The DBO theory of action and distributed cognition

Abstract

French

The DBO (Desires, Beliefs and Opportunities) theory of action proposed by analytical sociologists aims to provide an action-theoretical basis for building explanatory theories in sociology. Peter Hedström claims that the DBO theory is realistic because it does not make assumptions that are known to be false or seriously incompatible with the current scientific understanding about the nature of human action and cognition. This article nevertheless aims to show that the DBO theory is not only incomplete but also that its background assumptions are unrealistic, in the sense that they do not fit with the distributed nature of action-related cognition, which has recently become a growing topic of interest in cognitive sciences. The author also indicates that the neglect of the distributed and embodied aspects of cognition in the DBO theory leads to various biases in the process of constructing mechanism-based explanations in social sciences. Finally, an alternative approach to action theory is sketched on the basis of this critique.

Keywords

action theory analytical sociology collective agency distributed cognition mechanism-based explanation

The nature of human action and the possibility of action-based causal explanations have been debated in sociology since the beginning of the discipline. Recently, an intellectual movement called analytical sociology has taken these themes under systematic investigation. One of the pioneers of analytical sociology, Peter Hedström (2005), has formulated a DBO (Desires, Beliefs and Opportunities) theory of action which aims to provide an appropriate micro-foundation for explanatory theories in sociology by using the words desire, belief and opportunity as the primary theoretical terms. The DBO theory is not intended as an entirely new approach to human action, but instead aims to provide a systematic account of the action-theoretical views that are already presupposed in many theoretical and empirical studies associated with analytical sociology.

In this article, I ask how realistic the DBO theory is from the perspective of recent cognitive scientific studies, especially those related to distributed cognition. The issue of realisticness thus concerns the compatibility between the assumptions of the DBO theory and recent approaches in cognitive science. In my view, this question is legitimate because Hedström (2005: 38) not only presents the DBO theory as an analytically realistic model of ‘the type of actors involved in the action’ but also explicitly requires that ‘[t]heories of action should be based on empirically grounded knowledge’ (2005: 65). He specifies these views by writing that, though analytically realistic models are incomplete in the sense that they ‘intentionally move out of focus all elements that are deemed inessential to the problem at hand’ … they nevertheless are ‘realistic because the elements that are retained are believed to reflect the real processes at work’ (2005: 38). On the basis of the analysis and evaluation of its background assumptions, I argue that the DBO theory contains many unrealistic views on the nature of human cognition. I also aim to show that, when used as a general framework for constructing mechanism-based explanations of social macro-phenomena, it leads to unjustified bias towards explanations in terms of individuals’ intentional actions and interactions.

In the assessment of the background assumptions of the DBO theory, I draw mainly on the perspective of distributed cognition. I have chosen this perspective because it contains theoretical views and empirical findings on action-related cognition that are highly relevant to social-scientific action theories. The distributed cognition approach emerged in the 1980s and 1990s within cognitive science, though it surely has antecedents in other fields including anthropology and sociology (Clark, 1998; Hutchins, 1995; Salomon, 1993). The basic idea of this approach is that many cognitive processes in our everyday life transcend the boundaries of the skulls and skins of individuals since they are distributed across many communicatively interacting individuals and the cognitive artifacts (i.e. culturally produced entities employed in information processing) used by them. It thus rejects individualist and disembedded interpretations of human cognition and emphasizes the materially, socially and temporally distributed nature of many cognitive processes that take place in everyday action. The topics concerned with distri- buted cognition have recently received some attention in the fields of cognitive social theory (for a review, see Strydom, 2007) and cultural sociology (for reviews, see DiMaggio, 1997; Lizardo & Strand, 2009). They are nevertheless seldom explicitly invoked in action-theoretical discussions which address methodological issues about causal explanation.

The DBO theory and analytical sociology

Although analytical sociology as an intellectual movement is a fairly recent pheno- menon, the roots of the ideas associated with this movement can be traced to classical sociology (see Manzo, 2010). For example, the works of Max Weber, Alexis de Tocqueville and Robert Merton are commonly cited as important antecedents. Raymond Boudon, James Coleman, Jon Elster and Thomas Schelling are in turn seen as contemporary social scientists whose influence in the formation of analytical sociology has been profound (Hedström, 2005; Hedström & Swedberg, 1998; Manzo, 2010). To my knowledge, Hedström’s (2005) book, Dissecting the Social: On the principles of analytical sociology, is nevertheless the first book-length exposition of the basic concepts and doctrines of this approach.

Analytical sociology, according to Hedström (2005: 1), ‘seeks precise, abstract, realistic and action-based explanations for various social phenomena’. Explanations of this kind are constructed by dissecting (or decomposing) the complex social phenomenon to be explained into its essential constituent entities and activities. While focusing on crucial explanatory factors, these explanations abstract from the unimportant details. By following this procedure, analytical sociologists aim to provide relatively general explanatory theories which refer to social mechanisms that generate the range of social phenomena they are explaining (2005: 2–3). In general terms, a mechanism can be understood as ‘a constellation of entities and activities that are organized such that they regularly bring about a particular type of outcome’ (2005: 25). Social mechanisms, in turn, are composed of individual actors and their intentional actions (2005: 28–29). The DBO theory of action, therefore, aims at providing the basic theoretical concepts for building explanatory theories that refer to social mechanisms (2005: 37–38). Hedström exemplifies this approach to social explanation by referring, among other examples, to Merton’s notion of self-fulfilling prophecy, Harrison White’s theory of vacancy chains and Schelling’s analysis of the various segregation processes. It is important to note, however, that not all analytical sociologists share a common definition of the concept of social mechanism or a common theory of action, even though the interest in mechanism-based explanations unifies them (see Hedström & Swedberg, 1998; Hedström & Ylikoski, 2010).

Hedström and Petri Ylikoski (2010: 50–52) nevertheless argue that the following four general ideas are shared by the most acceptable definitions of the concept of social mechanism:

‘a mechanism is identified by the kind of effect or phenomenon it produces’;

‘a mechanism is an irreducibly causal notion’;

‘the mechanism has a structure’;

‘mechanisms form a hierarchy’ because ‘a mechanism at one level presupposes or takes for granted the existence of certain entities with characteristic properties and activities’ and ‘it is expected that there are lower-level mechanisms that explain them’.¹

The mechanism-based model of explanation is also typically presented as an alternative to the covering-law model of explanation and its statistical variants (e.g. Gross, 2009; Hedstöm, 2005: 15–33; Hedström & Swedberg, 1998; Hedström & Ylikoski, 2010). This means that mechanism-based explanations are not conceived of as subsumptions of the described phenomenon under a set of law statements (or statements about statistical regularities), but rather as descriptions of the causal mechanism(s) that brought about the phenomena to be explained. This is also the reason why the search for mechanisms helps to distinguish genuine causal relations from merely coincidental associations. Furthermore, analytical sociologists often emphasize that mechanism-based explanations open black boxes by disclosing the structure of the mechanism that produces the phenomenon under explanation (e.g. Boudon, 1998; Elster, 1989) and criticize such correlation-based ‘statistical causal models’ where ‘variables have replaced actors as the active subjects with causal powers’ (Hedström, 2005: 105). In analytical sociology, then, statistical associations are regarded either as phenomena that require theoretical causal explanation in terms of social mechanisms or as evidence for the proposed theoretical explanation.

In addition to the notion of social mechanism, a methodological doctrine of structural individualism is commonly advocated by the current proponents of analytical sociology. According to Hedström and Peter Bearman, structural individualism holds that ‘all social facts, their structure and change, are in principle explicable in terms of individuals, their properties, actions, and relations to one another. It differs from traditional methodological individualism … by emphasizing the explanatory importance of relations and relational structures’ (Hedstrom & Bearman, 2009: 8; see also Hedström, 2005: 5). In accordance with this view, structures and properties of social networks are counted among the factors that explain the outcomes of the mobilization efforts in Hedström’s (1994) empirical study on the Swedish trade unions. Hedström (2005: 68–74, 115) nevertheless warns us about ontologizing methodological distinctions between the properties of individuals and the properties of collectives by ascribing autonomous causal powers to the latter, and insists that ‘it is individuals, not social entities, which are endowed with causal powers’ (2005: 115). Hence, structural individualists favor such bottom-up explanatory strategies in which macro-level social phenomena are explained by modeling the complex social interactions on the part of many individual actors that are thought to generate these phenomena (e.g. Hedström, 2005: 75–100).

