K

Knowledge Set at the centre of a network of related notions like reality, meaning, understanding, reason and truth, knowledge is probably the most debated concept in the history of Western thought. With the rise of scientific psychology in the second half of the nineteenth century and of cognitive science another hundred years later, the notion of knowledge has become crucial to our comprehension of ourselves, insofar as the mind has been conceived of as a knowledge engine. There is no way of encapsulating the concept of knowledge in a brief entry, so we will limit ourselves to outlining a few roles that it plays in pragmatics and communication. We recommend that the reader consult the entries on cognitive science and representation and computation for further discussion and references. Pre-theoretically at least, the following

domains of knowledge can be kept distinct in the study of communication: (1) knowledge that each agent has about herself; (2) knowledge that each agent has about the interlocutor(s); (3) knowledge that each agent has about the process of communication; (4) knowledge that each agent has about the specific communicative exchange in which she is currently engaged; and (5) knowledge that each agent has about the world within which the exchange takes place. There can be overlaps between two or more of these domains. How each such domain is viewed and described depends on how one conceives of knowledge. Furthermore, since the cybernetic turn at the middle of the twentieth century (Wiener 1948; Ashby 1956; Bateson 1972) the

mind has been viewed as a knowledge engine: a system that dynamically generates and employs knowledge. Therefore, the question concerning the nature of knowledge is closely related to the question concerning the nature and the ‘structure’ of the mind and of its activities. There are basically two large perspectives concerning these issues, which we will call descriptionist and constructivist perspectives. Because we are accustomed to dealing with

knowledge as it appears to be hosted in sentences, drawings and other external, publicly accessible reifications of human thought, it may appear commonsensical to think that knowledge has an objective nature and that it exists before the subject and independently of it. The subject then may or may not have access to knowledge. In this perspective, knowledge is conceived of as reified into a (potentially infinite) set of statements whose meaning and truth are independent not only of the individual subject, but of any subject at all. Large bodies of mind and brain sciences have straightforwardly adopted this view to describe knowledge as a mental phenomenon. To the traditional locus of ‘accessible’ knowledge – the external world of books and encyclopaedias – these research communities have added a second, substantially identical locus placed in the realm of mental events and activities. Such events and activities, like the knowledge that they employ, are mostly or completely outside the scope of subjectivity. In this tradition, the mind is made up of an array of subpersonal components and processes that collect, store, retrieve and manipulate representations as pieces of knowledge. The functioning of the representational mind, that is, of the

knowledge engine, is independent of the subjective mind and impenetrable to it. The subjective mind may access some products of the functioning of the representational mind, possibly adding qualitative feels (qualia) to them, but its role is marginal, often to the point of theoretical or practical eliminativism. The unconscious in Freudian psychology, the computational machinery of cognitive science, or language processing in mainstream psycholinguistics are examples of mental functions and activities that are thought to be representational and knowledgeintensive, yet to fall outside the scope of subjectivity (Searle 1992 criticizes this view). Analogously, several conditions of neuropsychological interest are often explained in terms of the patient’s defective access or lack of access to relevant portions of her own knowledge base. The standard interpretation of anomia (nominal aphasia), for example, is in terms of the inability to access (part of) one’s own lexicon. Many theorists, however, do not accept this

view. An alternative is to regard mental knowledge (or all kinds of knowledge) as the subject’s construction rather than as an objectively given, albeit internalized, description which she can or cannot access. Constructivism is a broad array of theories characterized by the idea that knowledge is inseparable from the subject. In this perspective, knowledge is a property of the meaningful ways in which an agent interacts with the world and with herself. By definition, a meaningful interaction exists only in the first person, and the first person exists only in the here and now. Therefore, knowledge is actively and continuously constructed by the interacting mind. The representational mind (that is, the knowledge engine) and the subjective or phenomenal mind (that is, consciousness and experience) become one and the same thing. A consequence of this view is the need to distinguish between the knowledge that an agent’s mind actually represents at a certain moment – like her experience of walking and the relevant thoughts and bodily feelings – and the knowledge that is merely embedded in the way she acts, like the ‘knowledge’ about balancing that her body implements as she walks. While in the descriptionist approach outlined above the difference between the two is at most a question of ‘levels of representation’, here the difference is

