[2] And yet it has become something of an obligation for books in cognitive science to contain at least a passing reference to Heidegger. For example, consider this fairly typical comment concerning connectionism in Bechtel and Abrahamsen's Connectionism and the Mind: "Thus it provides hope of situating cognitive processing in the world, and so begins to elucidate what Heidegger may have had in mind when he emphasized that our cognitive system exists enmeshed in the world in which we do things, where we have skills and social practices that facilitate our interaction with objects" (Bechtel and Abrahamsen 1991: 126).
[3] This particular quotation points to the explanation for this interest in Heidegger among cognitive scientists. The passage indirectly suggests that Heidegger is relevant to cognitive science insofar as he puts forward a set of constraints which any cognitive system must satisfy if it is to count as similar to our own. Bechtel and Abrahamsen correctly attribute to Heidegger the position that there are three essential features of our cognitive system and its relation to its environment. First, Heidegger is said to have emphasized that our cognitive system "exists enmeshed in the world." Second, it is asserted that Heidegger claimed that the "world" is to be thought of as the field in which we do things, presumably as opposed to primarily thinking of it as a set of objects. Third, it is claimed that Heidegger thought that our doings, our interactions with objects within this world, is facilitated by "skills and social practices." Now, if what Heidegger says about us and our relation to the world is true, then these three essential facts about us amount to a set of conditions on having a cognitive system similar to ours and any attempt to model that system must reproduce those features. So, Heidegger is at least relevant to cognitive science insofar as he puts forward a set of constraints which must be satisfied by any system which could count as thinking in the same sense we think.
[4] We owe a debt to Hubert Dreyfus for pointing out this potential relevance of Heidegger to cognitive science. In a long series of publications beginning with What Computers Can't Do, Dreyfus has insisted that Heidegger's work has profound implications for cognitive science in general and for the pursuit of artificial intelligence in particular. According to Dreyfus,these implications begin with the requirements that any thinking entity must "be-in-the-world," that "the world" in which we are is the context in which significant action can take place, rather than a set of decontextualized objects, and that our primary way of being-in-the-world is through skillfully coping with it in accordance with a variety of social practices. The terminology of "skills," "coping," and "social practices" is Dreyfus', not Heidegger's (although the positions are certainly to be found in Heidegger himself). So Dreyfus' interpretation of Heidegger is the proximate source of the constraints on our cognitive system listed by Bechtel and Abrahamsen.
[5] But Dreyfus doesn't stop with this list of conditions on being similar to us. Instead, he goes on to argue that if we understand these conditions properly we will see that it is at least highly unlikely, if not entirely impossible, for a digital computer, which manipulates formal symbols in accordance with a set of algorithms organized in a program, to ever satisfy those Heideggerean constraints, and thus count as thinking.
[6] In this paper I will concentrate on this further inference which Dreyfus draws from Heidegger's work. I will try to show that while Dreyfus is quite correct in thinking that Heidegger provides a set of conditions on what it is to think, and that if Heidegger is right about those conditions no program which provides a set of rules for manipulating formal symbols could specify what it is to satisfy those conditions, he is wrong to think that there is anything in Heidegger which gives us reason to think that these conditions could or could not in fact be satisfied by some digital computer running some program. That is, I will argue that if Heidegger is right, we have reason to believe that no computer program could tell us what it is to think, but that there is nothing in Heidegger which counts against the possibility that some computer may actually think.
[7] So it is central to my thesis that we distinguish two issues. The first issue concerns what Heidegger would call a "question of being": "What is it to be a thinker?" The second issue concerns a matter of fact: "Which actual entities might count as thinkers?" We can better see that and how these issues are distinct if we distinguish clearly between two questions concerning computers and thought.
[9] Presumably what motivates this possibility is the recognition that some machines running in accordance with such programs for the manipulation of formal symbols, suitably connected to their environments through program entry and exit rules, have, in restricted and well defined areas, proved successful at behaving in ways which seem to require rationality. This success in modeling rationality locally has suggested the possibility of modeling rationality globally. And if one can program a machine to be rational, this suggests that what it is that the machine is doing which qualifies it as rational is specified by that program. Finally, insofar as rationality is considered at least necessary, and perhaps sufficient, for thought, this in turn suggests that if we can describe what constitutes a rational entity we would thereby describe what constitutes a thinking entity.
[10] One can also take the question of whether the mind is a program in an entirely different direction. In this sense, the question concerns what actual digital computers, organized so as to follow a set of formal algorithms, might be capable of doing. Whereas the first question asks concerning what is meant in saying that something thinks, and raises the possibility that "thinking" might be defined in terms of programs which, when followed, cause rational behavior, this second question raises the issue of whether a computer operating according to such a program could actually count as thinking. Now, if we conclude in answer to the first question that to think just is to operate according to some program, then, of course, any machine which does so operate also thinks. On the other hand, if we conclude in answer to the first question that what it is to think can not be captured in any program, this leaves the second question open. In order to answer it we would need to know at least two things. How ought we to understand what it is to think? And, is it possible for a thing which acts in accordance with some program also independently to satisfy whatever necessary constraints there are on thinking? The answer to the question of whether a computer could actually think thus turns on the answer to the ontological question. For this reason, Heidegger would have called it an "ontic" question, a question about a being rather than about being.(2)
[11] Dreyfus seems to think that Heidegger's work has important consequences for both of these issues, and that if Heidegger is right it is not only impossible to define thought in terms of programmed behavior but also impossible that any computer could actually think. And, as I will show, if Heidegger is right regarding "being-in-the-world," "skills" and "social practices," then this does indeed have important negative implications for the possibility that thought can be identified with behavior in accordance with a set of operations defined over formal symbols. As I will also show, however, Heidegger's work does leave open the possibility that some digital computer might actually think. But to see this we must have some understanding of the question which Heidegger himself was trying to answer.