Status and concepts of the DBO theory

The DBO theory is influenced by the rational-choice approach though it is intended to be more comprehensive and realistic than the latter. Hedström (2005: 61), for instance, emphasizes that ‘DBO theory makes no assumption that actors act rationally …; it only assumes that they act reasonably and with intention’. Rational action in the sense of rational-choice theories is then conceived of as a special case of the DBO theory rather than the only form of human action (Hedström, 2005: 41, 61; see also Elster, 1989). Furthermore, as Gianluca Manzo (2010: 151–157) indicates, rational action in analytical sociology is not necessarily conceptualized as the expected utility-maximizing behavior of individuals, due to the fact that other models of human rationality have also been formulated by analytical sociologists. They include Raymond Boudon’s (e.g. 1998) model of cognitive rationality, Michael Hedström’s (2005: 48–49) model of mimetic rationality and Macy’s (1997) model of emergent rationality. For these reasons, critiques raised against rational- choice theories do not automatically apply to the DBO theory.

In the DBO theory, as noted above, ‘[d]esires (D), beliefs (B) and opportunities (O) are the primary theoretical terms upon which the analysis of action and interaction will be based’ (Hedström, 2005: 38). Although these three terms form the conceptual core of the DBO theory, Hedström does not deny that other factors, too, such as emotions and social relations, may play an important role in some action-based explanations. The concept of belief is defined as ‘a proposition about the world held to be true’, and the concept of desire as ‘a wish or want’ (2005: 38). Drawing on analytical philosophy of action, Hedström interprets desires and beliefs as intentional mental events (or states) of individual actors, and claims that an individual’s desires and beliefs ‘can be said to cause an action in the sense of providing reasons for the action’ (2005: 39). The concept of opportunity in turn refers to ‘the actual set of action alternatives that exists independently of the actor’s beliefs about them’ (2005: 39). Opportunities not only provide action possibilities for individual actors, but the set of available opportunities also constrains the actions of individuals. Although the opportunities exist independently of the actors’ beliefs, they nevertheless influence his/her actions via his/her beliefs about them (2005: 39). Hedström thus contends that ‘[t]he cause of an action is a constellation of desires, beliefs and opportunities in the light of which the action appears reasonable’ (2005: 39).

The DBO theory recognizes various causal interconnections between an individual’s desires, beliefs and opportunities (2005: 40; see also Elster, 1989). Individual actors may, for example, unconsciously modify their desires so that they end up desiring only what they believe they can get. The expression ‘sour grapes’, originating with the fable ‘The fox and the Grapes’, nicely exemplifies the logic of this psychological mechanism, termed ‘adaptive preferences’. Another example in which actors’ beliefs affect their desires is a psychological process wherein actors end up desiring only the objects that they believe are beyond their reach. This mechanism of counter-adaptive preferences is captured in the saying: ‘The grass is always greener on the other side of the fence’. Hedström (2005: 43) also argues that the mental states of individual actors, including their desires and beliefs, can be affected by the actions of other individuals. Indeed, many of the most interesting social mechanisms – such as self-fulfilling prophecy, dissonance-driven desire formation and rational imitation – involve social interactions of this kind (2005: 47–59; see also Elster, 1989; Rydgren, 2009). This view clearly differentiates the DBO theory from the atomistic interpretations of individual minds.

It is worth stressing that the unintended consequences of intentional actions of interacting individuals play a crucial role in many social mechanisms. For example, Merton’s notion of self-fulfilling prophecy refers to a process in which an initially false belief becomes true as the unintended consequence of actions and interactions that are based on this belief (e.g. Hedström, 2005: 48). Elster (1989: 95) offers another example wherein erosion is the unintended and undesired outcome of many farmers trying to get more land by felling trees because large-scale deforestation leads to erosion.

The background assumptions of the DBO theory

On the basis of the previous discussion, I have listed six background assumptions of the DBO theory of action.

(1) The first assumption says that, from the perspective of explanatory social research, interacting individuals should be conceived of as the only causally efficacious actors in society. This view is evident in Hedström’s (e.g. 2005: 19, 115) explicit denial of the existence of the causal powers of social entities and his methodological position of structural individualism. Structural individualism states that all social phenomena are ultimately explainable in terms of interrelated actions and interactions of individuals – including their intended and unintended outcomes. The term ‘social emergence’, according to his view, should be interpreted as an epistemic notion that refers to social properties that ‘cannot, in practice, be predicted by knowing everything that there is to know about the pre-emergent properties of the parts’ of social phenomena (2005: 74–75). Hedström (2005: 75–100; see also Hedström & Bearman, 2009: 9–10) nevertheless accepts the use of concepts that refer to social relations between individuals, structures of interaction networks and norms in mechanism-based causal explanations. In all events, his critique is aimed at ascriptions of ontologically emergent causal powers to social entities and at explanations of social phenomena that refer to causal powers of this kind (Hedström, 2005: 70–74).²

(2) The second assumption states that for the purposes of explanatory social research, agency (capacity for intentional action) of individual actors can be taken as a black box. It is important to note that Hedström (2005) nowhere denies the existence of the lower-level psychological and neurophysiological mechanisms that underlie individual agency. He does not, however, pay detailed attention to these mechanisms and occasionally appears to suggest that, whatever these mechanisms may be, their details are not sociologically relevant. In their recent paper, Hedström & Ylikoski (2010: 60) nevertheless explicitly admit the need for an account of human agency that is based on knowledge about the lower-level psychological mechanisms. To my knowledge, no one has yet attempted to systematically ground the DBO theory in theories and results of the current psychological and cognitive sciences. For this reason, it is fair to say that agency is taken as a black box by the proponents of this theory.

(3) According to the third assumption, desires and beliefs are best understood as causally efficacious mental events and states of individuals. As we saw, this view is explicitly advocated by Hedström (2005: 39). Though he insists that only individuals can have desires and beliefs, he admits that the desires and beliefs of an individual may influence each other and be affected by the actions of other individuals. This assumption and some of his other views (see below) also suggest that the conscious cognitive processes, such as decision-making and deliberation, should be understood as sequences of mental events in individuals.

(4) The fourth assumption is that actors think before they act. Hedström (2005: 59) explicitly describes the ‘baseline’ of actions of a single individual as follows: ‘among the set of alternatives, the actor chooses the action believed to bring about the desired outcome’ (emphasis added). Hence, in the DBO theory – as in many other social-scientific action theories – the actor’s decision-making between action alternatives is temporally separated from his/her concrete bodily action that aims to realize his/her choices. The rationality (or the reasonability) of action thus concerns the cognitive process in which the actor chooses the best means that he/she believes will enable him/her to achieve the desired outcome.

This interpretation might be challenged by insisting that social mechanisms, according to Hedström, often involve series of decision–action cycles by many interacting actors whose actions modify each other’s desires, beliefs and opportunities. Though this statement is certainly correct, it is also perfectly compatible with the view that the sequences of social interaction of this kind can be – at least in principle – decomposed into the interconnected ‘unit acts’ of individuals. It is also further assumed in the DBO theory that these unit acts can be divided into two distinct temporal phases: the process of choice and the bodily action following it. Even in cases such as dissonance-reduction processes, where the causal influence of others’ actions is mediated by the focal actor’s psychological processes which ‘operate behind the back of the actor’ (Hedström, 2005: 51), other people’s actions are said to affect the actions of the focal actor only indirectly via modifications of his/her beliefs or desires. These modified desires and beliefs in turn form the input to the decision-making process that precedes his/her bodily actions.

(5) The fifth assumption says that in explanations of actions of an individual, his/her conscious intentional mental events have the explanatory primacy over the other causes of action. Hedström (2005: 36) states that any viable action theory ‘should explain an action in intentional terms’, which means ‘that we should explain an action by reference to the future state it was intended to bring about’. Presumably this act of intending has to be a conscious one. The idea behind these views appears to be the following Aristotelian principle: Unless some external factor (or weakness of the will) prevents it, the right type of action follows once the actor has consciously chosen the action alternative that he/she pursues. As we have seen, Hedström (2005: 43, 53–54) nevertheless accepts that unconscious cognitive processes, such as wishful thinking and dissonance reduction, may affect the beliefs and desires of individuals, and thus function as the causes of the desires and beliefs of an individual actor. According to his DBO theory, conscious choices that are assumed to reflect the desires and beliefs of individual actors are nevertheless conceived of as the primary (immediate) causes of the actions of individuals.

