action we must specify what need it serves. This theorist is perhaps impressed by the actions of certain animal species of a somewhat primitive type. The natural response would be that this is too simple, too parochial: not every action is prompted by a biological need, and actions sometimes require practical reasoning involving instrumental belief. The need theory applies to some cases but certainly not to all. Well, belief-desire psychology, as currently understood, is rather like that: it fits some cases well enough, but it is too uniform and simple. There are a variety of different kinds of motivational state, ranging across a wide spectrum. In some ways the theory is too complex (because of basic actions) while in others it is not complex enough (because of cases like false beliefs about one’s desires). Thus reasons for action are of different types, not always resolving into the two-factor model of CT. Pluralism about reasons is the indicated position.[2]

Colin McGinn

[1]Sub-intentional actions, such as rolling one’s tongue around one’s mouth or tapping one’s foot nervously, seem particularly unsuitable to the belief-desire treatment: what instrumental belief do I have when my tongue is rolling around pointlessly? Just as the heart has no instrumental belief when performing the act of pumping blood, so many of our more automatic actions are free of cognitive supervision.

[2]All reasons may be causes, but the causes can vary as to type. Desires themselves can come in many types, from the moral to the animalistic. Nor is there less variety in the concept of belief. It is variety all the way down.

Subjectivity and Symbolism

Mechanism in physics was a unified theory of the physical world, positing only bodies in space and contact causation. Newton’s mechanics undermined this unity by postulating action at a distance, thus introducing another kind of causation. Electromagnetic theory undermined it further. Things don’t always operate by impact and reaction. In the case of psychology mechanism took the form of behaviorism, specifically stimulus-response psychology: the stimulus makes its impact and the response is the elicited reaction. The organism is just another body in space being pushed around by impinging forces. The reflex is the perfect exemplification of this theoretical framework: stimulus strikes body, body reacts (e.g. the patellar reflex). The discovery of conditioned reflexes held out the prospect of extending this basic mechanistic model to all behavior—psychology is thus the study of environmental impacts and the resulting motions of bodies. This too is a unified theory, employing a single conceptual apparatus to characterize everything the mind does and is. It is a unified theory that lasted much longer than the mechanism that was undermined by Newton and Clerk Maxwell. Mechanism about the mind persisted long after mechanism about the body had met its demise. Mechanism in physics replaced mysterious teleological conceptions of physical action (Aristotle), while mechanism in psychology replaced mysterious dualist conceptions of the mind (Descartes): but the former kind of mechanism succumbed to the mysteries inherent in Newton’s discoveries, while the latter kind soldiered on.

There were always rumblings against psychological mechanism, despite its dominance during much of the twentieth century; but the rumblings reached a crescendo during the latter part of the century. For convenience I am going to locate these dissenting voices in the persons of Thomas Nagel and Jerry Fodor; and I am going to summarize their contributions in two words—“subjectivity” and “symbolism”. By now this is a very familiar story, so I won’t spell things out. In brief: Nagel drew attention to consciousness, deploying the phrase “what it is like”, and insisting that there is more to the mind than physiology; Fodor likewise resisted physical reductionism and argued that the mind works symbolically, coining the phrase “the language of thought”, and suggesting that the mind is a computational system. Sensations are the paradigms of the subjective; thoughts are the paradigms of the symbolic. Some scuffles ensued regarding the scope of these basic categories, mainly having to do with whether everything symbolic is also subjective, and vice versa; but a consensus emerged that the mind has both sorts of property. It felta certain way and it representeda certain way. Thus people described conscious experience as consisting of qualia while thought processes were described as consisting of symbols: phenomenology and syntax, respectively. Two approaches to the mind thus came into prominence (one might cite Husserl and Chomsky as the father figures of these developments—or going back further, Brentano and Turing). One approach emphasized the first-person perspective, introspection, and lived experience; the other emphasized language (natural and formal), grammar, and symbolic processing.