[13] Heidegger rarely discussed "thought" or "cognition," and "consciousness" rarely appears as an issue in his early writing. Instead, he focuses on intentionality. Heidegger follows Brentano and Husserl in taking intentionality as the mark of the mental sine qua non. So whatever Heidegger has to say which is relevant to questions concerning thought and computers is going to be relevant to issues concerning intentionality. And, since the possible intentionality of a state is centrally tied to the possible semantics of that state, Heidegger's thought will be most directly relevant to the semantics of mental states.
[14] In fact, intentionality was one of the two topics which centrally concerned the early Heidegger. The other topic was being. From Heidegger's perspective, these two topics are necessarily related, and he approached this relationship from both sides.
[15] On the one hand, Heidegger felt that any investigation of "being" demanded that one first have some comprehension of how any understanding of being was at all possible. Heidegger distinguishes at least three "questions of being": (1) What is that which is?; (2) What do we understand by the word "being"?; (3) How is it possible to have any understanding of being? The first question is ambiguous between the population question, what kinds of entities are there, and the question of what the criteria are by which we should answer the population question. The second question concerns the meaning of "being," and as Heidegger sees it, this question is more fundamental than the first in that any answer to either of the forms of the first question presupposes an answer to the second. But, Heidegger thinks, even the question of the meaning of being (in this sense) rests upon a yet more fundamental question, the third "question of being," "How is any understanding of being possible?" That is, if one is to understand what it is for something to be, one must first understand what it is to intend something as something which is, and how it is possible to intend something as something which is. And if one is to understand what is meant by "being" in general, one must first understand what it is and how it is possible to intend "being" as such and in general. But, as intentionality in general is marked by the fact that any intention is directed towards something which is taken as something which is, either as mere object of an intention or, in the more usual case, as transcending that act itself, one understands what it is to intend "being" just in case one understands intentionality as such.
[16] Heidegger's conviction that one can only investigate what is meant by "being" by investigating the character of the intention directed towards being is rooted in an assumption which Heidegger inherited from Kant by way of Husserl: the necessary conditions on the objects of an experience or an intention are rooted in the necessary conditions on that experience or intention itself. The validity of that assumption is not directly at issue here. What is relevant, however, is that from Heidegger's perspective, any investigation of being must first articulate the nature and possibility of intentionality, and so this assumption drives Heidegger from the "question of being" to an articulation of the character of intentionality.
[17] On the other hand, Heidegger felt that there was a basic gap in Husserl's articulation of intentionality. Heidegger thought that Husserl focused exclusively on the essential characteristics of various sorts of intentional acts and their objects. How is perception as such distinguishable from imagination, for example, or what is it that distinguishes the object of a perceptual experience from the object of an imaginative experience? But from Heidegger's perspective this left out the crucial issue of what one is saying when one says that some thing is actually intending some other thing. "This, then, is the result of our deliberations: in [Husserl's] elaborating intentionality as the thematic field of phenomenology, the question of the being of the intentional is left undiscussed" (Heidegger 1985: 113). Another way to put "the question of the being of the intentional" is to ask "What are the necessary conditions on the possibility of intentionality?" A more contemporary way to put this question is to ask, "What conditions must be satisfied when some event is correctly described in intentional terms?"
[18] This question, however, seems to have a ready answer. Ever since Brentano resurrected the medieval term "intentional," it has been common to define "intentional" in such a way that a state is intentional just in case it is about or directed towards something. And Heidegger is certainly aware that Husserl has this definition available to him, and Heidegger gives us no reason to think that he himself doesn't accept the definition. So Heidegger's question concerning the being of the intentional can't just be a request for a definition of "intentional." And, in fact, it isn't. Rather, Heidegger is asking whether or not attributions to a state of intentionality in the usual sense necessarily imply that that state also has other features. In particular, Heidegger is interested in whether for a state to be intentional it must be a state of some specific type of entity, and if so, what is the most general way to characterize that type of entity. Heidegger's name for entities which can support intentional states is "Dasein." And, for Heidegger, to ask after the most general features of a type of entity is to ask after the meaning of its being. So, as he makes explicit in Being and Time, Heidegger's basic question regarding the being of the intentional should be understood as the question of the meaning of the being of Dasein.
[19] Heidegger takes the question of the meaning of the being of Dasein in a radical direction. On the face of it, to ask what it means for a class of entities to be is just to ask for the properties which an individual must have in order to belong to that class. And if membership in that class is deemed to be necessary for that individual to be the individual it is, then these properties are essential to the individual. But Heidegger has little interest in the issue of Dasein's essential properties. After all, he thinks that the tradition has already identified what those properties are: the intentional ones. Instead, Heidegger is struck by the fact that intentional attributions come in packages. It seems to Heidegger that if some entity has one intentional state, then it has an entire set of them, and that some entities just don't seem to be the right sort of thing to have any intentional states. So Heidegger wants to know what it is which characterizes an entity as being a candidate for having intentional states. How must an entity be, if it is to have intentions? To put the question in modern dress: How are we describing something when we describe it in such a way that it has the full set of intentional states? What must an entity be like if we are to be able to successfully take an intentional stance towards it?
[20] There are some good reasons to think that if some state of some entity is intentional, then other states of that entity are also intentional. In typical cases, particular intentional attributions come in packages of attributions to individual entities. Jane can't believe that the gun is loaded, unless she also believes many other things, and a case can be made that Jane also can't hope to avoid getting shot, or desire to leave the room, or act in an attempt to leave the room unless she has many beliefs.
[21] In asking regarding the meaning of the being of Dasein, Heidegger is basing himself on these facts. His orientation towards intentionality is formed in and through attention to the systematic character of intentional attributions. Unsurprisingly, this orientation predisposes Heidegger in favor of some variety of holism regarding the intentional. To understand what is involved in an intentional state one must understand the being of the entity which has intentional states, and recognize why and how it is that such an entity has a multiplicity of such states. For Heidegger, to understand what it is for a state to be intentional one must understand the basic structure of Dasein. And to understand the basic structure of Dasein is to comprehend what must be globally true of the entities which have intentions. This strategy distinguishes Heidegger's approach to issues regarding intentionality. Instead of focusing on the intentional character of individual states or events, he focuses on the distinctive character of the subjects of intentional states.