(6) Finally, according to the sixth assumption, collaborative learning can be ignored in action-based explanations.³ For the purposes of this article, collaborative learning can be understood as consisting of cognitive processes occurring in situations where many people solve problems, acquire new skills, study course material, produce new knowledge or modify their social practices by acting together (for discussions and various definitions of this concept, see Dillenbourg, 1999; Tomasello, 1999). Collaborative learning usually involves a group of people enacting complementary roles and sharing attention and experiences by means of different kinds of representational media (e.g. texts, pictures, diagrams, maps, physical models). Organizational learning can be seen as a special case of collaborative learning that takes place in some organization. Although I admit that the DBO theory may prove useful in constructing explanations of the outcomes of the trial-and-error type of individual learning as well as some simple varieties of social learning, processes of collaborative learning clearly fall outside its scope because the DBO theory assumes that learning is a process wherein individual actors use ‘information about the past to decide what to do in future’ (Hedström, 2005: 41).

These six assumptions include views and presuppositions about the nature of action-related cognition. In the following, I evaluate the realisticness of these assumptions, basing my assessments mainly on the perspective of distributed cognition. Hence, distributed cognition and its relevance to action theory form the topics of the next section.

The distributed nature of human cognition

The cognitive scientist Edwin Hutchins’s (1995) influential work Cognition in the Wild has already become a classic study of distributed cognition. In this book, Hutchins employs methods of cognitive ethnography and computer simulation to investigate the navigation of a large US Navy vessel. One of the main arguments of the study is that the pilotage (i.e. navigation near land) of the ship is performed by the navigation team as a whole, due to the fact that no single member of the team is cognitively capable of bringing the vessel into port. Hutchins demonstrates in detail why many cognitive processes occurring in pilotage – such as the so-called fix cycles that enable the team to locate the ship on the navigation chart – cannot be adequately explained by focusing solely on the cognitive processes that take place in the heads of individuals, but should be understood instead as materially, socially and temporally distributed. This means that in pilotage the members of the navigation team interact with each other by performing complementary roles that involve cognitive division of labor and exploitation of various cognitive artifacts and external structures. Hence, it can be said that the members of the team participate in the larger cognitive processes that involve materially mediated communicative interactions and spatiotemporally coordinated uses of cognitive artifacts. According to Hutchins’s study, the most important cognitive artifacts employed in pilotage are navigation charts, nautical slide-rules, hoyes, gyrocompasses, alidades, fathometers, bearing record logs, written procedures and phone circuits (note that the study deals with the time before GPS navigation). In addition to the human members of the team, these various cognitive artifacts are counted as proper parts of the distributed cognitive system, which, unlike its individual members, is cognitively capable of navigating the vessel into the port. Hutchins (1995: xiii–xiv) thus maintains that the navigation team, understood as a distributed cognitive system, exhibits ontologically emergent cognitive properties⁴ and, therefore, forms the proper unit of analysis in his study, which seeks to understand the cognitive aspects of navigation.

In addition to navigation teams, proponents of distributed cognition have studied distributed cognitive processes in various educational contexts (Dillenbourg, 1999; Salomon, 1993), in the cockpit of a modern jet aircraft (Hutchins & Klausen, 1996), in a biomedical laboratory (Nersessian, 2009; Nersessian et al., 2003), in a psychiatric emergency department (Cohen et al., 2006), in the heart room of a hospital (Hazlehurst, McMullen & Gorman, 2007) and in other organizations, such as businesses (Secchi, 2010). The common feature of these various distributed cognitive systems is that they are all functionally organized to perform a certain cognitive task or to achieve some more or less specifically defined goal. Furthermore, unlike most systems studied in natural sciences, distributed cognitive systems are hybrids, in the sense that their components (i.e. people and cognitive artifacts) are radically different in nature. Because of the mode of organization of their components, these systems have ontologically emergent cognitive properties that differ both from the non-relational cognitive properties of their components and their aggregates. I suggest that cognitive efficiency achieved by means of coordinated division of cognitive labor, collective problem-solving ability, collective decision-making ability, distributed (or transactive or collective) memory and collective bias can all be seen as plausible examples of ontologically emergent cognitive properties of distributed cognitive systems – all of these collective properties have been examined in empirical studies on distributed cognition (e.g. Hutchins, 1995; Smith, 2008; Theiner, Allen & Goldstone, 2010; Wegner, 1986).

As proponents of distributed cognition have argued, emergent cognitive properties of this kind cannot be explained from a reductionist perspective which ignores the mode of material, social and temporal organization of the components (i.e. interacting people and their cognitive artifacts) of those dynamic cognitive systems whose emergent properties they are (e.g. Clark, 1997; Hutchins, 1995). Hutchins’s 1995 study, however, nicely exemplifies how the initial emergence and the reproduction of the cognitive capacities of the navigation team can be explained in terms of the properties of its human members, their cognitive artifacts and socio-culturally coordinated interactions in structured environments – occasionally he also refers to some aspects of the larger military culture in the US navy and its social structure. It is worth stressing that, in many sociologically interesting cases, power and authority relations between two or more individuals acting in different roles (or positions) make a difference to the outcomes of the distributed cognitive processes. Hutchins (1995: 14–17), for example, discusses how the positions in military rank constitute relations of authority and military identities that affect the distributed cognitive processes (e.g. the chains of command and the instruction of neophytes) among the members of the navigation team. For these reasons, advocates of distributed cognition conceive the cognitive properties of distributed cognitive systems as legitimate objects of study and explanation (see Clark, 1997; Hutchins, 1995; Salomon, 1993; Theiner, Allen & Goldstone, 2010). In other words, their understanding of the nature of emergent cognitive properties is primarily ontological, in Wimsatt’s (2007) sense, although they are also interested in epistemological implications of the existence of emergent properties of this kind.

In addition to broadening the unit of analysis in studies on action-related cognition, the perspective of distributed cognition has highlighted the historical tendency to reduce our information-processing load by externalizing computational processes and experiences into our cognitive artifacts and culturally evolved environments. To take a simple example, we tend to use a pencil and paper (or a calculator if possible) when faced with an arithmetical problem, such as multiplying 137 by 243, that exceeds our mental calculation skills. In this case, it can be said that a person plus the pencil and paper used by him/her compose an extended cognitive system⁵ that is capable of solving the arithmetical problem. Hence, when switching from mental calculation to calculation on paper, we transform the difficult cognitive problem into a series of simpler operations that are quite easy to carry out. This, in turn, is enabled by the symbolic system of arithmetic that we use in our arithmetical calculations, which is itself a product of cultural evolution. It has been argued accordingly that many of the culturally produced cognitive artifacts (e.g. pictures, diagrams, texts, computers, models, simulations, maps, symbolic systems) and artificial environments (e.g. navigation bridges, aircraft cockpits, emergency clinics, computerized offices, organizations) similarly transform our cognitive tasks and facilitate their implementation (for other examples, see Clark, 1997, 1998; Donald, 2001; Hutchins 1995; Nersessian, 2009). Moreover, not only do we utilize cognitive artifacts in our online cognitive processing, but we also tend to reduce our memory load into them (see e.g. Clark, 1997; Clark & Chalmers, 1998; Donald, 2001; Hutchins, 1995). Handwritten notes, written documents, pictures, books, maps, cellphones and databases function as important storages of information in many of our everyday activities. The emergence of some of these cognitive artifacts, such as written language, printed books, maps and newspapers, has required the accumulation of inventions and experiences over many generations. Arguably, the diffusion and uses of some cognitive artifacts have also affected significantly the ways in which people think, interact and perceive some aspects of reality (see Donald, 2001, chap. 7; Hutchins, 1995). Besides extending and modifying our cognitive capacities, the evolution of new cognitive artifacts has also contributed to the emergence of new kinds of psychological and social problems related, for example, to information overload, inequalities in access to information and massive surveillance of individuals’ activities by means of information technology.

From the point of view of the social sciences, the most interesting dimension of distributed cognition is nevertheless socially distributed cognition. This phenomenon is not entirely new to social scientists due to the fact that attributions of cognitive properties to collectives have been common in sociology and other social sciences since Emile Durkheim’s day. Notions such as ‘collective representation’, ‘collective consciousness’, ‘ideology’, ‘collective frame’, ‘myth’, ‘symbol system’ and ‘collective memory’ are often invoked in sociological discussions (e.g. Benford & Snow, 2000; Berger & Luckmann, 1966; DiMaggio, 1997; Misztal, 2003; Strydom, 2007; Wegner, 1986; Wilson 2005). Many social scientists also talk about collective (or corporate) actors, such as integrated groups and organizations, and attribute cognitive states (e.g. beliefs, intentions, purposes and goals) as well as cognitive capacities (e.g. capacity to form collective goals and to make collective decisions) to them. Many of the traditional uses of these terms have nevertheless been either metaphorical or have presupposed an untenable holistic social ontology. In the latter case, the nature of the group-level cognitive phenomena and their relation to the cognition of individuals has remained notoriously obscure. However, it has been recently suggested that the perspective of distributed cognition can be employed not only to clarify and explain the group-level cognitive phenomena but also to rebut many of the traditional arguments that have been raised against the attributions of cognitive properties to collectives (see Hutchins, 1995; Smith, 2008; Sutton, 2008; Sutton et al., 2010; Theiner, Allen & Goldstone, 2010; Tollefsen, 2006). Here I deal with one common objection that comprehends all cognitive phenomena in individualistic terms and rejects attributions of cognitive properties to collectives on the pretext that collectives cannot have consciousness, self-awareness and will of their own (e.g. Elster, 1989: 154–158).

