A cognitive account of expertise: Why Rational Choice Theory is (often) a fiction

The premises of Rational Choice Theory

In order to examine closely the fundamentals that bring together the rich and diverse developments of RCT, we need to get back to the basics. First, Bayesian explanations are modeled upon an agent with a coherent structure of preferences. It assumes the existence of an agent with a limited and known set of individual preferences in a world that is taken for granted. An internalist explanation locates the causal and motivation power of action somewhere between the neurological phenomena and the environment, at the psychological level of mental states. The outcome of the action is considered a given. Since agents construct their preferences based on supposed outcomes, they are at the same time the motor for the action. As we will explain, this circularity is what makes RCT an atomistic explanation, isolating the decision-making process from its context.

The second premise tells us that our choice is based on a comprehensive and exclusive access to information, so we are rationally in control of the situation. But we often discover new preferences, or act to modify them, lie, or deceive. In sociological research, we continually run into inconsistent agents. Thus, the vast majority of lay and professional decisions fall outside the RCT model. For instance, the assumption of intransitivity describes the structure of individual preferences as consistent and equivalent. In everyday life, we constantly change preferences. When ordering in a restaurant, how many times have we wavered between two options, A or B (soup or salad), and knowing that A was available, ordered C (pasta) when the waiter finally came along? We are (biologically and socially) inconsistent, and recent findings in biology support this claim, as we elaborate in later sections (Damasio, 1994; Premack, 2004; Resnick, 1994; Tooby & Cosmides, 1992).

Third, the Bayesian model defines rationality as essentially teleological: agents gather information from others, from the environment, or from their own actions, to fulfill their wishes. Expert performances are often substantive and not instrumental, as we will demonstrate in later sections.

Fourth, the RCT model represents a continuation of the psychological explanations of common sense, without going through the necessary epistemological break (Bachelard, 1986). Folk Psychology gave rise to AI models. The classic model of AI (or GOFAI) represents the clearest academic attempt to demonstrate Western dualism empirically. This epistemic tradition, stemming from Descartes and ultimately Plato, extends to the Illustration ideal of Rationality. Studies in cognitive science—and also in anthropology, sociology, and social psychology—propose that the belief in a unique, rational, golden standard transcending the social context is a form of reification. From within the social sciences, one cannot take expertise for granted, but must look for the mechanisms of social production of cognition. As Carr (2010) claims “It is precisely the widely held ideas that language primarily functions to denote preexisting states and that those states are the inner property of speakers that so frequently naturalize expertise as something one has rather than something one does” (p. 26).

Fifth, we are always directed to a maximization model of rationality, which constitutes a black box for decision-making. Despite its failure as an empirical program, the key to RCT success is to be found not so much in its explanatory scope as in its capacity for producing formal predictions. While its psychological model is a black box for action and cognition, this allows for the production of simulations and formal models, often with a mathematical basis and delicate computational geometries. A narrow definition of what constitutes behavior and rational decision-making is a requirement of the model’s formality. As we will point out, experimenters are well aware of their explanatory limitations, but are seduced by the possibility of creating simulation models, the results of which are all too readily published within the scientific community.

These five assumptions have been slightly modified by three developments in RCT: game theory, bounded rationality, and heuristics. These schools of thought explain some of the sample behaviors and decisions that fall outside the classical Bayesian model. However, we argue here that while these approaches expand the capacity for explanation, they do so without substantial alteration to the basic model.

The social openness of intentionality

The Rational Choice model revolves around the belief that the mind is a manipulator of symbolic information. This type of explanation is ambiguous and based on cognitive dualism. As we will show in this section, the great paradox of RCT experimentation is that the removal of social rules, leaving just the black box of psychological mental states, makes the agent superfluous (Clark, 1997). ²

GOFAI describes the mind as an organizational system functionally independent of its substance. Its central working hypothesis may be summarized as follows: “A physical symbol system has the necessary skills sufficient for general intelligent action” (Simon & Kaplan, as cited in Clark, 2001, p. 28). Symbols acquire meaning through formal deductive logic or thanks to visual and perceptual processes that interact with short-term memory (Chase & Simon, 1973). In an address given at UC Berkeley in 2010, John Searle claims insufficient empirical evidence for proving the existence of psychological standards below the social level or above the neuronal level. If decision-making follows unconscious psychological rules, it should be possible to make them conscious at some point. Verbalized rules could then be considered a product of intentional states in social contexts of interaction. We introduce an example from psychology of perception presented by Searle to illustrate the superfluous nature of rules that do not have a neurological nor a social basis.