[23] As Heidegger articulates it, being-in-the-world is quite complex. But for Heidegger it is at least the case that any entity which has being-in-the-world as its mode of being displays two fundamental features. First, it acts purposefully in a goal-directed fashion. Second, in pursuing its goals any Dasein uses tools as tools, where to use a tool as a tool is to use it as it should be used because it should be used in that way. As how a tool should be used is determined by a context of habitual social practice, to say that being-in-the-world essentially involves tool use implies that an entity can be in the world only if its goal directed activity is norm governed in the specific sense that in general it does what it is appropriate for it to do socially, rather than what is appropriate merely given what it wants to achieve. My discussion of being-in-the-world in this section focuses on each of these features of Dasein in turn.
[24] Heidegger tells us that the basic structure of Dasein is "care." In Being and Time care is articulated as: "...ahead-of-itself-Being-already-in-(the-world) as being-alongside (entities encountered within-the-world)" (Heidegger 1962: 237). Although I will make no effort to explicate the "care structure," there are two aspects of Heidegger's discussion of care which are relevant here. First, Heidegger insists that only entities with the care structure can count as having intentions. "It could be shown from the phenomenon of care as the basic structure of Dasein that what phenomenology took to be intentionality and how it took it is fragmentary, a phenomenon regarded merely from the outside. But what is meant by intentionality - the bare and isolated directing-itself-towards - must still be set back into the unified basic structure of being-ahead-of-itself-in-already-being-involved-in" (Heidegger 1985: 302-303). Second, Heidegger's specification of "care," as "ahead-of-itself," etc., is an attempt to articulate the essential character of what Heidegger refers to as "the basic state of Dasein": being-in-the-world. So the basic state of beings with intentions is named "being-in-the-world," and the essence of this way of being is care; "...being-in-the-world is essentially care..." (Heidegger 1962: 237).
[25] "Being-in-the-world" is, of course, a compound expression, and while Heidegger insists that it is a "unitary phenomenon," he himself doesn't hesitate to reflect on its various aspects in a number of different ways. Terminologically, "being-in-the-world" is composed of "being-in" and "world." So it makes some textual sense to approach Heidegger's being-in-the-world by way of "being-in" and "world."
[26] Heidegger approaches "being-in" in several ways, but for our purposes the most illuminating manner of approach is by way of an early lecture course, The History of the Concept of Time. There he gives a list of "possible modes of in-being belonging to everydayness": "working on something with something, cultivating and caring for something, putting something to use, employing something for something, holding something in trust, giving up, letting something get lost, interrogating, discussing, accomplishing, exploring, considering, determining something" (Heidegger 1985: 159). Heidegger characterizes the common denominator of these modes of being-in as "concern," but there is another aspect of this list which jumps up at you. All of these states involve overt behavior of embodied persons, overt behavior described in intentional terms. So if Heidegger thinks that only entities which are being-in-the-world can have intentional states, and something is in the world just in case it is being-in, then the fact that Heidegger takes overt intentional performances as paradigmatic of "modes of being-in" gives a clear indication that Heidegger means to assert that only entities which act intentionally can have any intentional states at all.
[27] Heidegger is certainly not unique in thinking that overt behavior can count as intentional. Almost everyone thinks that. After all, whether one takes the mark of intentionality to be the informal "being directed towards" or the more formal linguistic criterion of being described using verbs which create apparently nonextensional contexts, intentional actions pass with flying colors. What is more significant is that Heidegger also takes a stand on the question of what it is in virtue of which an act has intentional content at all. On many views, the intentionality of overt acts is derivative from the intentionality of the states which explain or cause them. This position is held by a very diverse group of thinkers. John Searle and Jerry Fodor, who share little else in common, agree that the intentional content of overt acts is derivative from the intentionality of the internal states which cause them, even if they disagree on the character of those internal states. Heidegger, on the other hand, rejects this view. For Heidegger, to understand intentionality itself, in all of its forms, we must "set it back into the unified basic structure" of being-in-the-world. That is, rather than the intentionality of internal states being seen as explaining the intentionality of acts, Heidegger attempts to understand all intentionality in terms of being-in. And the paradigm cases of the modes of being-in are goal directed acts. So the intentionality of overt action is taken as primitive relative to the intentionality of internal states, rather than the reverse.
[28] Thus, we can reach this conclusion regarding Heidegger's views concerning the entity which can have intentions: given that acting intentionally is paradigmatic of the modes of being-in, and being-in is essential to being-in-the-world, and being-in-the-world is the basic structure of entities which can have intentions, to say that only entities which have being-in-the-world as their mode of being can have intentions is to say at least that acting intentionally is necessary to having any intentional states whatsoever.
[29] Heidegger's privileging of overt intentional acts as the paradigm cases of intentionality suggests to him that it would be inappropriate to describe those acts and their intentionality in terms derived from a model which emphasizes the intentionality of the states which cause or explains those acts. So Heidegger never treats the directionality of goal directed acts as resolvable into the intentionality of beliefs and desires. He thus resists the temptation to say of an act of hammering that what makes it an act of hammering is that the agent has a desire to make a nail fast and believes that moving the hammer in just this way will result in making the nail fast. Instead, he starts with a description of the acts themselves as goal directed, and asks what other intentions are implied in such descriptions.