This objection can be confronted by noting that the attribution of ontologically emergent cognitive capacities to distributed cognitive systems does not mean that these systems and their individual members must display an identical set of cognitive properties (Theiner, Allen & Goldstone, 2010: 383). In order to see this, we can employ a slightly modified version of the parity principle introduced by Andy Clark and David Chalmers (1998: 8). This socialized version of the principle presented by Georg Theiner et al. (2010: 384) says that ‘[i]f, in confronting some task, a group collectively functions in a process which, were it done in the head, would be accepted as a cognitive process, then that group is performing that cognitive process’. Now, given this functionalist characterization of cognitive processes and their respective capacities, the acceptance of the existence of the emergent cognitive capacities of collectives (e.g. social groups and organizations) does not automatically commit us to the existence of collective consciousness or collective self-awareness or collective will. In order to argue for the notion of emergent group cognition, we must only point out that some group-level capacities closely resemble (or are analogical to) certain cognitive capacities of individuals.

Furthermore, in many cases cognitive capacities, such as individual memory and collective memory, have the same name both at the individual and the systemic levels. Although these kinds of cognitive capacities of individuals and collectives can be referred to by using the same abstract term ‘memory’, they need not be conceived of as functionally identical. This is because their specific roles in human action and their underlying causal mechanisms may be different. For example, collective memories of an organization may be publicly represented and stored by means of written documents and can thus be directly contested, whereas memories stored in the brain of an individual are private as long as he/she does not communicate them to others (this is not to deny that our private memories, too, tend to change as we interact with other people). Even though they partially overlap each other (since individuals are parts of organizations), the underlying causal processes of the emergence and maintenance of a collective memory of an organization differ from the underlying neuropsychological mechanisms of the biological memories of its individual members. The former may include processes of commemoration, collaborated recall, ritual, founding and maintenance of an archive, writing and telling narratives about the history of the organization; whilst the latter may be described to a large extent in terms of strengthening of the synaptic connections caused by new experiences and activation of the relevant brain areas when triggered by suitable environmental stimuli. Still, the exercise of both of these cognitive capacities performs similar cognitive tasks and produces similar cognitive outcomes. Hence, despite their difference in functional organization at the more concrete level, there is an abstract level of functional description that applies to both of them: both can be described in terms of ‘storing’, ‘transforming’ and ‘retrieving’ information. For these reasons the word ‘memory’ appears to apply to both at the individual and the collective levels (for discussions of the complex relations between individual and collective memory as well as of the related empirical research, see Sutton, 2008; Sutton et al., 2010).

I believe that these brief and somewhat simplified remarks already show how the nature of collective-level cognitive properties may be further conceptually clarified from the perspective of distributed cognition without invoking the dubious notions of collective consciousness, collective self-awareness or collective will. Theiner, Allen & Goldstone (2010: 379) also propose wisely that, ‘[b]ecause of the heterogeneity among different cognitive and mental predicates … abstract arguments about group minds or extended minds should be replaced by specific discussions tied to particular properties: group memory, group problem solving, etc.’ I come back to the notion of collective decision-making in organizations in the next section.

Finally, advocates of distributed cognition emphasize that human cognition is not only distributed but also embodied, in the sense that our bodily structures and processes as well as our sensori-motor interactions with our environments are constitutive of many cognitive abilities, skills and habits that we have (e.g. Clark, 1997, 2008; Hutchins, 1995, 2008; Smith, 2008). Even if they admit the crucial role of bodily (including neural) processes in all cognitive phenomena, the theories of distributed cognition and extended mind are critical of ‘brainbound’ (Clark, 2008) views on human cognition, which see the brain as the only site of cognition. The problem with the brainbound perspective is that it does not do justice to our active bodily relationship to our environment nor to the role of cognitive artifacts and other people in many of our cognitive processes (Clark, 2008; Hutchins, 1995). It has been accordingly suggested that cumulative cultural evolution must be taken into account in theories on the biological evolution of human cognition, because our ancestors’ cultural inventions have changed the selective landscapes in the course of the biological evolution of human brains and cognition (e.g. Clark, 2008; Donald, 2001; Hutchins, 2008; Sterelny, 2003; Tomasello, 1999). From this perspective, humans are said to be a largely self-made species, whose members continually construct their cognitive niches by inventing new cognitive artifacts (including languages and new social arrangements) and epistemically engineering their environments (Sterelny, 2003, 2010). Advocates of distributed cognition thus suggest that our brain processes, too, should be investigated by relating them to their larger bodily and socio-material environment (e.g. Clark, 1997; 2008, chap. 7; Donald, 2001).

Evaluating the background assumptions

After this brief overview of the perspective of distributed cognition, I move on to consider the six background assumptions of the DBO-theory that were introduced above.

1 From the perspective of explanatory social research, interacting individuals should be conceived of as the only causally efficacious actors in society

It can be argued that the structural individualism advocated by Hedström (2005; Hedström & Bearman, 2009) and many other analytical sociologists is not an entirely coherent position. This is because it accepts ‘macro-level facts’, such as the structures of social networks and norms, as explanatory factors in social scientific research, but denies the existence of causal powers of social entities. For example, Hedström & Bearman (2009: 13) write that ‘[i]n order to understand collective dynamics we must study collectivity as a whole, but we must not study it as a collective entity’ since ‘[o]nly by taking into account the individual entities [i.e. individual human beings, in the case of realist models about collective dynamics], and most critically the relations between them and their activities, can we understand the macro-structure we observe’. Hence, they want to take properties of collectivities into account in explanations of macro-phenomena, but do not admit that this requires postulation of collective entities whose properties are referred to in these explanations. In contrast to this view, a number of authors have pointed out that, by ascribing the explanatory role to network structures and norms, one implicitly presupposes the existence of causally efficacious social entities whose ontologically emergent causal powers cannot be eliminated or reduced to the aggregates of the intrinsic causal powers of individual actors (e.g. Elder-Vass, 2010: 24–26; Kaidesoja, 2012; Wan, 2011: 166–169). Using the term coined by Mario Bunge (1996: 254), the position of structural individualism may thus be dubbed ‘individholistic’, which means that it is an instance of individualism with a hidden holistic component that is left unanalyzed. As Bunge (1996: 255) writes, individholism is ‘an inconsistent stand halfway between individualism and systemism’, which can be transcended only by adopting an emergentist social ontology combined with a multi-level approach in methodology (see also Kaidesoja, 2012; Wan, 2011). While I agree with this argument, here I pursue a different though parallel line of argument.

In what follows, I try to show that the previous assumption is too restricted since, in addition to interacting individuals, functionally organized collectives (e.g. integrated social groups and organizations), too, can be understood as causally efficacious actors in their own right. For example, organizations – such as business firms, political parties, labor unions and universities – often have the following features that make them plausible candidates for collective actors:

– Organizations are typically designed to accomplish some more or less specific purpose(s) or goal(s), such as pursuit of profit, propagation and implementation of an ideology or advancement of the shared interests of its members, and organizational planning is the core mechanism in the workings of organizations. In order to endure, organizations also have to modify their goals, plans and policies to meet the new requirements of the changing environment.

– Members of organizations usually perform quite strictly defined roles that often involve division of cognitive labor and significant authority relations. People accordingly acquire new obligations, tasks, rights and duties from the moment they start working in the organization.

– Incentive systems may also be introduced into organizations in order to promote such actions, which enhance the achievement of the collectively defined goals and plans.

– Organizations tend to make decisions in a more or less centralized fashion by means of specific procedures and practices.

– Organizations tend to acquire, store and process information that is relevant for the achievement of their goals.

– Organizations may form collective identities, policies and cultures, which remain relatively unchanged even though their individual members change.

– Organizations are commonly held to be morally and legally responsible actors to whom we attribute rights, duties and responsibilities.