qualitative. (Of course, the agent may theorize about bodily balancing. The knowledge employed for this purpose, however, is not directly the ‘implicit’ bodily knowledge, but a more or less detailed description thereof.) Intentionality, that is, the semantic (meaningful) relation between the mind and the world, is a crucial notion in this framework. The forms that intentionality takes in a certain agent depend on the agent’s biological nature and endowment (Johnson 1987; Maturana and Varela 1987; Tirassa et al. 2000) as well as on her previous interactions and on the activities in which she is currently engaged. This typically leads to the adoption of a methodology that is close to or includes phenomenology (MerleauPonty 1945; Varela 1996). Let us now turn our attention to knowledge as

it is employed in communication. First, each agent has to possess knowledge about herself. First-person experience and the awareness thereof are the basis of human agency: the reasons why an agent communicates are grounded in her beliefs, emotions, goals (both global and local), values, and so on. The standard jargon of psychology and cognitive science, typically cast in terms of elementary, subpersonal systems and microprocesses, has a hard time capturing the nature and functioning of human planning, action, and understanding. Relevance theory (Sperber and Wilson 1995) is probably the most renowned theory of pragmatics grounded in the standard jargon and general perspective of cognitive science. It may be interesting to note that, however satisfactory as a theory of language comprehension, relevance theory has never been able to become a theory of language generation or of dialogue, nor has it ever even tried to. Most other theories of communication since

Grice (1957) are cast in terms of mental states or, more precisely, in some variation of the socalled Belief-Desire-Intention (BDI) framework (Rao and Georgeff 1992). Such a framework appears to be well suited to the description of human activity, whether private, generally social or specifically communicative (Tirassa 1999). Mental states can be studied in terms of their intrinsic nature, their functional role in the mind’s internal dynamics and in its relations with the external world, their biological and evolutionary import, and so on. The use of a

BDI framework, however, is still neutral about the nature of the mind. In the descriptionist framework, mental states are abstract, objectively given states of a control system, and actions are behaviours or operators that are performed in an objectively given internal or external world (e.g. Fikes and Nilsson 1971; Newell and Simon 1972; Newell 1990; Cohen and Levesque 1990a, 1990b). In this view, it makes sense to say that a thermostat can perceive the external temperature, decide to turn on (or off) the heater and act upon such beliefs and intentions, issuing appropriate commands to the heater itself. This is viewed not as a metaphor, but as a simple, handy model of what goes on inside an animal’s or a human being’s mind. Instead, in the phenomenological or constructivist framework (Guidano 1987, 1991), beliefs, desires and intentions are phenomenal experiences that exist in the first person. The same view is then extended to the mental states involved in communication and generally in social cognition (Gallagher 2001; Gallagher and Hutto 2008; Tirassa and Bosco 2008). Second, if communication is conceived of in

terms of an agent affecting the mental states of another (Grice 1957), then knowledge of one’s interlocutor(s) is also crucial. An agent must have an idea of what her partner’s mental states are in order to affect them and to understand how (and even that) another agent is trying to affect hers. The ability to possess knowledge of others’ minds is the faculty commonly known as mindreading, or theory of mind (Premack and Woodruff 1978). The descriptionist view will typically conceive of this kind of knowledge as a theory proper that is made up of a list of facts and a set of reasoning rules concerning the minds of others (Fodor 1992). In this case, there is no particular reason to distinguish it from the knowledge that the agent has about herself, which also consists in a similar set of propositions and rules. The constructivist view will more likely conceive of mindreading in terms of intersubjectivity, and will probably be careful to keep it distinct from self-knowledge. Indeed, there are complex ongoing debates concerning the nature of mindreading (Whiten 1991; Baron-Cohen et al. 1993; Lewis and Mitchell 1994; Carruthers and Smith 1996) and its relation with self-knowledge (Goldman 1993;

