What is an artifact?

I begin my analysis by reflecting on the role of artifacts in Routines Theory and how, and to what extent, this has evolved over time. The word ‘artifact’, or ‘artefact’, comes from the Latin words arte (ars, art) and factum (facere, to make). Artifacts are defined as ‘anything made by human art and workmanship’ (Oxford English Dictionary Online), or ‘something created by humans usually for a practical purpose’ (Merriam-Webster Online Dictionary, 1990: 105).

There are probably as many characterizations of the nature, properties, and role of artifacts as there are disciplines (see also Vilnai-Yavetz and Rafaeli, Reference Vilnai-Yavetz, Rafaeli, Rafaeli and Pratt2006). For the purpose of this paper, I find it useful to distinguish among three main perspectives: realist, constructivist, and actor-network theory (ANT).Footnote ¹ The realist argument is that artifacts have stable, inherent properties so that agency can be attributed to them straightforwardly. Realists, for example, have shown how the physical structure or operation of artifacts, such as the design of a technology, constrains or enables practices, beliefs, or social configurations (Mumford, Reference Mumford1964; Winner, Reference Winner1980). This view has been criticized by constructivists as it plays down the role of social factors in defining an artifact's properties. In contrast, the social constructivist approach emphasizes how representations by different social groups play a mediating role between the technology and its social context. The changes in practices that follow the introduction/use of an artifact or technology here depend upon the social representations of what the artifact/technology is and what it does. Technologies, in other words, do not have objective properties: they display ‘interpretive flexibility’ (Pinch and Bijker, Reference Pinch, Bijker, Bijker, Hughes and Pinch1987), and can be attributed very different functions and properties depending on the agencies that use or produce them. Absolute physical constraints in this framework do not exist, but what appears to be a constraint is in reality a social construction, a particular reading of the artifact that seems objective because it has become obdurate through negotiation and closure (Pfaffenberger, Reference Pfaffenberger1992).

According to the ANT view, neither realists nor constructivists are able to account for the complex interactions between people and things. Realists underestimate the power of humans when faced with the intrinsic influence of certain technologies, while constructivists underestimate the power of objects. According to ANT scholars (Callon, Reference Callon, Bijker, Hughes and Pinch1987; Latour, Reference Latour1987), the properties of artifacts neither are objective facts, as in the realist tradition, nor are they mere social constructions, as in the social constructivist tradition: artifacts and their properties are both real and constructed. They emerge from a heterogeneous network of social and technical elements (including social representations and natural forces/technical elements) that co-construct them. Any phenomenon in this framework is thus the outcome of the workings of heterogeneous socio-technical ensembles of elements also referred to as ‘hybrid collectives’ (Callon and Caliskan, Reference Callon and Caliskan2010). Within such collectives, both humans and non-humans (i.e. artifacts) can act, influence, and mediate (Latour, Reference Latour1987). Thus, in this view, an artifact's properties are neither inherent nor socially constructed, but relational and emergent: actants can form links with each other and this can give rise to networks of actants. When the associations in a network become stable or solidify, the network stabilizes. Properties emerge as black boxes, as apparently objective properties of actants. In this paper, I hope to demonstrate that this dynamic and emergent characterization of artifacts and their properties can provide vantage points to study routines dynamics. First, however, we need to delve deeper into the relationship between artifacts and routines.

Routines and artifacts

A plethora of artifacts is typically involved in the course of a routine. These may include letters, documents, clocks, calendars, meters, gauges, display boards, forms and invoices (Nelson and Winter, Reference Nelson and Winter1982), spatial arrangements of machines, tools and materials, and written codes of SOPs, or computer systems (Cohen et al., Reference Cohen, Burkhart, Dosi, Egidi, Marengo, Warglien and Winter1996). While all kinds of artifacts can play instrumental roles in rule- and routine-following, routines authors have traditionally placed an emphasis on ‘cognitive’ or ‘representational’ artifacts (Cohen et al., Reference Cohen, Burkhart, Dosi, Egidi, Marengo, Warglien and Winter1996). These include SOPs and formal rules, often referred to as ‘artifactual representations’ of routines (Hutchins and Hazelhurst, Reference Hutchins, Hazelhurst and Langton1991; Cohen et al., Reference Cohen, Burkhart, Dosi, Egidi, Marengo, Warglien and Winter1996; Lazaric and Denis, Reference Lazaric and Denis2001; D'Adderio, Reference D'Adderio2001, Reference D'Adderio2003, Reference D'Adderio2008a, Reference D'Adderiob; Pentland and Feldman, Reference Pentland and Feldman2005, Reference Pentland and Feldman2008). In Cohen et al. (Reference Cohen, Burkhart, Dosi, Egidi, Marengo, Warglien and Winter1996), for example SOPs are defined as a kind of representation, a formalized statement of what actions should occur. This separates SOPs from the actions occurring as routines are expressed in context, a separation that is desirable as ‘real behavior diverges substantially from formalized SOPs’ (Michael Cohen in Cohen et al., Reference Cohen, Burkhart, Dosi, Egidi, Marengo, Warglien and Winter1996: 673). Thus, in this view, routines can be coded in cognitive artifacts (representations) such as work-flow graphs, which facilitate their manipulation, to be subsequently ‘brought back to the field’ generating new expressions (actual routines). Written rules and procedures can serve as proxies for the ostensive aspect of routines (Pentland and Feldman, Reference Pentland and Feldman2005; D'Adderio, Reference D'Adderio2008), or they can be read as embodying specific configurations of views, goals, and dispositions by different agencies (D'Adderio, Reference D'Adderio2003, Reference D'Adderio2008a, Reference D'Adderiob). The relationship between SOP, rules, and ostensive can thus reveal the extent of divergence, whereas the match between the SOP or rule and the actual routine or performance can be understood as revealing the extent of control (Pentland and Feldman, Reference Pentland and Feldman2005). Increased control can be achieved by delegating the rule or procedure to machines or software, which tends to make them more persistent and durable (D'Adderio, Reference D'Adderio2008a, Reference D'Adderiob; Pentland and Feldman, Reference Pentland and Feldman2008).