The details of these abstractly described features of organizations have been widely discussed in the fields of organization theory and management. For the purposes of the present article, it is enough to point out that these properties are often ascribed to organizations both in everyday discussions and in organization studies. This is not to deny, however, that there are significant differences between different kinds of organizations. I also admit that not all organizations exhibit all of these features, and that the emergence and maintenance of them is dependent on the coordinated interactions of their members.

Now, if it is accepted that organizations often exhibit the previous features – or at least most of them – then it can be further argued that organizations possess emergent cognitive powers that cannot be analyzed as aggregates of the cognitive properties of their individual members or aggregates of their intentional cognitive actions. I suggest that it is precisely the emergent cognitive capacities of this kind that justify ascriptions to intentional agency to organizations, understood as functionally organized wholes, and necessitate studies that conceptualize organizations as collective actors. To make my case, I focus on the capacity of organizations to make collective decisions. I chose this feature of organizations for two reasons: it is crucially important for ascriptions of intentional agency to them; and it can be fruitfully analyzed from the perspective of distributed cognition.

Collective decisions in an organization are usually made by utilizing the complementary skills and expertise of its members who participate in the collaborative decision-making processes (e.g. Secchi, 2010). The size of the circle of those members who participate in decision-making may vary from one (e.g. dictatorship) to all members (e.g. participatory democracy) – in most cases it lies somewhere in between these poles. Members of the organization also typically exploit different kinds of cognitive artifacts as representational media that enable information sharing in their decision-making practices. Examples range from memos, reports, statistics and cost–benefit analysis to scenarios of the possible futures of the organization. Furthermore, the members engaged in collective decision-making usually enact quite strictly defined roles that involve authority relations. They may also function as representatives of those members of the organization who are excluded from the process of organizational-level decision-making.

These points are well taken by Davide Secchi (2010), who argues that decision-making in organizations is best understood by combining the perspective of distributed cognition and Herbert Simon’s notion of docility. The latter means the tendency of the people involved ‘to depend on suggestions, recommendations, persuasion, and information obtained through social channels as a major basis of choice’ (Simon, 1993: 156). The docility of the collaborative decision-makers is exemplified by their activity of taking advice from the other members of the organization as well as from the external consultants and stakeholders, whereas the docility of an organization is embodied in the structural arrangements that support the activities of advice-giving and advice-taking (Secchi, 2010: chaps 8–9). Hence, due to its materially and socially distributed nature, it is unlikely that decision-making in organizations can be adequately modeled in terms of aggregation of decisions of individuals. In some special cases, such as decisions made by means of majority voting, this sort of reductionist perspective may nevertheless be sufficient.

The previous points can be further supported by reference to the so-called discursive dilemma and some other impossibility results obtained in recent discussions on judgment aggregation. These findings suggest that social groups engaged in rational decision-making on logically connected issues (e.g. political parties, expert panels or court juries) cannot exclusively rely on the aggregation procedures of their members’ decisions (e.g. majority voting) because different ways of implementing these procedures tend to produce inconsistent results (List & Pettit, 2002; Pettit, 2004; Spiekermann, 2010). Unfortunately, the space limitations of the present article do not permit a detailed review of this argument. On the basis of these results, Philip Pettit (2004: 182) nevertheless writes that ‘it is reasonable, even compulsory, to think of the integrated collectivity as an intentional subject’, and continues:

The basis of this claim is that the integrated collectivity … is going to display all the functional marks of an intentional subject and that there is no reason to discount those marks as mere appearances. Within relevant domains it will generally act in a manner that is rationalized by independently discernible representations and goals; and within relevant domains it will generally form and unform those representations in a manner that is rationalized by the evidence that we take to be at its disposal. In particular, it will manifest this sort of functional organization, not just one time, but over time; it will display a degree of constancy as well as the degree of coherence that we expect in any intentional subject. (Pettit, 2004: 182)

Although I doubt that integrated collectives of this kind display ‘all the functional marks of an intentional subject’, I nevertheless agree that they exhibit sufficiently many of them to call them intentional actors. Kai Spiekermann (2010: 410–411) also argues that the above-mentioned impossibility results imply that groups and organizations engaged in collective decision-making on logically related issues tend to form ontologically emergent (in William Wimsatt’s sense of the term) cognitive capacities that enable them to make collective decisions and exhibit collective rationality over significant periods of time.

These arguments may nevertheless be contested by noting that these sorts of cognitive capacities on the part of groups and organizations tend to be fragile and to break down quickly because individual members of organizations usually care more about their own interest and desires than those of their organizations. This counterclaim can be answered by noting that it presupposes an egoistic view of human nature that can be challenged on the basis of empirical research and everyday experience. Recent studies on the ontogeny and evolution of human cognition indicate that human beings have cognitive, motivational and emotional predispositions not only to identify themselves with their groups but also to recognize and adapt their behavior according to the norms, complementary roles and shared intentions that prevail in these groups (e.g. Tomasello, 2009). Of course, these predispositions may be either enforced or inhibited by the parents and larger socio-cultural environment of an individual during his/her psychological development, but they nevertheless appear to have a basis in the biological make-up of our species. Michael Tomasello (2009), drawing on the results of these studies, further argues that insofar as people identify themselves with a group, they both are cognitively capable of recognizing the group’s goals (as distinct from their individual goals) and tend to promote the group goals by participating in the collaborative activities even if their actions are in conflict with their individual self-interests. I suggest that these kinds of cognitive features of individuals and the cognitive skills they acquire when participating in collaborative interactions enable the formation and maintenance of the emergent capacity of organizations to make collective decisions. These points are not meant to deny that some individuals may try to advance their private interest by using integrated groups and organizations as instruments. The above remarks nevertheless do suggest that, in contrast to the views of many methodological individualists, situations where the self-interests of the group members override their group commitments and obligations appear to be rarer than is assumed in the above counter-argument. Furthermore, in the case of organizations, these situations can be counteracted by introducing various incentive systems.

So far I have argued that there are good reasons to conceive of organizations and perhaps some other integrated groups as causally efficacious intentional actors. Now I would like to suggest that there are accordingly cases where it is necessary to explain macro-level social phenomena in terms of organizations and other collectives that interact with each other in certain contexts – without reducing group actions to those of individuals. This view can be supported by considering Arthur Stinchcombe’s (1998) theoretical model of the mechanism of monopolistic competition between corporate groups (i.e. corporations, universities and nation-states) located in certain competitive institutionalized fields (i.e. markets, prestige systems and world power system) that involve field-specific rankings in terms of profits, prestige or power. By the term ‘monopolist competition’ Stinchcombe (1998: 268) refers to ‘competition in which each firm or other corporate group (at least, each corporate group toward the top in profits, prestige, or power) delivers a unique product or dominates the unique territory, but there are more or less close substitutes that make them subject to competitive pressure’. His mechanism model then aims to explain the empirically established phenomenon of ‘autocorrelation of profits, prestige, and power of highly ranked corporate groups’ (1998: 267) in these competitive fields. That is, it seeks to answer the question why some corporate groups (e.g. the University of Harvard) tend to maintain their high status and related monopoly advantages over significant periods of time in relation to the other groups with whom they compete (e.g. the other US universities). The key feature of the general explanatory mechanism proposed by the model concerns the competence of corporate groups to appropriate their ‘going-concern value’ (John R Commons) by monopolizing the field-specific opportunities. High-status groups are thus able to turn their field-specific liberties into exploitable opportunities more effectively than the competing groups. Stinchcombe (1998: 286) writes that ‘[a] central feature enabling organizations to outperform alternative actors in their fields … is that they can act corporately faster than consensus can form’. This suggests that high-status organizations must develop collective decision-making practices and routines that are different from the time-consuming procedures that aggregate individual decisions. In this model, a corporate group is thus understood as ‘a network that can work together, and so can do together the activities that bring in benefit from the opportunities that it exploits’ (1998: 281). In order to do this, it has to have ‘a common understanding of “the collective welfare”’ (1998: 286) and a related policy as well as internal information flows, resources and incentive systems that are relevant for this policy (1998: 286–287). In other words, Stinchcombe (e.g. 1998: 296) explicitly assumes that corporate groups are unified causal actors who exploit their benefits or maximize their utilities (e.g. profit, reputation and national power) as collectives in their respective competitive fields. Whether or not the details of this explanatory model are correct, it can be concluded that it is very hard to see how ‘autocorrelation of profits, prestige, and power of highly ranked corporate groups’ (1998: 267) can be explained if one assumes that interacting individuals are the only explanatorily relevant causal actors (for similar arguments, see Mayntz, 2004; Wan, 2011, chap. 7).