The result of these innovations is not a unified theory; it is a disunified theory.[1]The general assumption is that the mind has two aspects, or consists of two sorts of faculty, or is made of two sorts of thing. True, it is subjective in some parts of its being; but also true, it is symbolic in other parts of its being. A theory of mind must therefore recognize a fundamental duality in what the mind is and how it operates. Metaphors have arisen to capture this duality: on the one hand, the stream or river of consciousness,[2]the mosaic of qualia, the theater of the mind; on the other hand, the mind as a computer, thought as inner speech, reasoning as calculating. It is not supposed that qualia are units of computation, and it is not supposed that words in the language of thought are qualitative contents of consciousness. The ontology is quite different, as is the role assigned to the entities postulated. One might say that subjectivity is an analogue phenomenon while symbolism is a digital phenomenon—continuous versus discrete, fluid versus segmented. The chief characteristics of mental symbolism are infinite productivity and syntactic concatenation, while the chief characteristics of mental subjectivity are not described in this way but are thought of a kind of flowing or pulsating (those metaphors!). Feelings don’t combine like words to generate an infinite array of syntactically structured strings, but neither do mental symbols afford a rich subjective life. These ontologies exist side by side, but they don’t interpenetrate—they don’t integrate. There is no unified theory of the subjective and the symbolic. [3]

Here we may be reminded of the current situation in physics, and indeed I think the comparison is apt. The theory of gravitation and the theory of elementary particles form different theories that are not unified or integrated. Indeed, the two theories operate with very different principles and laws. Yet the macro world and the micro world are not separate disjoint worlds; they overlap. What we have is a unified reality and a fragmentation of theory—not a happy state of affairs. There are some who detect actual tensions between the two theories, if not outright contradiction. Similarly, we have two theories of the mind, also not integrated or unified, and apparently about different things. For example, we can approach a given conscious thought in two ways: as a subjective state of consciousness imbued with a characteristic phenomenology, or as a symbolic structure functioning in a computational process (mostly unconscious). Yet we are looking at the same thing—just as a macroscopic object is the same thing as a congeries of microscopic objects. Surely there must be some way to bring these two descriptions together. But where is it to be found? Ideally we would be able to take a subjective description and derive from it a symbolic description, or vice versa—we would see these descriptions as aspects of a single reality. But there is a chasm between them, a dualism withinthe mind. What have qualia and symbols got to do with each other? Couldn’t you have one without the other? Is one more basic? What makes a symbol have a subjective character, and what makes qualia have symbolic properties? The two seem to stare at each other across a vast divide. It is not as if we have a computational theory of subjectivity or a phenomenological theory of symbolism—whatever either of those things might be.[4]The quantum world looks alien to the gravitational world and the subjective world looks alien to the symbolic world, but these worlds must be parts of a seamless whole in some way. In particular, an adequate theory in psychology would integrate the subjective and symbolic perspectives.

I am not saying it can’t be done; I am only saying that we don’t presently know how. The concepts are lacking; the theory is fugitive. At least mechanism avoided this kind of theoretical disunity. Starting with the idea of a reflex (innate and hardwired) behaviorism tried to generalize to cover all aspects of the mind, employing a single conceptual apparatus. That was a dismal failure—an absurd leap of faith. But the apparatus that replaced it is radically bifurcated and dubiously connected; we don’t even know how far the subjective and the symbolic overlap. It would be different if the two theories dealt with different components of the mind—say, subjective theory with sensations and symbolic theory with thought—but that is far from being the case, since sensations occur in perceptual and cognitive processes and thought is imbued with subjectivity. Subjectivity and symbolism exist in the mind in intimate and inextricable connection. So there really oughtto be a unifying theory, but we don’t have any idea of what it might look like. What would be nice is some explanation for how mental symbols are necessarily infused with subjectivity. Spoken symbols have the phenomenology associated with their sensory modality (mostly hearing), but symbols in the language of thought are not sensed in any way, so their phenomenological aspect must have some other source. Granted that the language of thought is innate, is it that the phenomenology of thought is coded into the genetic basis of its lexicon and syntax? The mind reels. Consciousness and computation are not separate aspects of the mind, existing is isolation from each other. A mind (a human mind) is a conscious symbolizer: it symbolizes in the mode of consciousness (as well as unconsciously). Its nature is subjective symbolism or symbolic subjectivity.

I have no suggestions to make about how to integrate these two aspects of mentality; I merely wish to point out the lacuna. As in physics, we live in an era of theoretical fragmentation with respect to the mind, following upon the heady unity promised by a general mechanism. Perhaps the future will bring the kind of theoretical unity that made mechanism so attractive to our ancestors, perhaps not.

Colin McGinn

[1]Functionalism undertook to demonstrate theoretical unity across everything that dwells in the mind, but it succumbed to an onslaught of objections, which I won’t repeat.

[2]Also consciousness as like a mysterious flame or a translucent rainbow or a type of veil or a genie drawn from a lamp or a shimmering force-field over the brain or steam from an engine—and no doubt others.