Along the same line of thought, Rafaeli and Pratt (Reference Rafaeli, Pratt, Rafaeli and Pratt2006) have advocated the need to study ‘linked artifacts’, such as written procedures and multiple performances, which hold a special relationship as they evolve from one another or are meant to replicate each other. Key issues highlighted here involve understanding how multiple representations co-evolve, what is the relationship between an actual routine and its ‘copy’, and the implications for performances of introducing a tangible artifact that captures the routine. I will address these key aspects later in the paper. At this stage, it will suffice to emphasize the important role that formal procedures and rules play in routines reproduction. In recognition of their key roles, rules and SOPs will be the main focus of the rest of the paper.

When focusing upon the role of artifacts in Routines Theory I thus identify three approaches: first, artifacts in early Routines Theory (Section 3 in this paper); second, artifacts and agency (Section 4); and, third, artifacts at the centre of the routine (Section 5). This journey will afford substantial progress towards accomplishing the shift required to bring artifacts to the centre of routines.

Artifacts as ‘external memory’

Routines scholars have been concerned with the role of artifacts in routinized performances from the outset. In early contributions, artifacts are part of, and help constitute, the context that provides the background to routinized performances. According to Winter (in Cohen et al., Reference Cohen, Burkhart, Dosi, Egidi, Marengo, Warglien and Winter1996), for example, context dependence is a fundamental feature of routines, such that ‘their effectiveness is not measured by what is achieved in principle but by what is achieved in practice; this generally means that the routine might be declared effective in some specific contexts, but perhaps not in others’ (p. 662). Winter (ibid.) highlights two key aspects of context: physical, which is expressed in the local/artifactual complements to the routine, including the material tools and physical plant layout and equipment; and motivational or relational, which captures the agents’ discretion to decide whether, when, and how to perform a routine.

These early contributions emphasize the role of artifacts as ‘external memory’ (Nelson and Winter, Reference Nelson and Winter1982); artifacts help humans deal with solving complex problems by sharing some of the cognitive burden. This notion resonates with the later anthropological notion of distributed cognition (Hutchins, Reference Hutchins1995), pointing to the fact that a routine does not reside in any one place, and certainly not solely in the human mind, but is instead distributed across people and artifacts, including rules and technologies. I will address these key notions later in the paper. At this point, I want simply to highlight the fact that Routines scholars have pointed from the very outset to the complex and distributed web of artifacts and coordinating relationships that constitutes and supports a routine (Winter, Reference Winter and Montgomery1995, Szalanski, Reference Szulanski1996).Footnote ² This bears important consequences for capturing the nature of routines and the role of artifacts.

Artifactual ‘representations’ vs ‘expressions’

There is a further contribution in earlier Routines Theory, which also points to the critical role of artifacts and has been similarly overlooked in subsequent work. Early scholars have devoted specific attention to the key relationship between formalized, or artifactual ‘representations’ of routines (including SOPs and formal rules) and routines in practice, or ‘expressions’. Cohen et al. (Reference Cohen, Burkhart, Dosi, Egidi, Marengo, Warglien and Winter1996), for example, while agreeing that tacit knowledge and non-deliberate behaviour are involved in the operation of routines, also highlight that stability and replicability depend on ‘the following of explicitly stated rules and instructions . . . and a carefully designed artifactual environment within which workers can learn coordinated behaviors’ (Warglien in Cohen et al., Reference Cohen, Burkhart, Dosi, Egidi, Marengo, Warglien and Winter1996: 659). According to early authors, thus, artifactual representations of routines – as ‘cognitive artifacts’, which include rules and SOPs – play a central role in the evolution of routines, including their transfer and replication. Transferring ‘best practice’, it follows, rather than involving the straightforward uplift and transplantation of routines to a new site, involves ‘a large effort to set up a “technology of replication” that usually implies (i) learning a language within which to code successful routines, (ii) creating cognitive artifacts that can be diffused (through flowcharts and other replicable representations), (iii) translating the high-level description contained in the cognitive artifact in actual practice, generating a new routine adapted to the new context’ (Hutchins and Hazelhurst, Reference Hutchins, Hazelhurst and Langton1991 in Cohen et al., Reference Cohen, Burkhart, Dosi, Egidi, Marengo, Warglien and Winter1996: 674). Thus, in earlier routines work, procedures and rules as artifactual representations of routines are key and intrinsic aspects of routines production and reproduction (see also Schultz, Reference Schultz and Becker2008).

The discussion above shows that the exact nature and role of formal rules and procedures (as representations), and the extent to which these artifacts are able to shape the course of actual routines (expressions), were key topics in early routines studies. Nevertheless, this pioneering discussion was limited to identifying ‘interactions’ and ‘tensions’ between these two aspects of routines and thus failed to make significant progress towards assessing their reciprocal influence. To achieve this objective, we need more advanced and finer-grained theoretical tools and methodologies.

Artifacts are ‘not the routine’

The strong emphasis on artifacts found in earlier Routines literature was not pursued in subsequent contributions with the exception of a rather narrow stream of routines studies (see Becker et al., Reference Becker, Lazaric, Nelson and Winter2005; Pentland and Feldman, Reference Pentland and Feldman2005, Reference Pentland and Feldman2008; Cohen, Reference Cohen2007; Adler and Obstfeld, Reference Adler and Obstfeld2007; D'Adderio, Reference D'Adderio2001, Reference D'Adderio2003, Reference D'Adderio2008a). The relative lack of interest in artifacts must be understood as part of the authors’ conscious effort to move away from a reductive view of routines as ‘things’ and bring agency back into the picture. This meant attributing artifactual representations a lesser role when compared with actual routines (expressed in context). Thus, the observation that ‘what finally generates value is the routine (the expression) rather than the code (the representation)’ (Cohen et al., Reference Cohen, Burkhart, Dosi, Egidi, Marengo, Warglien and Winter1996: 674), while entirely accurate, underlies a representational view of routines that implicitly relegates their artifactual representations to the lower rank of imperfect imitations. The crux of the matter here is precisely what function can we ascribe to a category of artifacts that plays such a crucial role in the evolution of routines.