To sum up: I argued that collectives, especially organizations, can be fruitfully conceptualized as distributed cognitive systems composed of human members and the artifacts used by them. I further pointed out that these systems can be said to possess ontologically emergent (i.e. non-aggregative) cognitive powers (e.g. capacities to make collective decisions, to form and implement collective goals and plans, to maintain a collective memory and to solve cognitive problems). I also suggested that it is precisely capacities of this kind that justify the ascriptions of agency to collectives in explanations of social macro-phenomena. Emergent cognitive capacities should not, however, be attributed to collectives uncritically. On the contrary, we should also analyze the underlying mechanisms of these capacities in terms of interacting individuals (including their beliefs, desires, abilities and skills); cognitive artifacts (including written codes and documents) used by them in the coordination of their actions and interactions; the roles enacted by them; and the institutionalized norms followed and reproduced by them (see Clark, 1997: 186–192). Indeed, far from taking the emergent cognitive properties of collectives as black boxes, proponents of distributed cognition have carefully examined many of their underlying mechanisms in their empirical studies on distributed cognitive systems (e.g. Cohen et al., 2006; Hutchins, 1995; Theiner, Allen & Goldstone, 2010). It is nevertheless important to stress that explanations of the emergent cognitive powers of collectives in terms of their underlying mechanisms do not explain these powers away – these sorts of explanations are not eliminative reductions. On the contrary, once we have acquired explanatory understanding of how the emergent cognitive capacities of collectives emerge and function in terms of their underlying mechanisms, these capacities acquire a more robust ontological status, and references to them in explanations of social macro-phenomena become accordingly justified. Though I have focused on the cognitive capacities of collectives that constitute their agency, my intention is not to deny that in some cases distributed cognitive processes in social groups may turn out to be less efficient than the aggregates of individual cognitive processes. Situations of this kind are analyzed in studies on collective biases (e.g. Smith, 2008) and groupthink (e.g. Turner & Pratkanis, 1989).

2 For the purposes of explanatory social research, agency (capacity for intentional action) of individual actors can be taken as a black box

Not all explanatory tasks in social sciences require a detailed analysis of the mechanisms that underlie the agency of individuals. For example, in many survey studies on social macro-phenomena, such as distributions and changes of attitudes in a certain population of people, questions regarding the underlying mechanisms of individual agency can usually be safely ignored. But there are other explanatory tasks in social research where it is necessary to consider whether and how individual actors are capable of acting intentionally. For example, if you are trying to find out why some senior citizens manage to get by without help in their homes while others cannot, the issue of agency and its underlying mechanisms are of central importance. In studies of this kind, the perspective of distributed cognition is likely to prove useful because, as indicated above, the research on distributed (and embodied) cognition has shown that many of the cognitive capacities and abilities underlying our agency are context dependent. This means that not only our brain structures and bodily abilities but also our cognitive artifacts and socio-material environments are constitutive of our agency (e.g. Clark, 1997; Lizardo & Strand, 2009). To continue the above example, it may be assumed that elderly couples have formed over time a distributed (or transactive) memory system that enables them to rely on each other’s complementary expertise in order to perform various intentional actions (see Wegner, Raymond & Erber, 1991; also Wegner, 1986). Hence, it can be hypothesized that widows and widowers tend to suffer from the disruption of the distributed memory system of this kind, which may have detrimental effects on their agency because they may suddenly lack some of the beliefs and skills that are necessary for living in modern society.

Furthermore, as Tomasello (1999, 2009) argues on the basis of several empirical studies, our abilities to act intentionally, to understand the intentions of others and to act as members of groups depend on both our inherited biological predispositions (e.g. predispositions to imitate others and share attention with them) and our communicative interactions that take place in culturally produced environments during our cognitive development. For these reasons, he contends that human agency is best understood as a both biological and socio-cultural phenomenon at the same time. In my view, Tomasello’s and others’ studies on the development and nature of the cognitive mechanisms underlying human agency should not be ignored in explanatory social research insofar as we want to base our explanations on the concepts of empirically grounded theories of action.

3 Desires and beliefs are best understood as causally efficacious mental events and states of individuals

Studies on distributed, situated and embodied cognition in everyday contexts have shown that an environmentally sensitive (or extended) concept of belief is needed in order to understand many of our everyday actions due to the fact that our beliefs are often more or less tied to the specific social and material environments in which we act (e.g. Clark, 1997; Hutchins, 1995; see also Lizardo & Strand, 2009). In this view, the beliefs of individuals should not be regarded as abstract propositional contents of their mental states but rather as action-related representational relations between embodied individuals and their environments. It has also been suggested that in actions that take place in socio-materially structured environments we often rely on ‘external’ resources (including various representational media and ‘information storages’) rather than on detailed ‘internal’ representations of our environments (e.g. Clark, 1997; Hutchins, 1995). This is not to deny, however, that we are also capable of forming such ‘internal’ representations which are decoupled from any specific type of action or behavioral response and which may thus influence many types of actions (Sterelny, 2003). Recent inquiries on the phenomenon of contextual priming nevertheless indicate that the activation of our ‘internal’ representations too is often partially dependent on the contextual cues. John Bargh and Ezequiel Morsella, for example, write that ‘[i]n contextual priming, the mere presence of certain events and people automatically activates our representations of them, and concomitantly, all of the internal information (goals, knowledge, affect) stored in those representations that is relevant to responding back’ (Bargh & Morsella, 2008: 76). Neuroscientific evidence suggests that these kinds of priming effects depend on the functioning of the so-called mirror neurons in the premotor cortex in our brains (Bargh & Morsella, 2008: 76). Furthermore, in cases where a belief or a memory – in order to avoid confusion, the terms ‘ideology’ and ‘collective memory’ may be more appropriate here – is distributed across a functionally organized social group, it can be best conceived of as an emergent property of the group rather than an aggregate of the properties of its individual members (Theiner, Allen & Goldstone, 2010; see also Tollefsen, 2006).

The concept of desire, too, needs reconsideration from the perspective of distributed cognition because not all desires can be adequately conceptualized as mental events or states of individuals. It can be first of all doubted, in the light of new cognitive scientific research, whether desires and preferences of individuals are mental states at all (e.g. Clark, 1997; Freese, 2009). It may, for example, be more fitting to conceptualize desires as embodied dispositions to act rather than as mental states. Furthermore, as was noted above, there are good reasons to think that organizations and other integrated groups are able to form collective desires (or goals) that are distinct from the aggregates of the desires of their individual members. Jeremy Freese (2009, 104–105; also Satz & Farejohn, 1994) even suggests that organizations may be able to act on their collective preferences more coherently and consistently than can individual actors.

The above considerations indicate the need to reconsider the definitions of the concepts of belief and desire in terms of disembodied mental states and events of individuals. I am not denying that in some action-related, embodied and context-sensitive sense, individuals have causally efficacious beliefs and desires, but I want to suggest that so do integrated groups and organizations. I also admit that interacting individuals play a crucial role in forming, sustaining and transforming the beliefs and desires of collective actors, but I insist, at the same time, that the beliefs and desires of collectives cannot be reduced to the aggregates of desires and beliefs of individuals. Hedström and some other analytical sociologists are nevertheless right to emphasize the causal interactions between individuals’ desires, beliefs and opportunities as well as the influence of others on them. From this perspective, the concepts of belief and desire in the DBO theory may already be said to contain some action-related and context-sensitive elements.

4 Actors think before they act

There surely are circumstances where our actions can be adequately modeled by dissecting them into the decision–action sequences. As pragmatists (see Gross, 2009; Joas 1996; Kilpinen, 2000) in particular argue, many of our everyday doings nevertheless fall outside this description. For example, episodes of habitual and routine-like action in relatively stable environments (e.g. riding from home to work on a bicycle using a familiar route) do not involve sequences of temporal phases of decision-making immediately followed by bodily actions (for other examples, see Noë, 2009, chap. 5; Wood, Kashy & Quinn, 2002). Thanks to our habits, we are free to focus our conscious thoughts on other things (than our immediate action) during habitual actions. Sometimes our deeply entrenched habits may also hinder our ability to react to unexpected changes in our environment in an appropriate way (Wood, Kashy & Quinn, 2002). On the basis of considerations of this kind, Neil Gross (2009: 368) goes so far as to suggest that social mechanisms can best be understood in terms of chains of problem-situations and actors’ habitual responses to them. This is probably an overstatement since some social mechanisms, such as the joint planning and implementation of plans, surely involve conscious decision-making that cannot be described solely in terms of ‘habitual responses’. Moreover, even though I argued above that organizations and other integrated collectives can be understood as intentional actors, collective-level routines (e.g. Stinchcombe, 1998: 285–286) and collectively enacted habits (e.g. Gross, 2009: 370–371) that do not necessarily involve decision-making appear to be common in them as well. The previous assumption of the DBO theory, however, excludes habits and routines from the set of potential explanatory factors in mechanism-based explanations.