[30] The paradigmatic modes of being-in, working on something or producing something, for instance, involve in addition to the goal of the act an apparent intention directed towards an entity as a certain type of tool. For example, if it is correct to describe a certain event as a hammering (that is, as directed towards the goal of a nail being made fast), it is also correct to say that the agent is treating the object which is being used as a hammer, and the entity which it is used on as a nail. For Heidegger, the primary way in which one intends hammers as hammers is by hammering, and every time one hammers one intends something as a hammer. "Equipment can genuinely show itself only in dealings cut to its own measure (hammering with a hammer, for example)..." (Heidegger 1962: 98). So it is appropriate to say that some agent is engaging in a goal directed act of production, say, only if it is also appropriate to say that she is intending some tool as a tool. But, then, since Heidegger thinks that engaging in practical activity is a necessary condition on a subject having any intentions at all, and an agent engages in practical activity only if she intends tools as tools by using them, then the necessary conditions on tool use are at the same time necessary conditions on intentionality as such. But what are the necessary conditions on intending tools as tools?
[31] The best way to come to grips with this question is by way of the other aspect of Heidegger's "being-in-the-world," the world. Heidegger thinks that the intentional actions which characterize Dasein as being-in always presuppose a certain sort of context. "Being-in" is always "being-in-the-world." How are we to interpret this talk about the world? The crucial clues come from Heidegger's specification of what we understand when we understand the world. When we understand world we understand a functionality context. "We always already understand world in holding ourselves in a contexture of functionality." A "contexture of functionality" is a structure of relations which specifies various jobs which objects can be used for, and which types objects in terms of how they are to be used. What we intend an entity as when we intend it as a tool is something which is for a certain use, which is "in order to" accomplish a certain end. But this equipmental role, this specific "in order to" which characterizes the type to which the tool belongs, is itself specifiable only in terms of its relations with other equipmental roles. Hammers are for hammering, but hammering is for using nails to secure boards, so that the frame of, e.g., a house can be made strong, etc. Heidegger's holism reveals itself in his claim that since each equipmental role is defined only in terms of its relations with others, the holistic context of such relations has a priority over each of the individual roles. This broadest context of functions, of what is to be done with different things, makes up the functionality contexture which is understood when we understand the world.
Each individual piece of equipment is by its nature equipment-for - for traveling, for writing, for flying. Each one has its immanent reference to that for which it is what it is. It is always something for, pointing to a for-which. The specific structure of equipment is constituted by a contexture of the what for, in order to. Each particular equipmental thing has as such a specific reference to another particular thing. Every entity that we uncover as equipment has with it a specific functionality. The contexture of the what for or in order to is a whole of functionality relations.... The functionality whole... is the prius, within which specific beings, as beings of this or that character, are as they are and exhibit themselves correspondingly. (Heidegger 1982: 163-164)[32] The widest functionality contexture is the world as "that for which one lets entities be encountered." That is, one can intend tools as tools only if one intends the entity which is a tool as to be used ("in-order-to") in a certain way, and to intend an entity as to be used in some way is to implicitly place it within the context of in order to relations which comprises the world. What it is to intend something as a hammer is to intend it as to be used to hammer nails, but one can do that only if one intends nails as a certain sort of tool which is to be used to make boards fast, and boards are tools.... So, for Heidegger, every actual tool using activity, as such, is only possible insofar as the Dasein which acts is "in the world," engaging in productive activity against a background of implicitly defined functional roles. Each particular act makes use of some object and in doing so treats it as to be used to attain some particular end. But one can intend an object as to be used in a definite way only insofar as one implicitly intends the other objects with which and on which the tool is to be used as themselves to be used in various ways. Every actual intentional tool use also involves an implicit understanding of the functional context in terms of which the tool used is the sort of tool it is. In Heidegger's terms, every occasion of being-in is being-in the world.
[33] Given Heidegger's emphasis on activity, on "in order to"s and "for the sake of"s there is an odd fact about Heidegger's philosophy of action. The odd fact is that Heidegger doesn't have any philosophy of action. Instead, he has a discussion of tool use. The attentive reader will have noticed that there was a certain slippage in the discussion several pages back. I began my articulation of Heidegger's concept of being-in by discussing the claim that only agents which act intentionally can have any intentions. But Heidegger's paradigmatic modes of being-in all involve tool use, and his notion of the world specifies a context of functional roles which define tools as tools. But surely not every goal directed act is a tool using act. One would think that if anything counts as a goal directed activity it would be a tiger hunting its prey, but tigers typically don't use tools in doing so. So it seems that Heidegger is committed to a stronger thesis than the one I attributed to him. For Heidegger it is not merely the case that to have intentions an entity must be a subject of goal directed activity, it is also the case that the entity must be a subject of tool using activity. But what could possibly motivate such a claim?
[34] Heidegger believes that full fledged intentionality demands a certain sort of normativity. For Heidegger, an agent counts as having intentions only if she engages in tool use, and to engage in tool use is, by and large, to use the tools as she should use tools of that sort in realizing the goals of her tool using acts. But to use tools as they should be used displays a different sort of normativity than that displayed in non-tool using goal directed activity. That Heidegger thinks that acting in light of tool using norms is necessary for intentionality is probably a function of thinking that this kind of normativity is necessary for language use, and thinking that language is necessary for genuine intentionality.(3) To understand Heidegger's views on how tool using action is essential to intentionality as such it is best to compare his account of how action involves a normative dimension with Donald Davidson's position (Davidson 1980, 1984).
[35] For Davidson, an event is an act if it is part of a pattern of events which are coherent in the sense that they are integrated into a system in which, by and large, the agent does what she should do, given her reasons for action. Those reasons, in turn, are understood in terms of the agent's beliefs and desires. An individual act is rational, in the narrowest sense, if it is an act which the agent should perform given the belief and desire of the agent which explain that act. So it is rational for me to flip the switch if I believe that doing so will turn on the light and I want the light to be on. Conversely, the content of a given state of the agent is partially fixed by the normative role of that state in the system of the agent's acts. To oversimplify, for a state to be the belief that flipping the switch will turn on the light is for it to be a state such that, given that one is in that state, if one also had the desire to turn on the light, then one should flip the switch. (One must add to this the principle of charity that the agent with this state mostly does what she should.) These relations suggest that one interpret the goal of an act to be the realization of a state of affairs which satisfies the desire which partially rationalizes the act. I believe that flipping the switch will turn on the light, and I want that light to be on, so I flip the switch. I should flip the switch insofar as the flipping is an act which is toward the light being on. Thus the content of the desire which is ingredient in the explanation of the act is covariant with the direction or goal of the act itself. In turn, the content of the belief which is relevant to an act's explanation would then be the state of affairs, which, if it were actual, would make it rational for the agent to do as she does, given her desire.