On the one hand, it is important to reiterate that formal procedures, SOPs, and rules are not the routine (Bourdieu, Reference Bourdieu1977; Cohen et al., Reference Cohen, Burkhart, Dosi, Egidi, Marengo, Warglien and Winter1996; Suchman, Reference Suchman1983; Pentland and Feldman, Reference Pentland and Feldman2008; D'Adderio, Reference D'Adderio2008a). They are formal, explicit, synthetic, selective, partial representations of routines (D'Adderio, Reference D'Adderio2003, Reference D'Adderio2008a). Distinguishing in this way between formal procedures and routines is fundamental to avoid the categorical mistake of assuming that it is sufficient to design a procedure to achieve a certain performance, as presumed in much managerialist literature. Pentland and Feldman (Reference Pentland and Feldman2008), for example, have warned practitioners against ‘the folly of designing artifacts while hoping for patterns of actions’ (p. 1). While practitioners all too often design artifacts in their attempt to control and prescribe routines, they are committing the clear mistake of confusing artifacts with the routine's ‘ostensive’ (abstract) aspect (ibid.: xx).

At best artifacts can be ‘partial representations’ containing selected and codified configurations of the multiple (ostensive) views and physical arrangements that make up a routine (D'Adderio, Reference D'Adderio2008a). Moreover, while such artifacts are designed for the very purpose of directing performances, the actor's complicity in following their course is often required. SOPs, thus, rarely prescribe work, as actors are often able to modify them, or even altogether neglect to include them in their performances. Thus, in Suchman's case (Reference Suchman1983), office routines differ from their representations. In Feldman's example (Reference Feldman2000), routines are not incorporated into performances; the workflow and freeze processes fail to fully prescribe the routine in D'Adderio (Reference D'Adderio2003, Reference D'Adderio2008a respectively); and in Lazaric and Denis (Reference Lazaric and Denis2001), SOPs fail to determine actions.

On the other hand, while this distinction is important, there is more to it. Even though artifacts very rarely coincide precisely with the routine, they nonetheless still play a key role in their evolution. It is precisely for this reason that artifacts, in general, and artifactual representations, in particular, deserve further investigation. For example, addressing the relationship between representations (routines-in-theory) and expressions (routines-in-practice) in depth and in detail can provide vantage points to characterize the evolution of routines (D'Adderio, Reference D'Adderio2003). This includes uncovering the processes of translation from formal routines to actual performances and vice-versa, from performances back to procedures, or, in other words, the co-production of procedures and performances (D'Adderio, Reference D'Adderio2001). By adopting the notion of translation, a term dear to the Sociology of Technology, I intend to emphasize the fact that the movement from formal procedures to performances, and vice-versa, entails not so much a straightforward process of codification and de-codification (of performances into procedures and vice-versa) but one of deep transformation, of socio-technical ‘actualization’ and ‘virtualization’ involving both actors and artifacts (as in the notion of ‘translation routines’ in ibid).

Building on this line of thought, more recent work by D'Adderio has further theorized the mutual influence of procedures (routines-in-theory) and performances (routines-in-practice) in the cases of the engineering freeze process (2008a) and of the exact replication of routines (2008b) both in high-technology manufacturing contexts. Drawing from Performativity Theory, the first study examines the introduction of software-embedded SOPs in the Engineering Freeze process and shows how (computer-embedded) procedures and rules play a fundamental influence in shaping performances: formal routines are dynamically performed through iterative cycles of framing, by which there is convergence between model and reality, overflowing (by which there is divergence between model and reality) and further reframing (D'Adderio, Reference D'Adderio2008a,Reference D'Adderiob) (Figure 1).

Source: D'Adderio (Reference D'Adderio2008).

Figure 1. Reciprocal influence of procedures and performances

The second case study builds on these findings to capture the key role of artifacts, or SOPs, as intermediaries and mediators in the transfer of a production capability and associated routines in manufacturing. This work characterizes the deeper mechanisms by which procedures influence performances at both origin and destination sites (D'Adderio, Reference D'Adderio2008b) and shows how similarity between origin and destination is the emergent outcome of iterative cycles by which the formal, computer-embedded, procedures are performed at both sites (Figure 2). These examples demonstrate clearly that, in order to achieve further insights into routines dynamics, we need to go beyond a passive view of procedural artifacts to analyse their complex and active role in the production and reproduction of routines. This role entails enrolling a plethora of other artifacts (including objects, tools, and technologies) that make their reproduction possible, or even likely. Understanding routines, in other words, involves taking artifacts seriously.

Source: D'Adderio (Reference D'Adderio2008b).

Figure 2. Relationship between procedures and performances in artifact-mediated routines transfer

Artifacts and routines: dead or alive?

In order to advance our understanding of the role of artifacts in routines’ change and stability, we need therefore to turn our attention to the character of artifactual representations and the socio-technical mechanisms that regulate their evolution. Michael Cohen's (Reference Cohen2007) distinction between ‘dead’ and ‘live’ routines, inspired by Dewey's influential theory (Reference Dewey1922), is highly relevant here. While ‘dead routines’ tend to be rigid, mindless, and are typically codified in artifacts, tools, and technologies, ‘live routines’ are flexible, mindful, and involve the contribution of actors, their experience and learning. The distinction between live and dead routines delivers a powerful message about the nature of routines and the role of agency in routines, which has been lacking in earlier routines literature. While this analytical distinction is useful, however, it requires further qualification when it comes to characterizing the role of artifacts. In particular, I suggest that these notions must be treated with care to ensure that the connotation ‘dead’, when attributed to a routine, does not result in obscuring the nature and role of artifactual representations of routines and their own internal mechanisms (dead meaning not so much inconsequential, but hardened, solidified). After all, there are entire bodies of scholarly knowledge that draw from ‘dead artifacts’ to inform the way in which we understand and characterize organizations, cultures, and societies. Anthropologists and archaeologists, in particular, are able to reconstruct entire societies from their scattered, partial remains and vestiges. From anthropology, thus, we learn that human agencies leave traces of their knowledge and work in artifacts (Thomas, Reference Thomas1991). Another way to express this is to say that artifacts are front-loaded with the habits, intentions, and rationales held by the agencies by which they have been created, adopted, and adapted.