[3]It is worth noting that psychophysics has never been a symbolic theory, any more than reflexology has been. The law-governed dependence of sensation intensity on stimulus intensity is not a computational process—the sensory systems don’t deducesensation intensity from stimulus intensity. Rather, there are psychophysical reflexes underlying the laws of psychophysics. We have no tendency to invoke a “language of sensation” to explain the facts of psychophysics.

[4]No one writes papers entitled “The Logical Form of Consciousness” or “A Subjectivist Theory of Universal Grammar”. Those sound like category mistakes.

The Anti-Ontological Argument

The ontological argument proceeds from the premise that God contains all perfections to the conclusion that God exists. The anti-ontological argument proceeds from the premise that God contains all perfections to the conclusion that God does notexist. It thus precisely reverses the traditional argument deriving from Anselm.[1]Anselm argued that God could not be merely imaginary, because his definition as the most perfect conceivable being logically implies his existence, existence counting as a type of perfection. The anti-ontological argument contends that God mustbe merely imaginary, because his definition as the most perfect conceivable being logically implies his non-existence, since no absolutely perfect being can possibly exist (though such a being can certainly be imagined). This argument, like its traditional counterpart, operates with a very simple principle, which may be stated thus: For any kind K, nothing that exists could ever fall under the concept perfect K. Consider, for example, the kind hammerand form the concept perfect hammer: you have now gone from a concept with existent instances to a concept with no existent instances. Why? Because there can be no such thing as a perfect hammer, since every existent hammer will be imperfect in one respect or another. Only imaginaryhammers are perfect; real hammers always have flaws and drawbacks and weaknesses. Every hammer will fail to hit the nail on the head once in a while; every hammer will occasionally hit your finger instead; every hammer costs money to buy; every hammer gets rusty and decays; every hammer weighs something; every hammer takes several knocks to drive the nail home. There is no point in going into a hardware store and asking for the perfect hammer; there is no such thing. There are only more or less imperfect hammers. True, you can conceiveof a perfect hammer stipulated to have none of the defects listed, but there can’t be such a hammer. That’s not how the world works: there are always downsides and side effects and boundary conditions. It is the same for knives, motorcars, houses, musical instruments, and other artifacts. Ditto for organs of the body, and indeed whole organisms. Nor are persons any different: there is no perfect teacher or perfect policeman or perfect violinist. Everything is flawed in one way or another. Everything fails to live up to our imaginary ideals. Only non-existent objects lack any imperfection, because they are so stipulated. The real world always carries attributes that exceed and diverge from the ideal function of a type of object; thus no object functions ideally, as a matter of principle. Nothing is functionally perfect.

But God is defined as all perfect, absolutely perfect in every respect, without flaw or failing of any kind. The ancient Greek gods were not so defined, so there is no logical or metaphysical obstacle to their existence: but the god of monotheistic Christianity (and other religions) is decreed to be entirely without imperfection. Certainly we can imagine such a being, given the powers of the human imagination; but by our principle this being could not exist in the real world. The more we stipulate God’s supreme perfection the more we remove him from the realm of reality. In the real world God’s great powers and virtues would come with accompanying drawbacks, such as unintended side effects or large expenditures of energy or the exclusion of alternative desirable states of affairs. But we are told that none of this is true of God: when God acts there are no unfortunate correlates, no cons to the pros. His being and his actions are perfection through and through. He doesn’t even take up space thereby preventing other good things from existing! But this is simply defining him out of existence, like allowing that no actual terrestrial hammers are perfect while insisting that there are perfect hammers somewhere else in the universe (“transcendental hammers”). If someone were seriously to claim that, you would naturally ask what kind of hammer this could be: what material is it made from, how does it escape the laws of nature, how does it operate? In the case of God we are schooled not to ask these kinds of question, but the price is that we are merely confusing the imaginary with the real. The concept perfect beinghas no existent extension, where beingis taken to mean some sort of person-like entity. Thus that concept logically implies the non-existence of whatever falls under it (i.e. imaginary objects). If we take the concept of perfection to imply something functional, such as performing the office of a god perfectly, then no existent entity could ever be functionally perfect in the sense intended. In fact, when God is conceived as imperfect, as he sometimes is in the Old Testament, we have a clearer idea of what an existent entity of this kind might be like; but once we stipulate that there is nothing imperfect about him we enter the realm of the unreal and merely imaginary. Just as there are no perfect circles in the real world, only ideally, so there are no perfect gods in the real world, only ideally. The only circles that existare imperfect.