The Ponzo Illusion (Figure 1) consists of seeing the upper line longer than the one at the bottom, even though the two are actually the same length. Within the context of the image, the upper line appears to be farther away. The misjudgment of perception comes from our mental capacity to judge an object according to its context. According to Euclidean perspective, the upper line must be longer than the lower one. Nevertheless, and this is the key argument, the degree of internalization of the Euclidean laws is subject to cultural and historical variations. Even basic perceptual laws such as the Ponzo Illusion supervene moments of social construction, based on an open and evolved biological apparatus (Tooby & Cosmides, 1992).

OPEN IN VIEWER

Figure 1.

The Ponzo Illusion.

Moreover, simplified decisions do not represent complex everyday life. Common experiments in Rational Choice Theory aim to explain decision-making from the principle of speed accuracy trade off. The agent seeks correct answers over a short period of time, showing that players make decisions based on the principle of efficiency maximization. In the experiment, agents watch a screen with fairground ducks and try to guess the existing ratio of red to yellow ducklings. This is an experiment without replacement, such as tossing a coin, with a restricted explanatory scope, which is describing a principle of maximization in an experimental setting. RCT works in constrained social contexts, such as a total institution (Goffman, 1961) or in simplified economical or moral decisions, as explained in Simon’s (1991) concept of bounded rationality. In these social situations the agent’s capacity for cognition and decision-making is strictly limited and scripted. The RCT model subsumes the agent in the formal superstructure of institutions and shapes the workings of the mind. Folk theory of mind can only explain competitive situations where the agents’ intentionality is non-existent or irrelevant.

While Rational Choice considers instrumental rationality one of the key dimensions of decision-making, we claim that strategy is only one among many types of decision-making. Going back to the prisoner’s dilemma, wanting to cooperate is based on the tit-for-tat strategy. In everyday life, the need for cooperation is expressed in moral norms, as in “Do not do to others what you would not want done to yourself”: in social reciprocity (Simmel, 1908/1971), or in Kant’s imperative, this is a moral principle that encourages substantive rationality. Other possibilities for action would be to follow the authority of tradition or leadership. Strategy becomes one option among others.

The RCT premises ignore the heterogeneity of decision-making. John Elster’s (1998) interpretation of Antigone is an example of the insufficiency of the model. Antigone is sentenced to death by the city government because of a righteous act: her brother’s burial. The philosopher states that Antigone is being irrational because of the psychological weakness of the will. ³ However, Antigone would not exist without the social opposition between her family and Creon’s political will. She is torn between two legitimate authorities: that of the family and that of the city. Her stubbornness that sees her condemned to the death penalty is the product of her fidelity to the laws of the family, which are as social as those of the state. The Greek tragedy makes explicit a conflict between family and state laws, ultimately social institutions that fight for legitimate power and authority. Sophocles questions the power of the city, making clear that Antigone has the institution of the family to back her up. While Elster takes Antigone to be the representation of a conflict between the individual and the society, between the psychological and the social, Sophocles gives her a rationale in the social laws of sisterhood. Thus, both Antigone and Creon have good reasons (Boudon, 2003) to act as they do.

The biological basis for decision-making

Cognitive dualism maintains a gap between the neurobiological and the social level, and in so doing ignores both the social and the neuronal basis for decision-making. In this section we deem this causal gap is unnecessary: mental states can be explained as intentional states caused by the neuronal level, and shaped by the social dimension of interaction. We will take a stand towards a monistic explanation of decision-making and cognition.

Herbert Simon’s (1969) fable of the beach and the ant illustrates the classical AI and RC theoretical standpoint. If we see an ant moving through sand dunes, we can explain its history, objectives, direction, and the characteristics of its trajectory by looking at the characteristics of the land rather than attending to the ant’s mental content states. The functional emphasis on interaction glosses over the fact that computer circuits are materials other than the original biological matter.

Behavior is not a phenomenon sui generis. It is the product of mechanisms that process information … A cognitive description specifies what kinds of information the mechanism takes as input, what procedures it uses to transform that information, what kind of data structures (representations) those procedures operate on, and what kind of representations or behaviors it generates as output. (Simon & Kaplan, 1989, pp. 64–65)

The cognitive process as defined by Simon and Kaplan (1989) remains unique and general. The Nobel laureate’s use of interactivity is restricted by the vocabulary and assumptions of his own program. RCT goes well with Skinner’s (1945) behaviorism, and behaviorist accounts are certainly possible outcomes of its empirical application.