When extending this thinking to routines, we can see how previous experience and learning by organizational agencies are selected and codified into formal routines and procedures. So-called ‘best practices’Footnote ³, for example, originate from assumptions that come from sectoral and industrial experience, as well as the knowledge of analysts, regulators, and policy-makers. When they are embedded in artifacts, such as software packages, they also come to incorporate the knowledge, learning, and experience of software users and producers. In this process, layers of rationales, assumptions, and experience become over-imposed in artifacts (D'Adderio, Reference D'Adderio2003 and Reference D'Adderio2008b). In D'Adderio (Reference D'Adderio2008b), for example, we can see how configurations of rationales embedded in the central computer model reflect the outcome of performative struggles in which the artifact has been involved.

The immediate implication of this argument for the adoption of standard practices is that rationales delegated to artifacts inevitably interact (meaning that they may either complement or conflict) with the intentions and rationales of the locale where the procedure is used/adopted, as further discussed below. These routines may certainly be dead, but they nevertheless retain traces of previous lives, and, because of this, they are worthy of study and attention. It is in fact their very nature, their relative stability when compared with the variation that is present in performances, and their relative solidity when compared with abstract (ostensive) views that makes them ideal loci for study. Formal routines and procedures, in being often codified and embedded in material artifacts, are relatively stable and easy to observe, and can thus provide invaluable empirical vantage points to explore routines dynamics. Ultimately, artifacts are only part of what makes up the routine; nevertheless, they do play a key role that scholars have only just begun to explore.

Highlighting the key role of artifacts: routine-following is distributed

The first step in our quest to place artifacts at the centre of routines involves acknowledging the important role that they play in routine-following. Routinized performances typically involve not only humans, the focus of much mainstream routines literature, but also non-humans. The key role that non-humans play in the performance and evolution of routines can be usefully captured through the notion of distributed action/cognition (Hutchins, Reference Hutchins1991, Reference Hutchins1995).

Rooted in Cognitive Science (Hutchins, Reference Hutchins1991, Reference Hutchins1995; Hutchins and Hazelhurst, Reference Hutchins, Hazelhurst and Langton1991), the notions of distributed knowledge and action address the fact that there is more to cognition than what takes place in human minds. By describing processes in terms of functional relationships between brains, other people, and external objects, for example, Cognitive Science places an emphasis on the role of contextual elements in shaping individual and collaborative action and cognition. The role of artifacts as mediators is thus brought to the fore. Artifacts in this framework not only include man-made objects but also representational forms (i.e. spoken language, graphical models, text, etc.) which help propagate information in the material and social environment. Special attention is thus devoted to ‘cognitive artifacts’, tools developed for enhancing or facilitating human cognition whose close observation can provide valuable insights into the meaningful, constantly evolving practices that specific cultures produce and reproduce (Hutchins, Reference Hutchins1995).

Cognitive artifacts are ‘objects’ made by humans for the purpose of aiding, enhancing, or improving cognition. Scholars have included in this category a range of artifacts including not only forms, flowcharts, worksheets, schedules, and graphs (Hutchins, Reference Hutchins1995; Hutchins and Hazelhurst, Reference Hutchins, Hazelhurst and Langton1991) but also rules of thumb, proverbs, mnemonics, and memorized procedures, which play a similar role to objects in some cognitive processes (Norman, Reference Norman1993). Early work in Actor Network Theory (Latour, Reference Latour1987) has shown how the distribution of cognitive activity within socio-technical networks and between people and inscriptions accounted for much of the work of science. According to Latour, non-linguistic inscriptions or representations such as maps, charts, graphs, and tables enable the superimposition of representations of otherwise incommensurable items (1986). The physical and material arrangement of artifacts themselves has also been referred to as a cognitive artifact, including the arrangement of tools and machines on a factory floor (Nelson and Winter, Reference Nelson and Winter1982; Cohen et al., Reference Cohen, Burkhart, Dosi, Egidi, Marengo, Warglien and Winter1996; Pentland and Feldman, Reference Pentland and Feldman2005). A variety of artifacts including diagrams, checklists, forms, SOPs, rules, and procedures, some of which may be subsequently embedded in a software artifact, are often devised by managers to design or redesign a work process (Pentland and Feldman, Reference Pentland and Feldman2005, Reference Pentland and Feldman2008). Similarly, ‘intellectual’ or ‘conceptual’ equipment, including models, equations, and calculators, are described as fundamental in sustaining the performance of practices in financial markets (MacKenzie, Reference MacKenzie2009).

The notion of distributed cognition is not entirely new to Organizational Theory. According to Simon (Reference Simon1969) and Arthur (Reference Arthur1994), agents faced with complicated tasks conceive of tools, create rules and routines or set up organizations to relieve their brains and enhance their performance. In so doing, these authors have extended the actors’ cognitive capacities by ‘redistributing their brains, at least implicitly’ (Callon and Muniesa, Reference Callon and Muniesa2005: 1237, emphasis added). In early Routines Theory, analogously, the notion of distributed knowledge has provided ‘the natural locus of attention for lines of research focused on the role of artifacts’ (Cohen et al., Reference Cohen, Burkhart, Dosi, Egidi, Marengo, Warglien and Winter1996: 683).