It is also worth noting that many choice situations that are faced by individuals or corporate actors can be characterized as highly scaffolded, in the sense that they are heavily constrained by policies, infrastructure, social structures and institutional practices (Clark, 1997: 181–182; Satz & Farejohn, 1994). As Debra Satz and John Farejohn (1994) argue, the predictions of rational-choice theories about economic phenomena tend to be successful precisely in these kinds of environments, while failing miserably in other kinds of situations. Hence, the restricted predictive success of the rational-choice theories, they suggest, is explained by the structure of the environment rather than by the highly unrealistic assumptions about the nature of human agents made in these theories. In contrast to the rational-choice approach, the perspective of distributed cognition also suggests that the concept of human rationality has more to do with the ways we adapt to our changing environments by interacting with each other and exploiting our culturally evolved cognitive artifacts than with our internal calculative capacities (Clark, 1997, chap. 9; Secchi, 2010).

To conclude: insofar as it is taken as the general characterization of the nature of human action, the above assumption depicts action in an overly intellectualistic and individualistic way. For this reason, it is likely to lead to problematic biases in explanations of many social macro-phenomena.

5 In explanations of the actions of an individual, his/her conscious intentional mental events have the explanatory primacy over the other causes of action

As mentioned above, much of our individual action consists of environmentally embedded habits and routines that do not involve conscious choices, deliberation or decision-making. This is not to deny, however, that we are also able to consciously reflect on, deliberate on and transform our habits and routines, especially in situations where our habitual actions meet obstacles (Joas, 1996; Kilpinen, 2000). In any case, habitual types of action are not best explained by reference to the conscious intentions of the actors despite the fact that intentions may sometimes be part of their explanation. Recent studies on action-related social cognition also suggest that we have good reasons to count unconscious cognitive processes (in the sense that individual actors are not aware of these processes) among the causes of social interactions between individuals (for a review, see Bargh & Morsella, 2008). For example, people tend to imitate the actions (e.g. gestures) of strangers with whom they interact without being consciously aware of their actions. Several studies indicate that unconscious imitation of this kind functions as ‘a social glue’ between individuals by enhancing liking and bonding between them (Bargh & Morsella, 2008: 76).⁶ Experimental studies on touch in casual interpersonal interaction also suggest that touch unconsciously increases positive feelings and actions (e.g. helping) towards the toucher as well as the other members in his/her social group, but this effect seems to depend on initially positive feelings towards the toucher and the situation (Smith, 2008: 26–27). In addition to these examples, the role of unconscious, fast and automatic cognitive processes and their interplay with slow, deliberative and conscious cognitive processes is a growing topic of interest in recent dual-process theories on reasoning, judgment and social cognition (for a review, see Evans, 2008).

A proponent of the DBO theory might object that the previous processes are not instances of social action, which is the object of sociological inquiry. Since they happen at the unconscious level, they should be counted instead as mere social behavior. My answer to this is that proponents of DBO theory should reconsider their definition of the concept of social action since the recognition of unconscious cognitive processes may have far-reaching consequences for the sociological explanations of many social macro-phenomena (see Vaisey, 2008). For example, the previously described unconscious cognitive processes may well participate in the reproduction and strengthening of the segregation of ethnic groups in a society where different ethnic groups are concentrated in different residential areas. It is also important to note that not all of the explanations that refer to different-level mechanisms compete with each other but may also be complementary. For these reasons, we should remain open to the possible explanatory role of unconscious (and unintentional) cognitive processes when explaining outcomes of individual and social actions.

6 Collaborative learning can be ignored in action-based explanations

A number of empirical studies (e.g. Dillenbourg, 1999; Hutchins, 1995: chaps 6–8; Salomon, 1993) demonstrate that many learning processes in distributed cognitive systems are instances of collaborative learning. As mentioned above, these learning processes cannot be adequately understood by using the concepts of DBO theory. They may also be among the important explanatory factors, for example, in explanations of the differences in the performances between organizations (e.g. firms) competing with each other in some institutional context (e.g. markets). It can accordingly be hypothesized that the competence of corporate groups (or organizations) to maintain their high status and monopoly power in relation to the competing groups in Stinchcombe’s (1998) model of the monopolist competition is partly due to the effectiveness of the organizational learning processes that take place in these groups. In addition, collaborative learning should also be taken into account in explanations of the emergence and maintenance of many other cognitive capacities of collectives, such as work teams and school classes (see Dillenbourg, 1999; Hutchins, 1995; Salomon, 1993). So I think that the previous assumption is not sound.

Conclusion

The above arguments question the realisticness of the background assumptions of the DBO theory. They do not, however, imply that this theory is completely useless. On the contrary, the DBO theory may still be considered useful in studies where it is methodologically necessary to adopt unrealistic (and/or idealizing) assumptions, and where these assumptions do not exclude potentially crucial explanatory factors. What the above critique questions is the viability of the DBO theory to function as a general framework for building mechanism-based explanatory theories and models in sociology. As I have argued above, actors’ uses of cognitive artifacts, their materially mediated social interactions, unconscious cognitive processes and group-level cognitive capacities and processes – all of which are more or less neglected in the DBO theory – make a crucial difference to the outcomes of many social processes.

Perhaps some of the critical points presented above can be accommodated into the DBO theory by modifying its background assumptions (see Manzo, 2010). It may, for example, be possible to interpret collectives as causally efficacious actors by developing a new variety of DBO theory in which assumptions (1), (3) and (6) are modified. I hope to have shown, however, that Hedström’s version of the DBO theory lacks the conceptual resources that are needed for making sense of the action-related distributed cognitive processes, habitual action, the interplay between conscious and unconscious cognitive processes in social interaction, and the relations between individual and collective agency. I also fail to see how these phenomena might be conceptualized in terms of the modified versions of the DBO theory.

Hence, instead of modifying the DBO theory, I venture to suggest that our explanatory aspirations would be better served if we developed many partial and complementary theories of action that were each useful for a limited set of methodological purposes. In addition to action theories that focus on various aspects of individual action and interaction (e.g. rational-choice theories, DBO theory, individual learning theories, pragmatist theory of habitual action), we also need theories that focus on practical activities in which individuals participate (e.g. practice theory, activity theory) as well as theories that address the formation, maintenance, actions and interactions of collectives. If we accept this sort of action-theoretical pluralism, which is grounded in the multilevel nature of human action, we should develop many complementary (and partially overlapping) action theories for specific explanatory purposes, examine their limits of applicability and criticize all attempts to universalize a particular action theory.

Footnotes

Acknowledgements

I would like to thank Antti Gronow, Erkki Kilpinen, John Latsis, Petri Ylikoski and participants in the Philosophy of Science seminar (organized by the TINT-group) in Helsinki for their useful comments.

Funding

During the writing of this article I received financial support from the Kone Foundation, for which I am grateful.

Notes

Author biography

Tuukka Kaidesoja is a post-doctoral researcher in the Philosophy of Social Sciences at the Finnish Academy Centre of Excellence, University of Helsinki. His research interests include analytical sociology, critical realism, distributed cognition, and systemic social ontology and methodology.

References

Bargh

Morsella

(2008) The unconscious mind. Perspectives on Psychological Science 3(1): 73–79.

Benford

Snow

(2000) Framing processes and social movements: An overview and assessment. Annual Review of Sociology 26: 611–639.

Berger

Luckmann

(1966) The Social Construction of Reality: A treatise in the sociology of knowledge. Garden City, NY: Doubleday.

Boudon

(1998) Social mechanisms without black boxes. In: Hedström

Swedberg

(eds) Social Mechanisms: An analytical approach to social theory. Cambridge: Cambridge University Press, 172–203.

Bunge

(1996) Finding Philosophy in Social Science. New Haven, CT/London: Yale University Press.

Clark

(1997) Being There: Putting brain, body, and world together again. Cambridge, MA: MIT Press.

Clark

(1998) Embodied, situated, and distributed cognition. In: Bechtel

Graham

(eds) A Companion to Cognitive Science. Malden/Oxford/Melbourne: Blackwell Publishing, 506–517.

Clark

(2008) Supersizing the Mind: Embodiment, action, and cognitive extension. Oxford/New York, NY: Oxford University Press.

Clark

Chalmers

(1998) The extended mind. Analysis 58: 10–23.