[36] The normativity involved in this sort of instrumental rationality ultimately derives from the ends which are specified by the desire of the agent and the means which are sufficient to realize those ends under a variety of circumstances. I should do that which in the actual circumstances would satisfy my desire. And, given the minimal conditions on rationality embodied in the charity principle, I most often do that. But not always. One of the jobs of the concept of belief is to help fill in this gap. Often when I do what I should not, given my ends and the actual situation, I do what I should given my ends and my beliefs. So the content of my belief specifies the state of affairs in which I should have done what I in fact did do in order to realize my ends. So, for Davidson entities have intentions only if they have beliefs and desires and they have beliefs and desires only if we can interpret what they do teleologically, that is, as attempting to realize goals. And such entities both act teleologically and have intentions only if in general they act as they should, where the norms which determine what should be done are fixed by the desires of the agent and the actual situation.
[37] Action which involves equipment displays a characteristic type of normativity which is different in kind from that displayed by non-equipmental action. The only sense in which I should not do or believe something which is instrumental to one of my ends is that realizing that end might be incompatible with some further end of my own. There is a sense, however, in which a tool can be used incorrectly, even when it is used successfully to achieve some end of an agent. A hammer can be used successfully as a paperweight, but insofar as it is indeed a hammer, to use it in this way is to fail to use it as it should be used. It is to miss the fact that it is a hammer and not a paper weight; or, as Heidegger would say, it is to fail to understand the being of the hammer. So insofar as human action involves using tools as tools, that is, as to be used in certain characteristic ways, human action involves a kind of normativity which is absent from the non-tool using animal kingdom, and also absent from Davidson's understanding of the normative dimension of intentions.
[38] For Heidegger, we are in-the-world only if we use tools as tools. But what is it to use something as a tool? One might think that to use e.g., a hammer as a hammer is to act out of a desire to make a nail fast and a belief that one can make it fast if one moves this object in just this hammering way. But Heidegger thinks that this is too simple. Tools aren't merely things which can be used so as to achieve ends in certain circumstances. In fact, defective tools can be tools even if they are not capable of being used to achieve their characteristic ends. And objects which are capable of being so used need not be that sort of tool. Rather, tools are things which are to be used in certain ways, or should be used in those ways. The status of being a tool is a normative status, as Robert Brandom would say (Brandom 1994). To intend something as a hammer is to intend it as an entity which it is correct to use in certain situations, with certain other types of tools to achieve certain functionally described types of ends. And the conditions which specify when it is appropriate to use a tool of a certain type, and thereby define that tool type, are themselves normatively specified conditions. To intend a hammer as a hammer is to intend this entity as to be used as a hammer is to be used, and thus as normatively related to a whole host of other normatively characterized entity types and normatively characterized possible end and initial situations.
[39] This "to be used" character of equipment can't be derived from the ends of the agent. Indeed, this normative character of tools is precisely rooted in the distinction between the ends for which the tool can be used and the ends for which it is to be used. Heidegger thinks that language use demands using words as they are to be used, and so that language use displays this same form of normativity as tool use. And, insofar as language is necessary for intentionality, Heidegger thinks that being-in-the-world, that is, using tools as tools, is necessary for all intentional agents, and thus for all intentional states.
[40] But what is it that fixes the normative roles which define the tool types which are characteristic of the functional contexture in which agents which are in-the-world must act? The specific character of the world, the interlocking normative structures which specify how things are to be used, can't depend upon the ends of the individual agent's who actually use the tools. Rather, Heidegger thinks, these roles are fixed socially: I ought to act as "one" is to act, where the "one" is a distinctive social norm.
[41] Thus, for Heidegger, an entity can have intentions only if it is in-the-world, and it is in-the-world only if it uses equipment skillfully as it should be used according to prevailing social practices. But these claims are based on a fundamental decision regarding the priority of the intentionality of goal directed action over the intentionality of internal "mental" representations. And, it is legitimate to ask, "What motivates this decision?" In the next section I turn to this issue.
[43] Phenomenological approaches to the mind arise out of the fact that three distinct characteristics have traditionally been lumped together as "mental." First, mental states are intentional. Second, our individual intentional states, together with the relations among them, rationalize our actions. Third, it has traditionally been thought that all mental states are conscious. Now, it has been traditional to see these three "mental" regions as coextensive. And, since Descartes, it has also been frequently assumed that consciousness is incorrigible. Given this incorrigibility, the identification of the mental with the conscious suggested that it was possible to authoritatively investigate the intentional and rationalizing aspects of the mental by giving a description of those aspects from the first person standpoint of the agent's own conscious life. And, in that case, the phenomenological method is the obvious choice for an investigation of the necessary conditions on intentionality.
[44] Unfortunately for the claims of phenomenology, the identification of the mental with the conscious is no longer quite so obvious as it once seemed. Such a conflation now appears to beg several important questions. In particular, it now seems quite possible that I can have intentional states which influence and rationalize my behavior, but of which I am not aware. Equally, it is now quite apparent that many of those activities which were traditionally explained by appeal to mental states, such as playing chess, can be "performed" adequately by entities which follow rules for the manipulation of formally characterized symbols but which quite clearly have no states resembling our conscious ones. And it seems to many that a syntactic approach which treats the mind as a machine running a set of algorithms on formal symbols, and those symbols as related to the world by a set of representation relations, is at least a possible source of illumination on the mental. But if either of these is possible, then it could not be the case that all and only conscious states were intentional or rationalizing. And in that case, there would be no reason to think that the best we could do from a phenomenological first person perspective is the best we could do in providing an account of either rationality or intentionality.