Here it might be objected that perfect circles can exist in Plato’s heaven—the same kind of place in which God is supposed to exist. But this is a confused thought, because such ideal entities do not act in the real world: they may be claimed to have genuine existence (as opposed to be being merely imaginary idealizations) but they don’t do anything to change the course of events. God does: he is supposed to be an active agent, a powerful force, a driver of change. That is, he is supposed to be as other active agents are—a type of (very superior) person, not an inert abstract form. God isn’t a piece of abstract geometry eternally at rest; he is capable of intervening in history. But then he must have whatever imperfections come with the territory: he can’t be both of the world and yet not of it. If he has a will, he has to have whatever imperfections come with that—anything else is merely imaginative stipulation. To deny this is like saying that a perfect knife has an ideal cutting edge without recognizing that no existent knife can have a cutting edge free from all imperfection. We can say the words but no real knife could live up to them (it would have to be unable to cut the flesh of its user for one thing). Why should a god be any different?

You might say that an existent God could have all moralperfections. That is not as easy as it sounds, but anyway it won’t preserve the traditional Anselmian notion of God, since it is compatible with accepting multiple imperfections of other kinds. That is, there might be a morally perfect being with bad teeth, a limp, a weakness for chocolate, poor taste in music, and a horrible dress sense. Such a being could exist without violating our principle, but he doesn’t add up to God. It is the requirement that God be perfect in every way and in every particular—absolutely and totally perfect—that puts him beyond the realm of actual existence. We just can’t comprehend what this would be—just as we can’t comprehend an actual perfect circle (one drawn with the intention to approximate the platonic ideal). If we define God in this way, we define him out of existence: we place conditions on his existence that can’t be realistically met. Suppose we say (with Spinoza) that God is composed of an infinite immaterial substance: that substance will exclude other substances like it, thus precluding a second all-perfect God from existing (itself a drawback); but it also raises the question of how such a substance might operate and what prevents it from malfunctioning. How could any substance exist and yet be incapable of failing to function as intended? How could it not at least contain the seeds of imperfection? To think otherwise is to lapse into a fairytale land subject only to the laws of imagination. It is to deal in metaphysical nonsense. We can appreciate this point easily for hammers, knives, and policemen, but in the case of God piety prevents us from applying our principle consistently. We tacitly concede this when we ignore the question of God’s aesthetic properties: is God perfectly beautiful too? No existent thing is ever perfectly beautiful—the very idea is a fantasy—so how can God be perfectly beautiful? What does that even mean? Is he superlatively handsome? Does he have a lovely form than which no lovelier form can be conceived? Is his beard finer than the finest silk? None of this makes sense—so how does God possess all aesthetic perfections? The anti-ontological argument contends that any actual being with aesthetic qualities will have aesthetic imperfections or limitations; aesthetic perfection obtains only in the realm of the ideal or imaginary. The right thing to say is that we can conceive of an all-perfect being (or at least we can say those words) but that no such thing could exist in reality. Thus the definition of God as an all-perfect being logically implies that God is not real—just like the definition of an all-perfect knife (call it “Excalibur”) implies its unreality. If we knew there to exist gods that are less than perfect, we would accept no counter-example to our principle; suggesting that there is a different type of god that is perfect in every way would naturally arouse our suspicions, for it would violate our general conception of reality. We know that reality is less than ideal, and we know that we can conceive of things that are ideal; so we naturally reject the idea of the ideally real. We find the conjunction of the attributes of complete perfection and real existence to be contrary to reason.[2]The concept of absolute perfection is hyperbolic and fails to characterize the real world. It is this concept that is deployed in defining God as Anselm does in the ontological argument: but that definition asks too much of any actually existent entity. Thus the Anselmian definition of God, so far from entailing God’s existence, logically precludes it.

[1]In what follows I don’t attempt to say where Anselm’s argument goes wrong; instead I offer another argument with the opposite conclusion. If this argument is sound, we know that Anselm’s argument hasto go wrong somewhere. For the record, I think he is wrong to take existence as a type of perfection.

[2]I hope no one will protest that perfect numbers exist (a positive integer that is equal to the sum of its proper positive divisors): that is not the notion of perfection at issue. And of course there is no objection to the loose use of “perfect” in conversational contexts.