Gopnik 1993; Nichols and Stich 2002; Bosco et al. 2009). These debates often involve developmental aspects, which are particularly interesting in the area of communication, because many empirical studies appear to show that human infants have substantially no theory of mind during the first 9-12 months of life. If this were true, it would be impossible for infants to engage in intentional communication with their caregivers (Risjord 1996) and, at least under certain perspectives, to even acquire language during this period (Bloom 2000). These considerations have been influential in the development of alternative accounts of the nature and the ontogeny of mindreading and communication (Airenti 1998, 2004; Tirassa et al. 2006a, 2006b). Third, agents must have knowledge about the

process of communication itself. This is a complex area which includes knowledge of language, knowledge of speech acts (including their locutionary, illocutionary and perlocutionary aspects), knowledge of the nature and the structure of conversation (including the allocation of turns, the repair system, etc.), knowledge of rhetoric and of the structure of dialogue and discourse, and knowledge of social customs and habits (including the nature and the management of ‘face’, politeness, and so on). Here, the general logic of the difference between descriptionist and constructivist approaches can be captured as follows. The descriptionist approach will view dialogue and conversation as characterized by an intrinsic, grammar-like or script-like structure that prescribes to a greater or lesser extent what move is appropriate to the ongoing exchange or should be played by each interlocutor at each time, as well as how the various structural and linguistic components of such moves relate to each other and to the overall script (Schegloff and Sacks 1973; Rumelhart 1975; Schank and Abelson 1977; Airenti et al. 1993a). In the constructivist perspective, the descriptionist approach appears to unduly expel subjectivity from the study of the mind’s activities and to commit the fallacy that Searle (1990a) calls ‘explanatory inversion’: the belief that regularities found in conversation (or generally in human activities) are caused or generated by the application of internal rules. Thus, constructivist approaches to communication (Winograd and Flores 1986; Tirassa and

Bosco 2008) will view the meaning and relevance of the communicative moves and of their components as founded in each interlocutor’s dynamics of mental states and in how such dynamics relate to each agent’s subjectively perceived situation. Fourth, communicators must have knowledge

of the dynamics, the current status and the contents of the specific communicative exchange in which they are engaged. This has to include at least some (however rough) knowledge of each participant’s current status and situation, knowledge of what has gone on explicitly and implicitly up to the present time, knowledge of what has been mentioned or referenced in the previous utterances, knowledge of the presuppositions and the implicatures of such utterances, knowledge of the material setting of the conversation (for example, it may be a faceto-face interaction or an interaction that occurs on the telephone or on email), and so on. According to some theories (Clark 1992, 1996; Airenti et al. 1993a), a larger or smaller part of this knowledge has a peculiarly public status. The idea is that communicative actions take place in a joint (mental) space which includes part or all of the above phenomena: such space is jointly set up and modified by the interactants and becomes their common knowledge. Some mental states of the interactants also belong to this public space; of course, if communicative actions are thought of as mental events the two things become one. Here, the descriptionist view will conceive of the public space as something that exists before and independently of the mental dynamics (and, of course, the subjectivity) of the interactants. This is typical, for example, of the ‘joint plan’ theories of dialogue in artificial intelligence and classical cognitive science (Cohen et al. 1990; Grosz and Kraus 1996, 1999). The constructivist view will conceive of the public space as a one-sided mental state which is unilaterally built by each interactant. Successful interaction depends on each interlocutor constructing a public space which is identical to, or at least compatible with, those of the others (Tirassa and Bosco 2008). Sometimes the whole notion of public space is interpreted as

a form of collective intentionality (Bratman 1992; Searle 1995; Tuomela 1995). Fifth, the last kind of knowledge integral to

communication is the general world knowledge that is needed to deal with the current exchange. This may include, for example, knowledge of the entities to which reference has been made, background knowledge, ‘encyclopaedic’ knowledge, and so on. The descriptionist view will conceive of this kind of knowledge in terms of semantic networks (Collins and Quillian 1969; Quillian 1968; Woods 1975), frames (Minsky 1974) and other representational codes. The constructivist perspective will view it in terms of an agent’s capacity to experience the current situation as related to past ones and to the overall situation in which she finds herself (Clancey 1997a, 1997b; Glenberg 1997). Further, researchers adopting the constructivist perspective will probably view lexical knowledge as part of this category. In this perspective, to know what salt is in understanding a sentence like ‘Pass the salt, please’ is not different from knowing what happened at Waterloo in understanding ‘This is a Waterloo’. Actually, the very idea that words, phrases and sentences have intrinsic meanings that pre-exist interpretation may be viewed as a token of the descriptionist idea that knowledge is objective and pre-exists the subjective mind.