It is difficult to give a single definition of intelligence. Recent discoveries in the fields of cognitive science and developmental biology put into question the idea of a mind as a symbolic and linear information processor. Experiments in neuroscience and the social sciences show the decentralized characteristics of the human mind. Ontogenetically, symbols are physical items that are in the environment first and only later internalized (Hutchins, 1995, p. 370). Ramachandran and Blakeslee (1998) compare our mental abilities with a bag of tricks. Based on dynamic systems theory, authors who follow the Thesis of Radical Embodied Cognition (TREC) such as Thelen and Smith (1994) deem unnecessary the attribution of conceptual content, such as representations, categories, or concepts, to human behavior. Authors from embodied cognition such as Clark (1997, 2001, 2008a, 2008b) picture the brain as an organ of local control, considering physical coordination as the key element for cognition and action. From neuroscience, connectionism describes the brain as a nonlinear system (Churchland, 1995).

These findings include the environment not only for interaction, as in Simon’s proposal, but as part of an integrated approach (Barkow, Cosmides, & Tooby, 1992). Our cognitive processes are part of a larger system made of agents, artifacts, and interactions. As both embodied philosophers and connectionists claim, we do not operate, at least most of the time, following maps in our head, but by using the resources in our physical and social environment. The state of our environment inclines, disposes, encourages, or influences our sense of action. A key concept in explaining the workings of the cognitive system is that of affordance, deriving from the psychology of perception. An affordance refers to “the possibilities for use, intervention, and action offered by the local environment to a specific type of embodied agent” (Gibson as cited in Gibbs, 2006, p. 50). Thus, when a three-year-old wants to play with her younger brother’s toys, which are not appropriate for her current age and size, she is following experiences from the recent past, in which the toy is a solicitation, an affordance. ⁴

The process of tangram formation (see Figure 2) can further explain the holistic dimension of acting in the world. Tangrams are Japanese black puzzles that are assembled in a certain manner to create unique shapes. The resulting form (a Gestalt) does not appear until found by the player (following Picasso’s famous words Je ne cherche pas, je trouve: I don’t search, I find). Kirsh (2010) explains how expert players experiment physically with shapes and sizes, forming the correct final figures, which are multiple, but not infinite. Experts manipulate more effectively the black pieces over the surface. Thus, they are better in soliciting the functional and physical properties of the tangrams, in order to complete the puzzle more efficiently. They act by a holistic mechanism of pattern recognition and not only by normalized inferences. Following Dreyfus, one of the main authorities in phenomenology: “The parts get their meaning from the whole, even though that whole does not exist at any moment of time. It exists, as one may say, in the subject’s mind, as an intent… a Gestalt” (Dreyfus, 1979, p. 244).

OPEN IN VIEWER

Figure 2.

Tangrams.

Our cognition processes are not atomistic, which counters the first assumption from the Rational Choice Model. In contemporary cognitive science, as in Norman (1988), Zhang and Norman (1994), Perry (1999), Wright, Fields, and Harrison (1999), Hollan, Hutchins, and Kirsh (2000), Barsalou (2003), Resnick (1994), Kirsh (1991, 1995, 2010), and Chandrasekharan and Osbeck (2010), the environment becomes an ally or an opponent in the agent’s struggle to perform from the simplest to the more complex tasks. Emotions such as fear act as filters in our decision-making, and can also be a feature of our evolutionary biology equipment. ⁵ The prisoner’s dilemma happens because players are afraid to cooperate. Confessing is not the best option, but it is a reasonable decision if we assume that the other prisoner might talk after all: a rational assumption, if we consider the lack of empathy that occurs in most professional environments. ⁶ Kirsh and Maglio (1994) show how expert players in handling non-symbolic manipulation, as in a Tetris game, prefer to physically manipulate the buttons on the console instead of mentally rotating the figures. Moving the finger and pressing the button several times per second simplifies the internal decision-making process. Rumelhart (1989) points to three basic metacognitive abilities that describe biologically the ontology of the cognitive process: pattern matching, world modeling, and manipulating the environment.