[45] So if the only reasons Heidegger has for his account of intentionality are phenomenological, then he gives no reasons which need be considered seriously by anyone who denies that all intentional states must be conscious, or who even denies that the necessary conditions on intentionality must be available to consciousness. Fortunately, Heidegger does offer other than phenomenological grounds for his views regarding intentionality. Those grounds arise out of a reflection on the character of intentionality itself, rather than from any considerations arising out of or grounded in our consciousness of our own intentional states.
[46] From Heidegger's perspective the key fact about intentionality is that it is "transcendent." My intentional states are about entities which transcend both those states themselves and the fact that they are intended. That is, when I see a chair or believe that there is a chair in the corner it is the chair itself which I am intending, not any hypothetical mental object which only is if it is being intended. "When I look, I am not intent upon seeing a representation of something, but the chair" (Heidegger 1985: 35). "In conformity with its sense of direction, perception is directed toward a being which is extant" (Heidegger 1982: 63).
[47] Now this fact concerning the transcendence of our intentions gives us no guidance regarding how we are to understand how such transcendence is possible. And it looks as if we do need some guidance on this issue. The problem is that, on the one hand, intentional states appear to be relational. The verbs which establish intentional contexts all take objects which specify what the state described is about. But, on the other, the entities mentioned in the object clauses need not exist. I can hear voices when there is no one there and I can believe that unicorns are mammals even though there are no unicorns. So the relation which characterizes intentionality can't be a real relation between the intender and the object which is intended, as real relations presuppose the existence of both relata.
[48] These facts lead to a puzzle. Intentional states are typed and identified in part by the objects towards which they are directed. But the intentional character of these states can't be understood as a relation between these states and the actual object towards which they are directed. So how are we to understand the relational character of the intentional without committing ourselves to the existence of entities which need not exist?
[49] One classic response to this puzzle is to postulate a third entity, distinct from both the event of intention and the object which is intended, which is supposed to mediate between the subject and the object intended. Such entities allow one to distribute the problem. On one side, intentionality involves a real relation between two existing entities, the subject and the mediating "third." On the other side, the mediating entity is thought to represent, stand for, or mean some possible state of affairs or entity. And, since that state of affairs or entity which is represented is merely possible, it is possible for us to intend it even though it does not obtain or exist.
[50] Heidegger is a straightforward anti-representationalist. Aside from phenomenological doubts, the basic problem for Heidegger with all varieties of these mediating entities is that they just don't do the job for which they are designed. All of the heavy lifting is done by the representation relation. But this relation is at this point left entirely opaque. Instead of saying that I can see a chair which is not there we are taught to say that I intend an item which is there but which represents a possible chair which might not be there. But what are the identity conditions for possible chairs, as Quine might ask? And what relation, natural or unnatural, is the representation relation with possible chairs, as Heidegger might ask?
[51] So Heidegger rejects the representational theory of intentionality. Whether he is right to do so is another issue, of course. But at least until someone has actually succeeded in telling us exactly what being a representation consists in, it is rational to search for another account of intentionality. And this is precisely what Heidegger does.
[52] Formally, Heidegger's solution to the puzzle of intentionality is to take the "directedness towards" which is the defining property of the intentional as intentional as an intrinsic, non-relational property of the intentional comportment itself. Thought is open to or related to its object, transcends itself towards its object, in virtue of its own essential character as directed-towards, so the thought needn't involve a real relation with its object. This solution to the puzzle raises as many problems as it solves, however. It is simply impossible for the properties of classically conceived substances to involve relations to something else in themselves. The properties of substances are either intrinsic, in which case they characterize the substance apart from any relations that substance has to any other entity, or the properties are relational, in which case the property implies the existence of the other relatum in the relation.
[53] So, Heidegger concludes, intentional states just can't be states of substances. Rather, they must be states of a different sort of entity, Dasein. Now, we have seen that Dasein are essentially in the world, and to be in the world is to be an agent who uses tools as they are to be used because they are to be used in that way. So presumably Heidegger thinks that characterizing the subjects of intentional states as being-in-the-world helps to solve the puzzle of intentionality. And so he does.
[54] Here is how it is supposed to work. Remember the paradigmatic modes of being-in are "working on something, producing something, putting something to use," etc. All of these states involve overt behavior of embodied persons, overt behavior which is described in intentional or goal directed terms. And, from the standpoint of the intentionality puzzle, what is striking is that each of these states demand some relation between an agent and her environment, but need not involve a relation between the agent and the object which is mentioned in the characterization of the act itself. So, given what it is to be acting so as to produce a widget, it is possible, in virtue of real concrete relations between an agent and some entities in her environment, to be acting in order to achieve that end even though there are no widgets, never have been and never will be. Heidegger's strategy, then, is to solve the puzzle of intentionality by showing that because what it is to be acting so as to achieve a possible state of affairs is defined in terms of a manner of working on and with those things which do exist, in the case of intentional action one can intend a state of affairs that is not achieved by working on and with other things which really do exist. He then hopes to understand all other intentional states through their relations, explanatory and otherwise, with goal directed action.
[55] Consider the case of attempting to produce a house. There are two ways to understand what is meant when it is said that someone is attempting an act of house production. One interpretation is that when one produces a house one has a representation of a house present to the mind and that this representation of the house partially causes the acts of the agent. On this view, an act is an act of production iff it is caused in the right way by house representations. Alternatively, one can think that a certain behavior is directed in the way it is in virtue of its relations with other behavior and the natural and social environment in which it takes place. On this view, a series of motions can count as an attempt to build a house, iff they are coordinated in the right way with each other, the environment in which they occur, their effects on the environment, and the socially constituted norms for appropriately using tools to build houses. So, while there must be latitude given for the possibility of failure, for what I am currently doing to count as hammering in the cause of house production it must in general be done with certain functionally characterized types of things, on certain functionally characterized types of things, in order to accomplish certain socially characterized types of ends. And all of these can be correct descriptions of my behavior only if my overt behavior has a certain characteristic structure and a definite set of relations with the norms of my society.