The implications of treating routinized activity as distributed, however, have not been fully acknowledged in extant Routines Theory. To further our understanding, we thus need to turn to the work of researchers in Science and Technology Studies, especially that based on Actor-Network Theory. According to STS scholars, knowledge and actions are rarely individual; they mobilize entities, humans, and non-humans, which participate in the creation of knowledge or the performance of actions. This holds important implications for routine and rule-following: participation in (routinized) performances by artifacts is not quite as passive as more or less explicitly assumed in earlier theory and can only exceptionally be reduced to its purely instrumental dimension. According to Callon and Muniesa (Reference Callon and Muniesa2005), thus, the notion of distributed cognition leads us away from standard theories of action, which reserve agency for humans alone, towards the notion of distributed agency. Actor Network Theory (ANT) scholar Bruno Latour, for example, shows how calculative agencies are not human individuals but ‘collective hybrids’ (1987, see also Callon and Caliskan, Reference Callon and Caliskan2010). Latour argues that calculation does not take place solely in the confines of the human mind, but is distributed among humans (equipped with instruments) and non-humans. Analogously, for Callon and Muniesa (Reference Callon and Muniesa2005) agencies’ calculative capacities are not stand-alone but linked to distributed equipment. Distributed equipment includes bodies, tools, and machines (i.e. the broker's ear in MacKenzie, Reference MacKenzie2009; the stock ticker in Preda, Reference Preda, Callon, Millo and Muniesa2007; trading screens in Knorr Cetina, Reference Knorr Cetina, Knorr Cetina and Preda2005) as well as models, theories, equations, rules, and procedures (MacKenzie, Reference MacKenzie2009). Routine and rule-following are, in other words, truly distributed and material processes (Hatherly et al., Reference Hatherly, Leung and MacKenzie2007; D'Adderio, Reference D'Adderio2008a; MacKenzie, Reference MacKenzie2009).

A crucial consequence of treating cognition and action, including rule-following, as distributed, I propose, is acknowledging the fact that knowledge and action are not simply ‘distributed between’ but ‘stretched across’ actors and artifacts. This subtle but key distinction has often been overlooked by organizational scholars. These have traditionally taken a simpler interpretation of distributed cognition, merely as external source or repository of information or technology from which actors can arbitrarily draw during the course of their activities. The fact that cognition is distributed, however, bears deeper meaning as well as holding important consequences. The notion suggests that the actor's knowledge, skills, and competences depend on – and are at the same time configured by – the tools and artifacts they encounter or involve into their routinized performances. In other words, the skills and capabilities of actors are mediated and fundamentally transformed by the capabilities of the tools and instruments that they use in their work (Latour, Reference Latour2005). The notion of ‘distribution’ therefore holds implications for routines far beyond those commonly ascribed in the literature.

This distinction is evident in Hutchins’ example of navigation where he showed how the process of piloting a ship in and out of a harbour is a complex, rule-determined activity involving not only the coordination of crew members, but also the use of navigational instruments and maps (1995). A gyrocompass, for example, incorporates some essential rules of sea and land orientation and, for this reason, it requires that the pilots’ skills and activities be adjusted to its properties (Hutchins, Reference Hutchins1991). Similarly, in his ethnography of the two different scientific practices of ‘opticism’ and ‘digitality’, Lynch (Reference Lynch and Pickering1992) showed how rule- and routine-following depend on the complex relationships between material devices, theoretical optics, geometry, and modes of graphic represen-tation.

Along the same lines, this author (D'Adderio, Reference D'Adderio2001) showed how the Virtual Prototype did much more than embody the information and codified knowledge required to design and produce a new vehicle: it allowed for departments and functions across the extended organization (including suppliers and customers) to access development data in real time and feed back their knowledge and expertise into the shared virtual model. The Model also supported new practices and forms of calculation for product design and engineering which involved iterative cycles of ‘translations’. Through these translations some of the local knowledge held by local practitioner experts could be incorporated into the Model (virtualization, i.e. by using digital scanning techniques), and the global digital knowledge embedded into the Model could be used to produce local versions of the product (actualization, i.e. by using advanced CAD-printing techniques).Footnote ⁴ Translations involved conversions from the ‘language’ of local functions and experts (Galison, Reference Galison and Biagioli1999), present in multiple and divergent local formats, to the Model's ‘language’, ordered according to the software format, and back to the local language(s). While the software-embedded prototype did not contain all knowledge used or produced by the different functions involved, it nonetheless afforded new flexible ways to share data and learn about the product (ibid.).

Further examples can be found in more recent literatures. Preda (Reference Preda2004), for example, shows how by altering the equipment of traders through the introduction of the stock ticker, new ways of calculating decisions in financial markets could be afforded. In allowing prices to be known in real time and distant places, the ticker favoured new forms of arbitrage and speculation. For example, the availability of prices through displays allowed the development of techniques of graphic analysis of price variations such as chartism. Similarly, to the case of the Virtual Product in D'Adderio (Reference D'Adderio2001), the ticker did more than provide information, it constructed data that, owing to their format, produced specific effects of cognition and action Callon and Caliskan (Reference Callon and Caliskan2010). A similar argument is raised by Beunza and Stark (Reference Beunza and Stark2004) in their study of the trading room of a modern investment bank. Here heterogeneous forms of calculation were produced that depended on the devices used by traders to distribute their calculative activities (including trading robots, telephones, pricing tools, etc.).

The examples above illustrate how understanding routine- and rule-following as cases of distributed cognition can hold important implications for how we characterize artifacts. Artifacts in this light acquire a new role, one that goes beyond the passive intermediary role attributed to them in Organizational Theory to the role of mediators that participate in the co-creation of knowledge and transformation of actions (Latour, Reference Latour2005).

Artifacts as actants: the role of ‘inscriptions’

Analysing the role of artifacts as mediators entails moving beyond a characterization of artifacts as opaque, monolithic objects to account for the dynamics by which they are produced and reproduced as they become involved in routinized performances. In turn, this will provide valuable means to understand how artifacts influence those performances in which they are engaged. An important conceptual device to guide our exploration in such an uncharted territory, is to study artifacts in general, and the artifactual (coded or de-contextualized) routine, in particular, as ‘inscriptions’.