Reification or the epistemic error of attribution

Chess, puzzles, or mathematical games, which constitute most rational choice experiments, are training sites for a specific conceptual domain, that of analytical logic. RCT works with a type of rationality that is selectively applicable to games of logic or scientific enquiry, but does not explain everyday cognition. The generalization of the RCT model to all social arenas of knowledge production amounts to reifying the explanation of rationality. To consider that scientific rationality plays the same role in the supermarket, in moments of artistic creation, or when we bathe our daughter, is part of the well-spread scholastic fallacy of putting forward a theoretical principle—that is the product of a historical and cultural period of time—as a practical principle of cognition. RCT authors fall into reification, that is, taking an event-type (such as the definition of rationality in experimental contexts) as an event-token (rationality as observed in professional environments; Pylyshyn, 2003).

In Figure 3 we show the problem of the conditional as presented by Kalish and Montague (1964). It illustrates the possible contradiction between the logical criterion of truth and decision-making based on experience. Kalish and Montague apply an indirect derivation, a logical operation, which runs as follows: “If you can show that the assumption S [If P, then Q (or if not P, then Q)] leads to two contradictory propositions (Q and not Q), the assumption not-S must be wrong, and Q must be right.” The outcome is that Q is both X and not X, and so it is contradictory. Not Q is false, so Q must be right. However, in real life, Alfred can drop out of college, he can change his mind and start to enjoy studying because of new friends or subjects of interest, or he might end up with bad grades simply because of an unfair professor. Agents change their preferences, and their conception of work or fun, according to their experience and that of others.

OPEN IN VIEWER

Figure 3.

The specificity of formal reasoning (Kalish & Montague, 1964).

This example highlights the mismatch between mathematical and social reality. While an operation such as the indirect derivation makes sense in the domain of formal logic, it falls into contradiction in professional and everyday decision-making where P and Q necessarily have analogical content. ⁷ Unless P and Q are associated with meanings consistent with our cultural experience, we have nothing to say about their relationship. Experimental participants in game theory make mistakes all the time, specifically in logic, representation, and availability. Medical and graduate students make bad judgments. Their experimental decisions are often deemed as incompetent by the experimenters (Griggs & Cox, 1982; Tversky & Kahneman, 1974; Wason, 1966). According to cognitive anthropologist D’Andrade (1995), theirs is not a calculation but a representation error. The experimenters ask for types of reasoning skills that are not called upon in most professional environments. Logical premises that have no cultural coherence are not understood, or seem confusing. Rationality correlates with the type of social context the agent is in (Garfinkel, 1967, 2006). Therefore, our decisions are socially located.

Acknowledging the social root of mental categories such as concepts or schemas implies a redefinition of our object of research. Avoiding reification requires taking new data into account. Decision-making becomes physically open and can be explained through the observation of real cognitive and interactive processes. As stated by Sawyer (2005), Durkheim (1912/1998), and Bourdieu (1994, 1998), among others, it is not possible to understand cognition and decision-making without studying its social organization. More specifically, the technical and functional synchronization of resources are key in analyzing real work settings (Garbis & Artman, 1998; Muntanyola, 2011; Zhang & Norman, 1994).

Technical gossip (Knorr-Cetina, 1999), narratives (Silverstein, 2006), and other informal communication strategies are necessary tools for avoiding the confirmation bias or other errors in professional decision-making. Knorr-Cetina’s observations of laboratories, transportation control centers, classrooms, and other professional environments that require teamwork emphasize trust pathways and point towards the existence of technical gossip as the main communication pattern for expert decision-making. Expertise is based on linguistic codes, jargon, and scientific terms. The chief doctor in the hemodynamic unit must decide: (a) which parts of the body need to be visualized, (b) which steps are to be taken to move the fluoroscopy within the vascular system, and (c) which particular brand and type of stent the nurse must select from the models available at that particular moment (Muntanyola, 2012). In order to become a medical expert, it is necessary to be able to verbalize the options that are cognitively available at any given point in the functional process. Together with these narrative resources, modalities are crucial elements of communication. Not only verbal interaction, but also our bodies have an important role in decision-making (Alac, 2005; Boyer, 2005; Muntanyola, 2012; Myers, 2008). ⁸ The interaction goes through several stages of cognitive processing, which build into multimodal interactive systems (Alac, 2005). The centrality of embodied communication patterns explains its preponderance in professional environments, such as hospital units. Gesture, the speed of turn taking, and the existence of synchronized feedback are constant ingredients of a successful professional conversation. Detailed ethnographic analyses show how face-to-face interactions are paramount and dominate those interactions mediated by technological tools such as phone or email (Hutchins, 1995; Muntanyola, 2012; Myers, 2008).