[56] Given Heidegger's anti-representationalism he rejects the representational alternative as non-explanatory. Instead, he adopts the second view of productive and intentional action. But the goal directedness of action is, then, a case of how an agent's relationship with her environment can imply an intentional relation between the agent and a merely possible end. Since, for Heidegger, this is the only case in which an intentional relation can be made comprehensible in terms of real, albeit teleological and normative, relations, all intentional states are to be understood in terms of their relations with the purposeful activity of social agents. One is Dasein (has intentions) only if one acts, one acts only if one is in-the-world, and one is in-the-world only if one acts skillfully according to the practices of the society in which one lives.
[57] This, then, is the outline of the considerations which Heidegger advances supporting his views concerning the necessary conditions on intentional subjects and intentional states. Any account of intentionality must give us some understanding of the semantic character of intentional states. No semantics, no intentionality. The prime semantic fact is that a state can be about or directed towards something which is not or does not obtain, so intentions can't be understood as real relations between a subject and the objects specified in the description of the intentional state. Representational views of intentionality remain non-starters until they can give us an account of the representation relation, and at present they give us no such account. We do, however, have a case of an intentional state for which we can begin to understand how an intention can be directed towards a possible state of affairs even if that state of affairs does not obtain. In goal directed action, what it is for the action to be directed towards the possible goal which is specified in the description of the act can be understood in terms of relations between that act and other real entities, as long as those relations are characterized in normative and/or teleological terms. Since this is the only case in which we can understand the oddities of the semantic dimension of intentionality, we should understand the semantics of all intentions in terms of their relations with the semantics of "skillful coping." So the necessary conditions on the possibility of describing an agent as skillfully coping with her environment while following social norms, whatever those conditions might be, are at the same time the necessary conditions on that agent having any intentions whatsoever. That is, nothing can think unless it is being-in-the-world.
[59] Given the above description of Heidegger's work it should be clear that whatever relevance Heidegger has to the answer to these questions turns on identifying "thinking" with "having intentions," where to "have intentions" does not imply a primitive intentionality of internal states, but merely that an entity has states with intentional descriptions. Heidegger is primarily concerned with the being of the intentional, and being-in-the-world and skillful coping in accordance with social practices are conditions on any being having intentions. So I will take this identification of thought and intentionality for granted.
[60] Heidegger's conclusions regarding being-in-the-world, skills and practices, and intentionality essentially amount to an attempt to specify what it is for something to be like us in having intentions. As Heidegger himself is at pains to make clear, his questions concern the meaning of the being of Dasein, or what is presupposed about the subject when one makes intentional attributions. So one way to look at Heidegger's results would be to take them as a set of specifications which must be satisfied by any entity which could count as having intentions.(4)
[61] Clearly, then, Heidegger's work is most directly relevant to the first question, the question of being. And the relevance of Heidegger's work to that question is equally clear. Since for Heidegger what it is to have intentions is specified by being-in-the-world, and one is in-the-world only if one acts in accordance with social norms using tools as they should be used according to social norms for socially prescribed ends, one can't define what it is to have intentions in terms of behavior according to a set of algorithms defined over formal symbols. The problem is that such a set of algorithms prescribe what a system will do. But what it is to have intentions is defined in terms of what a system should do. For Heidegger, something is in-the-world only if, in general, it acts as it should act because it should act in that way. But there is no way to capture the content of a normative system which specifies the norms governing how an agent should act in terms of a set of rules which specifies how an agent will act. And, since it is a necessary condition for Dasein having intentions that it in general acts appropriately given some set of norms, one can't specify what it is that Dasein is doing which qualifies it as having intentions by saying that it is following any formal algorithm.
[62] One way this comes out is that, for Heidegger, an entity can count as having intentions only if it acts teleologically for ends. And one can be acting for end E even if what one does does not result in E, and one can not be acting in order to realize E even if what one does results in E (Okrent 1991). So the condition that an entity must act for ends if it is to count as having intentions can't be captured by specifying that it act according to the rules in any system of actual inputs and outputs. Further, for Heidegger, only agents who use tools appropriately count as having intentions in the full sense. And to use tools appropriately involves using them as they should be used, where this norm is defined in terms of a set of holistic relations among similarly normatively defined types of equipment and situations. To use A as it should be used is, in part, to use it with B in situation C in order to achieve D, in which one should use E with F to achieve ..., where C and D are defined in terms of what should be done in those situations and B, E, and F are defined only in terms of each other and the situations they should be used in and for. And none of these "shoulds" or "in order to"s is translatable into a "will."
[63] So, if Heidegger is right there is no way in which the mind can be a program in the sense that the program defines what it is to have a mind. What is interesting about this is that the deep structure of Heidegger's argument is similar with the structure of Searle's Chinese room argument, but the content is entirely different (Searle 1980). Searle argues, one will remember, that programs are syntactic, minds have a semantics, and syntax is insufficient for semantics; so programs are not minds. But Searle's reason for thinking that syntax is insufficient for a semantics is that he thinks that consciousness is necessary for semantic content, and that syntax is insufficient for consciousness. Heidegger, on the other hand, waives the consciousness requirement. Instead he argues, based on his discussion of the intentionality puzzle, that acting practically for ends is necessary for semantic content, and syntax is insufficient for acting for ends.
[64] There is a crucial difference between the structure of Heidegger's argument and the structure of Searle's, however. Since for Searle the semantics of thought is dependent on conscious states, only entities with conscious states could count as thinking, no matter how they behave. But for Heidegger the semantics of thought depends upon being-in-the-world, and being-in-the-world crucially involves a style of action. So it looks as if it might be possible for something to qualify as thinking in virtue of the style of its action alone. I turn to this possibility next.