Authors in the Science and Technology Studies (STS) tradition have shown how a complex range of rules and assumptions are embedded as ‘scripts’ within artifacts and technology during the design and usage stages (Barley, Reference Barley1986; Akrich, Reference Akrich, Bijker and Law1992; Latour, Reference Latour, Bijker and Law1992; Grint and Woolgar, Reference Grint and Woolgar1992). Creating scripts involves the socio-technical process of ‘inscription’ (Latour, Reference Latour, Bijker and Law1992) by which dominant interests or ‘programs for action’ are reflected in the form and functioning of a technology. Inscriptions are ways in which specific functions can be delegated to artifacts and technologies. Latour's hydraulic door closer, for example, absorbs the energy of those who open the door, retains it, and then releases it slowly in a manner similar to what one could expect from a well-trained butler (Latour, Reference Latour, Bijker and Law1992). While this does not imply that the assumptions that designers and users embody in technological artifacts are prescriptive (Latour, Reference Latour1987), it does point to the influence of scripts and their configurations on future performances that involve that particular technology or artifact. The notion of inscription thus can provide important ways to advance our characterization of artifacts and technologies as mediators that perform or influence the practices in which they are involved and the contexts in which they are embedded.

While useful for understanding artifacts in general, the notion of inscription can be especially valuable to characterize the influence of artifactual representations (i.e. formal procedures, SOPs, rules) on routines. Formal routines as inscriptions have some key characteristics. First, they are not neutral (Barley, Reference Barley1986; Akrich, Reference Akrich, Bijker and Law1992; Bakken and Hernes, Reference Bakken and Hernes2006; Preda, Reference Preda, Callon, Millo and Muniesa2007; D'Adderio, Reference D'Adderio2003, Reference D'Adderio2008a) but instead reflect the objectives, motivations, values, and dispositions of the agencies that use/produce them as well as the context from which they are abstracted. Second, they are selective, implying that the knowledge they contain has been sifted, ordered, and classified according to one or more rationales or ‘logics’ (Bowker and Star, Reference Bowker and Star1999; D'Adderio, Reference D'Adderio2003). In D'Adderio (Reference D'Adderio2008a), for example, the software-embedded SOP embodies the Design Engineers’ view of the world (expressed in their classification of the product structure according to vertical, parent–child relationships), which clashes with Production Engineers’ view of the world (expressed in their classification of the product structure according to horizontal, Boolean algebraic statements). Accounting for the role of technology- and artifact-embedded rules as inscriptions implies that focusing solely on actions by human agents to explain rule- and routine-following practice would leave us short of an explanation.

The idea that routines, and their artifactual representations, embody contrasting views, intentions, or logics that belong to various agencies is not totally removed from early assumptions in Organizational Theory, whereby authors have defined SOPs as the outcome of locally solved conflicts of interests (March and Simon, Reference March and Simon1958; Nelson and Winter, Reference Nelson and Winter1982): SOPs embody social norms and worldviews that belong to the agencies that have created/adopted them and ‘because of the reallocative effects involved in the change of rules any change would re-create new internal conflicts that could not be easily solved’ (Cohen et al., Reference Cohen, Burkhart, Dosi, Egidi, Marengo, Warglien and Winter1996: 694). The crucial implications of this notion for understanding routine-following, however, have not yet been fully explored in Routines Theory. This is where STS studies become especially useful.

From STS scholars, for example, we learn that, as long as the script finds no problem, the program for action that is expressed in the script is not opposed. When questioned, however, the script becomes visible and anti-programs can develop whereby users attempt to react to the assumptions embedded in the technology (Latour, Reference Latour1987). The implications of this are particularly relevant to routines transfer and replication: what happens, for example, when procedures loaded with embedded rationales that reflect a truce in a specific context are transferred to a new context? What kinds of conflicts are likely to arise as embedded rules and norms that belong to a previous context are performed in a new context, characterized by different rules and understandings? And what are the implications of this for the evolution of routines and their artifactual representations in the new context? In a study of the transfer of a complex capability and associated routines across two units of a global manufacturing enterprise, for example, this author has shown how selected assumptions and (ostensive) views that belong to the origin site are embedded in the shared Computer Model, which contains a codified version of all routines that make up the capability (D'Adderio, Reference D'Adderio2008b). The assumptions embedded in the SOPs and Model is subsequently performed at a destination where they shape local performances. The study also shows how the Model and associated procedures themselves change as a consequence of being performed at the destination. Deeper understandings of the role of artifacts can open up new avenues for routines research.

To make further progress in this promising direction, however, we need a theory that is able to analyse artifacts and chart the specific arrangements of knowledge, views, assumptions, and motivations that have produced them. In the case of artifactual representations, such a theory should allow us to capture the micro dynamics by which formal routines as inscriptions are brought to life (Cohen, Reference Cohen2007; Pentland and Feldman, Reference Pentland and Feldman2008; D'Adderio, Reference D'Adderio2008a) in a new setting, or in other words how they perform the new setting. In addition, our approach should allow us to capture the dynamics by which artifactual routines come into being, or are created by abstraction from an existing setting. Such a theory can enable us to capture how specific (ostensive) views and rationales come to be embedded in artifacts (and artifactual representations of routines), and the subsequent effect that artifacts, containing a specific configuration of scripts, have on performances (D'Adderio, Reference D'Adderio2008b). Applying the notion of script to routines, in other words, allows us to ‘situate the ostensive’ aspect by acknowledging that views and objectives do not simply reside in the actors’ minds but are also embedded in – or delegated to – artifacts. Taking the notion of scripts seriously implies recognizing that agency can be embedded in artifacts, both as traces of actions (as in a step-by-step instructions procedure) and of intentions, assumptions, rationales, and logics, and that this will have crucial implications for routines evolution. Before we can achieve a new theorization, however, a further step is required, one that allows us to capture the ways and degrees in which artifacts can influence routinized performances.

The influence of artifacts on routines: constraints, affordances and the ‘power of default’

The third step towards taking artifacts seriously entails analysing more closely the role and influence of artifacts, tools, and technologies on routines. This involves going beyond the current characterization of artifacts as simply ‘guiding’ and ‘constraining’. According to Callon and Muniesa (Reference Callon and Muniesa2005), for example, the artifact does not simply offer passive guidance: it offers an invitation. This notion, similar to the notion of affordance (Gibson, Reference Gibson1979), implies that the artifact, simply by being there, and available, proposes a specific kind of ‘calculation’ to the actor, who may accept (or not) the ‘invitation’ and ask the tool to perform the calculation (ibid.). The agencies’ calculative capacities are thus linked in complex ways to their equipment, which is distributed. The shift in emphasis towards a more active role for artifacts bears important consequences for routines. Far from implying that artifacts, including rules or procedures, determine action, or the course of a routine in our case, this concept suggests a richer and more meaningful role for artifacts than that currently allowed in the literature.