Cognition is no longer individual but distributed. Recent findings confirm the holistic nature of our cognitive abilities and have precipitated a change in the level of explanation in both cognitive psychology and anthropology. In the wake of Roberts (1987), Hutchins (1995, 2001), Hollan et al. (2000), and Kirsh (1995, 2010), the object of study is no longer the individual or the interaction, but the system or group of agents, instruments included. Distributed cognition recognizes the existence of a plurality of instruments and communication patterns among professionals as well as the existence of coordination moments that lead to an empirically sound account of decision-making. In the same cognitive ethnography of a hemodynamics unit, both the vascular doctor and a nurse measure the size of an injured artery in unison. Two parallel cognitive sub-processes are performed at the same time and place, distributed across two agents: the specialist doctor verbalizes the possible numerical values, while the specialist nurse measures the size of the injury on the computer screen with a rule and a piece of paper. The final conclusion is arrived at through a verbal negotiation of these two sub-processes.

The need for a pragmatic ontology

The classic AI program constitutes the experimental application of RCT. The empirical history of artificial intelligence developed in parallel with the generation of computer programs that could simulate and improve the performance of specialists. In the 1960s, the aim of an AI empirical program was to construct a General Problem Solver (GPS), starting with a single process to resolve all kinds of cognitive operations related to human activity, from the strictly logical to the mundane. Cognitive processing was approached from a syntactic point of view, without going into the semantic content of information (Dreyfus, 1979). The need for making explicit all possible decision alternatives outreached the creators’ imagination and the computational results of such GPS operations did not correspond to human decision-making. In a second phase, Minsky (1968) and others at MIT developed programs of semantic information processing, which were closer to specialist cognition. The computer stored information about normal cognitive processes in daily life and also from specific professional domains, such as INTERNIST for medical diagnosis. The most widespread functional programming language was LOGO (invented in 1967 at MIT). It was used broadly in high school computing classes in the late 1980s in both Europe and the US. A virtual turtle was able to “learn” from situational elements. Students could drive and “teach” the turtle, through programming direction and speed, how to move around the computer screen.

When replicating expert behavior, Minsky and others based their empirical work on a fragile assumption about human cognition. AI creators believed that intelligence, especially expert knowledge, was based on processing large volumes of information. This quantitative approach—according to which machines could think—became a caricature of reality. ⁹ Previously, success had been limited to formal domains, such as chess or statistics. In this same period, however, Weizenbaum (1966) created the ELIZA program (1964–1966 at MIT), where the computer responds to the users’ interrogation regardless of its semantic content. ¹⁰ Researchers focused on microworlds and other domain-specific programming. Winograd (1972) identified a number of environmental features that a computer needed to perform a sequence of actions. The machine, named SHRDLU, moved droids around without direct instructions, following a set of predefined rules. This experimental program was an important step towards externalizing the cognitive process, a trend in contemporary AI. Current success in robotics comes from machines than can interact with the environment and perform simple but cumbersome physical tasks, such as lifting heavy objects or cleaning around the house (Mahru-Z, KIST; You, 2010). Finally, Schank (1975; Schank & Cleary, 1995) attempted to replicate or simulate everyday activities (like ordering a meal) using scripts or narratives. However, the complexity of social life renders it near impossible to replicate in binary code all the decision-making processes involved in a night out.

Our cognitive processes cannot be separated from experience, where the exception is the norm. Weizenbaum, at the beginning of the GOFAI experimental program, claimed that a normal conversation builds up, on the fly, multiple contexts, sub-texts, etc., that cannot be defined prior to the interaction (Dreyfus, 1979, p. 219). The empirical failure of computational expertise demonstrates the fallacy of considering knowledge as the sum of isolated and atomized decisions. Contemporary cognitive science, specifically authors from embodied cognition (Clark, 2008a; Gibbs, 2006; Myers, 2008; Nuñez, 1995) and cognitive ethnography (Hutchins, 2006; Muntanyola, 2012), acknowledges the agent’s embeddedness in social situations. The objective account of behavior cannot be completed without the subjective insight of all the agents in the research process. As Harré and Gillet (1994) point out: “The experiment or observation has to enter into a discourse with the being studied and try to appreciate the shape of the subject’s cognitive world” (p. 21).