[65] Everything which I have said so far concerning Heidegger's relevance to cognitive science is in complete accord with Dreyfus' claims regarding Heidegger, although we read Heidegger quite differently. If Heidegger is right, then what it is to have intentions can't be captured in any set of formally defined rules. This result, it seems to me, is crucial, and, given Heidegger's primary interest in ontology, I am certain it is the result which Heidegger would have insisted upon. It also seems to me, however, that this ontological conclusion leaves the ontic question open. Regardless of what Dreyfus (or Heidegger himself, for that matter) might think, it doesn't follow from the fact (if it is a fact) that what it is to think can't be defined in terms of a program, that entities which behave according to a program could not also count as thinking. One can see that this is the case if one pursues the implications of a dialogue which Dreyfus initiated with the ghost of Alan Turing.
[66] In the article which can justly be said to have inaugurated the program of strong AI, Turing was willing to admit that "it is not possible to produce a set of rules purporting to describe what a man should do in every conceivable set of circumstances" (Turing 1950: 441). What Heidegger adds to this discussion is the assertion that what it is to have a mind must be understood in terms of acting in accordance with some set of rules describing what should be done in every conceivable set of circumstances (where those circumstance types are also defined normatively). And, if this is the case, then it follows straightforwardly from the premise which Turing accepts that no set of formal rules could specify what it is to have a mind. The only question which remains at issue then is "Could a machine which is programmed to behave in accordance with a set of formal rules also count as, in general, doing what it should, where what it should do is determined by a set of social practices?" That is, could such a machine also count as having intentions?
[67] Turing denied that it followed from the premise that he is willing to accept, that what a person should do can not be expressed in any set of rules, that there were no "laws of behavior," that is, "laws of nature as applied to a man's body." And, we might add, Turing was surely right in this denial. Dreyfus, however, is also correct when he responds that it doesn't follow from the fact that the behavior of human agents follow causal laws that these laws could be embodied in a computer program (Dreyfus 1979: 192). So it doesn't follow from the fact that everything in nature, including things which have intentions, fall under causal laws, that a digital computer could count as having intentions. But, it might fairly be objected, it also surely doesn't follow that these causal laws could not be embodied in a computer program. And if these causal laws could be embodied in such a program, then a machine could actually do everything that a person does in the same situations, physically described. And, it might be argued, in virtue of that fact, if the person counts as a Dasein, so would the machine.
[68] Well, given Heidegger's views, why couldn't this be the case? I can think of only two possible reasons to think that Heidegger's results preclude the possibility that some symbol processing computer might possibly count as having intentions. And it seems to me that Dreyfus has at times appealed to both of these possible grounds. On the one hand, one can think that Heidegger shows that some form of consciousness is required for intentionality, and that Heidegger has shown that machines following formal rules can't possibly be conscious. Unfortunately, while these considerations would in fact weigh against the possibility of a programmed machine having genuine intentions, as far as I can see Heidegger neither argues that consciousness is necessary for intentionality (although his Husserlian upbringing might predispose him in that direction) nor that formal syntactical engines couldn't be conscious. So there are no distinctively Heideggerean grounds for arguing in this way.
[69] On the other hand, one might try to appeal to the Heideggerean premise that the relations which specify what an agent must be if it is to have intentions can not be expressed in terms of a set of formal rules which could be embodied in a computer program. In that case one would attack the step in the above argument from "machine and human do the same things in the same situations, physically described" to "machine and human both count as having intentions." If, as Heidegger thinks, to have intentions one must in general do what one should do because one should do it, and what one should do can not be expressed in any set of formal rules, then one can't argue from "A and B do the same things, physically described" to "A and B both count as having intentions." One can't infer this because, if the premises are true, there can be no lawful relation between "doing the same thing physically described" and "doing the same thing intentionally described," so there is no way to infer from the fact that some machine does the same things in the same physical situations as a Dasein that they have the same intentions. The relations which are physically relevant need not be identical with the relations which are intentionally relevant, so the inference is blocked from physical similarity to intentional similarity.
[70] This, it seems to me, is quite a good argument. In fact, it reproduces in slightly altered form an argument which Davidson developed to show that, because there are systematically different types of criteria used to attribute physical and psychological states, there can be no psycho-physical laws (Davidson 1980). The conclusion to this argument, however, is that, even if it turned out to be possible to program actual human behavior patterns, physically described, this would be insufficient to guarantee that the entity running this program would also count as having intentions. Intuitively, this is right. If Heidegger is correct in thinking of intentionality as requiring a relation between an agent and a social context of practices, then if one alters the social environment of an entity, one also might be altering whether or not that entity had intentions, even if the physical description of the entity's behavior remains the same. But this still doesn't show that a digital computer could not have intentions.
[71] Precisely insofar as the two descriptions, "acting purposefully as one should given a set of social practices" and "acting in accordance with a set of rules for manipulating formal symbols," are logically independent of one another, the behavior of some agent might satisfy both descriptions. The only way we could ever find out whether we could build such an entity (or, maybe, even be one) is to try to build one and see (or to try to come up with a set of rules which adequately captures our own behavior, which, under another description, is also skillfully coping with our environment, by acting in accordance with social practices).
[72] Now, given the character of the necessary conditions which Heidegger places on intentionality, with the emphasis on social normativity and goal directedness, it may seem to one that it is unlikely that any entity could both be a digital computer and a Dasein. But a Heideggerean can't legitimately go further in her response to the ontic question. Even if what qualifies a thinker as a thinker is not that it behaves as some computer would, it does not follow that some computer could not also be a thinker.
[73] This result, however, is entirely in accord with the spirit of Heidegger's work. From a Heideggerean perspective, the important issues are all ontological. The only ontological question in this area is the question of the being of the intentional. And Heidegger's answer to this question is incompatible with the hypothesis that the mind is a computer in the sense that what it is to be a mind could be expressed in some program. This is the important result. Whether or not some computer could also count as thinking is, it seems to me, a much less interesting question.