The dominant way in which we understand the influence of artifacts and technologies in Social Science can be traced back to Wittgenstein. Drawing from his work, early social constructivists and ethnomethodologists, for example, have highlighted how (technology-embedded) rules and classifications are never deterministic but always interpreted (Garfinkel, Reference Garfinkel1967; Barnes, Reference Barnes1982; Bloor, Reference Bloor1997; Hatherly et al., Reference Hatherly, Leung and MacKenzie2007; Lynch, Reference Lynch and Pickering1992). In their view, and in accordance with one of Wittgenstein's prevailing interpretations, the irreducible interpretive flexibility of rules is such that – at least in theory – it can lead to infinite regression as ‘no course of action could be determined by a rule, because every course of action can be made out to accord with the rule’ (Wittgenstein, Reference Wittgenstein1967: 81). Due to the logical under-determination of behaviour by rules, there can be no closure as ‘The rule is, at any given time, what the practice has made it’ (Taylor, Reference Taylor, Schatzki, Cetina and von1993: 57–8, emphasis in original).

Organizational scholars belonging to the stream of Structuration Theory, have drawn on these approaches in conceptualizing the role that technology plays as a source of constraint and enabler in rule-following (Barley, Reference Barley1986; Orlikowski, Reference Orlikowski1992). According to Orlikowski (Reference Orlikowski1992), for example, technology constrains performance by facilitating it in a particular manner. A telephone, computer, hammer, or pencil can thus facilitate the performance of certain kinds of work and not others. The emphasis in these literatures however is firmly on human practices and the discretion of human agents to follow a procedure or rule or to choose to do otherwise (Giddens, Reference Giddens1993; Orlikowski, Reference Orlikowski1992). According to this framework, the properties embedded inside artifacts are never predetermined but rather ‘the capacity to modify the ‘rule’ that is drawn on in any action is an ever present possibility’ (Cassell, 1993 in Orlikowski, Reference Orlikowski2000: 411). Similarly, in his study of the introduction of CT scanners into two different hospitals, Barley (Reference Barley1986) showed how, while technologies do influence organizational structures, their influence depends on the context in which they are embedded. At the extreme, rules in this view exist only ‘virtually’ and are consequential only to the extent that they are enacted by users through practice.

While crucial in dispelling technological determinism, this view contains a somewhat partial characterization of the role of artifacts. While in fact, in rule-following as well as in other realms of human practice, artifacts do not determine actions, they nevertheless play an important role (Hatherly et al., Reference Hatherly, Leung and MacKenzie2007). Technical systems, for example, make it possible or easy to do certain things, and impossible to do others, so that, while in theory there can be infinite regression, ‘the logical open-endedness of the application of terms to particulars and the logical under-determination of behaviour by rules are foreclosed in practice’ (ibid.: 11: emphasis added). In other words, once embedded in artifacts, skills and tacit knowledge (Latour, Reference Latour, Bijker and Law1992), rules (Hutchins, Reference Hutchins1991; Preda, Reference Preda2000), and procedures (D'Adderio, Reference D'Adderio2003; Hatherly et al., Reference Hatherly, Leung and MacKenzie2007) tend to become more stable and durable, and this holds radical implications for rule-following behaviour. So while, in theory, actors can always choose to reject a tool or a rule, or decide to abide by it ceremoniously, in practice actors often follow the path of least resistance (MacKenzie, Reference MacKenzie2006).

This principle can be illustrated by analysing a category of technological artifacts, and is both interesting and relevant: software. Information systems, as bundles of inscriptions, play a fundamental role in influencing rule-following. Such systems are ‘neither merely neutral media nor simply means of increasing the efficiency of what unaided human beings might do’ (Hatherly et al., Reference Hatherly, Leung and MacKenzie2007: 32). They structure work, extend interactions, increase visibility of knowledge and actions, create a common platform for the accumulation of common knowledge, constrain the ability of practitioners to alter the results of another, regulate who has access to making changes, track the progress of changes, link multiple sites in different time/geographical locations, facilitate data sharing and the reception of feedback (Orlikowski, Reference Orlikowski2002). They solidify and stabilize rules, procedures, and classifications thus making it more difficult to avert them (Hatherly et al., Reference Hatherly, Leung and MacKenzie2007). Clearly, on the one hand, technical constraint is never absolute, and indeed many systems’ controls can be subverted if sufficient resources and incentives are applied to the task. On the other hand, however, there are reasons why the influence of technologies – in general – and information systems – in particular – is critical. While in theory it is always possible to bypass software-embedded controls, in practice this does not occur very often (D'Adderio, Reference D'Adderio2003, Reference D'Adderio2008a).

A first reason is that the assumptions, rules, procedures, and classifications embedded in software, both at design and usage stages, tend to become part of the users’ habitual background (Bourdieu, Reference Bourdieu1977), or ‘the way we do things around here’ and as such often go unquestioned. Secondly, as distributed and pervasive technologies, information systems are often entangled into a thick web of organizational relationships which make them difficult to avoid. Once adopted, for example, software can influence what kind of information should be created, selected, and shared, with whom, in which format and in what sequence. What is important here is that, while practitioners can choose to bypass the software, their boycott will hold consequences for them in terms of their ability to have their work taken into account by others in the organization. Thirdly, software tends to make information more visible across an organization thus making it easier to control that actions actually comply with the software. Fourthly and finally, while formal software controls can – in theory – be easily modified or entirely averted, in practice this requires the deployment of resources (i.e. time and software programming skills), which are often unavailable. In these circumstances, the ‘power of default’ of such artifacts (Koch, 1998 in D'Adderio, Reference D'Adderio2003; Pollock and Cornford, Reference Pollock and Cornford2004) will prevent adaptation and customization. An appreciation of the multiple ways and varying degrees to which artifacts and technological systems can influence routines and rule-following calls for the need to bring this analysis to the forefront of Routines Theory. This step involves careful consideration of artifacts and their embedded assumptions and dynamics in order to begin to understand their deeper implications for routines.