An efficient ontological perspective seems to be pragmatism, where decisions and cognitive processes are taken as dynamic elements in interaction. The study of communication patterns includes the study of gestures and facial metaphors, action, and movement based on the experience of the body. Space management and memorizing the choreography are two of the skills that make dancers experts in their chosen artistic field (Muntanyola & Kirsh, 2010). As shown in Figure 4, the dancer’s objective is to perform the right steps, in the correct order and place. The narrative is a case of distributed memory, since the sequence of movements spreads across the set of dancers. In order to explain what happened, the dancer constructs a narrative based on the conceptual blend spatial position (Fauconnier & Turner, 2002). This decision-making process is based on the other dancers’ moves, from which our first performer creates visual cues. Communication and the order of steps (i.e., the principles of interaction and social stratification) are participatory and inseparable from the reality of the performance. In short, this example shows how the social level of explanation penetrates the psychological level. Thus, the principles of interaction and social stratification become decisive in decision-making. Through this self-reflexive narrative, the dancer displays a specific spatial awareness, which constitutes a form of specialized knowledge. Moreover, spatial awareness becomes a style resource that identifies this dancer as contemporary, as opposed to a ballet dancer who uses music as a memory aid. Authors such as Williams (1985) put forward how the etymological jump from the English adjective expert to the noun expertise happened in the wake of 19th century division of labor. Thus, the dancer becomes an expert when capable of producing and communicating a certain form of cognition, a way of remembering, of moving, of creating new steps that do not belong to everyday life. This professional way of moving—recalling Goodwin’s (1994) well-known paper “Professional Vision”—lies at the root of the dancer’s expertise. Pragmatic observation replaces psychological cognitivism. We make decisions from our analysis of the Other.

OPEN IN VIEWER

Figure 4.

A dancer’s narrative on a spatial error on stage.

Conclusions

Individual cognitive processing includes gestures, tools, and material affordances, as well as social norms for communication and interaction.

Cognition as defined by the Rational Choice model is an ambiguous concept closer to fiction than to reality. Nevertheless, it is deeply rooted in our folk conception of individual psychology and decision-making. Participants in RCT experimental contexts stop playing, change the social rules of the game, manipulate results, and reinterpret choices. In The Purloined Letter (1845/1977), Edgar Allan Poe describes the Even & Odd game. Even & Odd presents a choice between a pure or a mixed strategy. Poe holds that if the other player is better at the game one should play randomly, while with an inferior player, whose intentions could be intuited through empathy, it makes sense to hazard an informed guess. Thus the other is not only a parameter to be taken into account, but a key dimension in the construction of the game process. Intentionality and empathy, two mechanisms that have evolved and are still evolving, are central to the decision-making process.

A review of the classical AI experimental program shows that simulators outside the domains of logic have not beaten the best human experts. Cognitive scientists, sociologists, and anthropologists alike flee the white room of experimental behavior and work with agents that actively participate in communicative and interactive patterns in the social world. More specifically, contemporary cognitive science moves away from the classical view of rational choice theory. We are discovering how we perform basic cognitive operations thanks to our socio-physical environment. As anticipated by Durkheim (1912/1998), Wittgenstein (1953), and now Clark (2008a, 2008b), institutions are cognitive requirements too. The synchronization of cognitive agents results in deeper interaction that shapes us as social beings.

This claim has two main consequences for research in decision-making. First, such a pragmatic approach assumes the need for experience in making reasonable decisions. Given the social nature of the cognitive mechanisms, specialists create multimodal interaction systems. Gestures are an important part of communication in professional environments. The doctor’s gestures and the dancer’s movements are not indicators of something beyond what we see, but cognitive decisions themselves. Speech is not only the main tool for conveying information, but also a key mechanism in solving problems and making decisions. Situated activity demonstrates how the instruments we use in our daily and professional lives shape the type of conceptual operation we perform. The communicative acts between members of a team are not about the cognitive process, but become the cognitive process.

Second, there is no single cognitive processing, but several: in real-life distributed cognition systems timing and social hierarchy become cognitive tools. Following Wittgenstein, Part II (1953), the findings support the impossibility of private language, thus, the impossibility of a private cognitive process. The current consensus on the biological status of cognitive processes is that our brain’s response to the demands of the physical environment results in our interaction with the world around us. Our cognitive nature leads to complex decision-making based on intentionality, empathy, and multimodality. This new integrated model in social sciences rejects the notion of an agent that is bounded, limited, and confronted by the physical and social world. The cognitive agent that we find in the social world is intentional and creatively embodied.

In all, our pragmatic stance looks for the social roots of expertise in the communication and synchronization patterns of specialists and experts working in their everyday life environments. Instead of looking for psychological rules for action, which as we have shown are trapped in the black box of Folk Psychology, we ought to explain the social rules for giving to certain acts and words the authority of expertise.