10.

Cohen

Blatter

Almeida

Shortliffe

. (2006) A cognitive blueprint of collaboration in context: Distributed cognition in the psychiatric emergency department. Artificial Intelligence in Medicine 37: 73–83.

11.

Dillenbourg

(ed.) (1999) Collaborative Learning: Cognitive and computational approaches. Oxford: Elsevier Science.

12.

DiMaggio

(1997) Culture and cognition. Annual Review of Sociology 23: 263–287.

13.

Donald

(2001) A Mind So Rare: The evolution of human consciousness. New York, NY/London: WW Norton & Company.

14.

Elder-Vass

(2010) The Causal Power of Social Structures: Emergence, structure and agency. Cambridge: Cambridge University Press.

15.

Elster

(1989) Nuts and Bolts for the Social Sciences. Cambridge: Cambridge University Press.

16.

Evans

(2008) Dual-processing accounts of reasoning, judgment, and social cognition. Annual Review of Psychology 59: 255–278.

17.

Freese

(2009) Preferences. In: Hedström

Bearman

(eds) The Oxford Handbook of Analytical Sociology. Oxford: Oxford University Press, 94–114.

18.

Gross

(2009) A pragmatist theory of social mechanisms. American Sociological Review 74: 358–379.

19.

Hazlehurst

McMullen

Gorman

(2007) Distributed cognition in the heart room: How situation awareness arises from coordinated communications during cardiac surgery. Journal of Biomedical Informatics 40: 539–551.

20.

Hedström

(1994) Contagious collectivities: On the spatial diffusion of Swedish trade unions, 1890–1940. American Journal of Sociology 90(5): 1157–1179.

21.

Hedström

(2005) Dissecting the Social: On the principles of analytical sociology. Cambridge: Cambridge University Press.

22.

Hedström

Bearman

(2009) What is analytical sociology all about? An introductory essay. In: Hedström

Bearman

(eds) The Oxford Handbook of Analytical Sociology. Oxford: Oxford University Press, 3–24.

23.

Hedström

Swedberg

(eds) (1998) Social Mechanisms: An analytical approach to social theory. Cambridge: Cambridge University Press.

24.

Hedström

Ylikoski

(2010) Causal mechanisms in the social sciences. Annual Review of Sociology 36: 49–67.

25.

Hutchins

(1995) Cognition in the Wild. Cambridge, MA/London: MIT Press.

26.

Hutchins

(2008) The role of cultural practices in the emergence of modern human intelligence. Philosophical Transactions of the Royal Society 363: 2011–2019.

27.

Hutchins

Klausen

(1996) Distributed cognition in an airline cockpit. In: Engeström

Middleton

(eds) Cognition and Communication at Work. Cambridge: Cambridge University Press, 15–34.

28.

Joas

(1996) The Creativity of Action. Cambridge/Malden, MA: Polity Press.

29.

Kaidesoja

(2009) Bhaskar and Bunge on social emergence. Journal for the Theory of Social Behaviour 39(2): 300–322.

30.

Kaidesoja

(2012) Naturalizing Critical Realist Social Ontology. London/New York, NY: Routledge. (forthcoming)

31.

Kilpinen

(2000) The Enormous Fly-Wheel of Society: Pragmatism’s habitual conception of action and social theory. (Research reports of the Department of Sociology no. 235, University of Helsinki)

32.

List

Pettit

(2002) Aggregating sets of judgements: An impossibility result. Economics and Philosophy 18(1): 89–110.

33.

Lizardo

Strand

(2009) Skills, toolkits, contexts and institutions: Clarifying the relationship between different approaches to cognition in cultural sociology. Poetics 38: 204–227.

34.

Macy

(1997) Identity, interest and emergent rationality: An evolutionary synthesis. Rationality and Society 9(4): 427–448.

35.

Manzo

(2010) Analytical sociology and its critics. European Journal of Sociology 5(1): 129–170.

36.

Mayntz

(2004) Mechanisms in the analysis of social macro-phenomena. Philosophy of the Social Sciences 34(2): 237–259.

37.

Misztal

(2003) Theories of Social Remembering. Maidenhead/Philadelphia, PA: Open University Press.

38.

Nersessian

(2009) Model-based reasoning in distributed cognitive systems. Philosophy of Science 75: 699–709.

39.

Nersessian

Kurz-Milcke

Newstetter

Davies

(2003) Research laboratories as evolving distributed cognitive systems. In: Alterman

Kirsh

(eds) Proceedings of the Twenty-Fifth Annual Conference of the Cognitive Science Society, Part II. Mahwah, NJ: Lawrence Erlbaum, 857–862.

40.

Noë

(2009) Out of Our Heads: Why you are not your brain, and other lessons from the biology of consciousness. New York, NY: Hill and Wang.

41.

Pettit

(2004) Groups with minds of their own. In: Smith

(ed.) Socializing Metaphysics: The nature of social reality. New York, NY: Rowman and Littlefield Publishers.

42.

Rydgren

(2009) Beliefs. In: Hedström

Bearman

(eds) The Oxford Handbook of Analytical Sociology. Oxford: Oxford University Press, 72–93.

43.

Salomon

(ed.) (1993) Distributed Cognitions: Psychological and educational considerations. Cambridge: Cambridge University Press.

44.

Satz

Farejohn

(1994) Rational choice and social theory. Journal of Philosophy 9102: 71–87.

45.

Secchi

(2010) Extendable Rationality: Understanding decision making in organizations. New York, NY/ Dordrecht/Heidelberg/London: Springer.

46.

Simon

(1993) Altruism and economics. American Economic Review 83(2): 156–161.

47.

Smith

(2008) Social relationships and groups: New insights on embodied and distributed cognition. Cognitive Systems Research 9: 24–32.

48.

Spiekermann

(2010) Judgement aggregation and distributed thinking. AI & Society 25: 401–412.

49.

Sterelny

(2003) Thought in a Hostile World: The evolution of human cognition. Malden, MA/Oxford (UK)/Carlton (Australia): Blackwell Publishing.

50.

Sterelny

(2010) Minds: Extended or scaffolded? Phenomenology and Cognitive Sciences 9(4): 465–481.

51.

Stinchcombe

(1998) Monopolist competition as a mechanism: Corporations, universities, and nation-states in competitive fields. In: Hedström

Swedberg

(eds) Social Mechanisms: An analytical approach to social theory. Cambridge: Cambridge University Press, 267–305.

52.

Strydom

(2007) A cartography of contemporary cognitive social theory. European Journal of Social Theory 10(3): 339–356.

53.

Sutton

(2008) Between individual and collective memory: Coordination, interaction, distribution. Social Research 75(1): 23–48.

54.

Sutton

Harris

Keil

Barnier

(2010) The psychology of memory, extended cognition, and socially distributed remembering. Phenomenology and Cognitive Sciences 9(4): 521–560.

55.

Theiner

Allen

Goldstone

(2010) Recognizing group cognition. Cognitive Systems Research 11: 378–395.

56.

Tollefsen

(2006) From extended mind to group minds. Cognitive Systems Research 7: 140–150.

57.

Tomasello

(1999) The Cultural Origins of Human Cognition. Cambridge, MA/London: Harvard University Press.

58.

Tomasello

(2009) Why We Cooperate. Cambridge, MA/London: MIT Press.

59.

Turner

Pratkanis

(1989) Twenty-five years of groupthink theory and research: Lessons from the evaluation of a theory. Organizational Behavior and Human Decision Processes 73(2–3): 105–115.

60.

Vaisey

(2008) Socrates, Skinner, and Aristotle: Three ways of thinking about culture in action. Sociological Forum 23(3): 603–613.

61.

Wan

Poe Yu-Ze

(2011) Reframing the Social: Emergentist systemism and social theory. Farnham/Burlington, VT: Ashgate.

62.

Wegner

(1986) Transactive memory: A contemporary analysis of the group mind. In: Mullen

Goethals

(eds) Theories of Group Behavior. New York, NY: Springer-Verlag, 185–208.

63.

Wegner

Raymond

Erber

(1991) Transactive memory in close relationships. Journal of Personal and Social Psychology 61(6): 923–929.

64.

Wilson

(2005) Collective memory, group minds, and the extended mind thesis. Cognitive Processing 6: 227–236.

65.

Wimsatt

(2007) Re-Engineering Philosophy for Limited Beings: Piecewise approximations to reality. Cambridge, MA/London: Harvard University Press.

66.

Wood

Kashy

Quinn

(2002) Habits in everyday life: Thought, emotion, and action. Journal of Personality and Social Psychology 6(83): 1281–1297.