Routines and artifacts: degrees of influence

A deeper characterization of the relationship between routines and artifacts entails moving beyond extremes to study the dynamics that regulate their mutual influence. To achieve this goal we need a finer-grained framework that is able to capture the influence of artifactual representations (routines in theory) on performances (routines in context) that goes beyond the extremes of prescription (full influence) and description (no influence). Some recent advances in Performativity Theory, and their even more recent application to Routines Theory, can provide a new characterization. According to Performativity scholars (Callon, Reference Callon and Callon1998; Callon and Muniesa, Reference Callon and Muniesa2005), while the two extreme cases, one where the tool or artifact is instrumentalized and the other where the actor is instrumentalized, are always possibilities, performativity, which involves some kind of reciprocal adaptation under conditions of uncertainty, is more often the norm. This is especially true when an artifact is entangled within a thick web of organizational relationships, as the earlier example of software illustrates (D'Adderio, Reference D'Adderio2008a).

This point has been raised in recent studies in the Sociology of Finance. In his studies of the market for financial derivatives, MacKenzie (Reference MacKenzie2003, Reference MacKenzie2005, and Reference MacKenzie2006) proposes a fine-grained identification of different categories of performativity (influence of models on reality): ‘generic performativity’, when an aspect of economics (a theory, model, concept, procedure, data-set, etc.) is simply used by participants in economic processes; ‘effective performativity’, when the practical use of an aspect of economics has an effect on economic processes; ‘Barnesian performativity’, when the practical use of an aspect of economics makes economic processes more like their depiction by economics; and ‘counter-performativity’ when the practical use of an aspect of economics makes economic processes less like their depiction by economics.

MacKenzie's framework is especially useful in highlighting how what previous theories considered the norm, are in fact often exceptions (Figure 3). At one extreme of performativity, there is prescription, which represents a very strong instance of performativity: automatic reproduction, pure repetition, no more recalcitrance, recurrent events (Sahlins, 1985 in Callon, Reference Callon, MacKenzie, Muniesa and Siu2007). Full prescription thus corresponds to ‘fiat lux et lux fuit’, as in the case of an automatically reproduced sequence of computer algorithms. At this extreme, which corresponds to the ‘framing view’ in Performativity Theory, there is very little adaptation as models are automatically reproduced. At the other extreme, there is the full demise or disuse of a model or tool, corresponding to the ‘overflowing view’ in Performativity Theory: the influence of the model is so weak that it is disused or rejected, and therefore not enacted in practice. One way to explain the demise of a tool or procedure, of course, is that individual agents have made the conscious choice to reject the model. Performativity, however, while not denying this possibility, affords us a more interesting explanation: the model as statement has not been able to put into motion a world in which it can function. In other words, the statement or formula has not been able to produce a successful socio-technical agencement.

Source: D'Adderio (Reference D'Adderio2008).

Figure 3. Procedures, rules and models: degrees of influence

A socio-technical agencement is the assemblage of heterogeneous elements that is required for the world contained in the statement to be actualized: ‘A formula that progressively discovers its world and a world that is put into motion by the formula describing it’ (MacKenzie, Reference MacKenzie2003 in Callon, Reference Callon, MacKenzie, Muniesa and Siu2007: 19). According to this notion, a model as statement fails when it has not been able to put into motion a world in which it can function, or, in other words, it has not been able to produce a successful socio-technical agencement. In the struggle between competitive performative programmes, some agencies are able to inscribe their own worldviews in artifacts (i.e. SOPs, rules). These agencies are the most likely to succeed in exerting their own influence: enrolling artifacts tends to create stronger agencements that, being tightly interconnected into a web of organizational relationships, are more stable and therefore more difficult to oppose or dismiss. The power of a SOPs or rules, thus, can be theorized as the emergent outcome of struggles among the world that it manages to create and other competing agencements. Some agencements are more, and some less, able to enrol people and materials and therefore are more or less successful. Through the notion of agencement, Performativity allows us to capture the influence of models beyond extremes: while full prescription and mere description are always possibilities, most of the time (and this is especially true in conditions of uncertainty) there is performativity, implying some kind of dynamic adaptation between model and reality (Callon, Reference Callon, MacKenzie, Muniesa and Siu2007).

MacKenzie's nuanced classification of the influence of theories and models of reality on reality itself is especially useful when translated to understand the interactions between procedures, rules (models of routines), and performances expressed in context (reality). Drawing from MacKenzie, this author has thus proposed a framework that captures the influence of formal routines or procedures (artifactual representations of routines) over routines in context (performances). The resulting evidence demonstrates that each school of thought, the mechanistic and the interpretivist, has placed itself at one of the extreme ends of MacKenzie's Performativity chart (Figure 2).

At one extreme of this debate (represented by the mechanistic school of thought) is the view that procedures and rules completely prescribe or determine actions. In interpreting the role of SOPs and rules as deterministic and equating actors to rule-following automata, this view corresponds to the ‘framing’ side of our performativity spectrum. This view constitutes an extreme given that, while SOPs and rules frame actions and viewpoints to an extent, full prescription is a rare and extreme outcome that entails automatic reproduction in the complete absence of adaptation or resistance. At the other extreme, is the agency-centred school that sees SOPs and rules as merely descriptive: simplified copies of the actual process that they (mostly inadequately) attempt to mirror. This view, centred on the ‘overflowing’ end of the Performativity framework, highlights the interpretative role of human agents who, in enacting rules, are able to modify or completely dismiss them (rules in this case are non-performative). While this extreme is also possible, this view does not account for the fact that, by incorporating beliefs into material devices, algorithms, procedures, and routines a model can have an effect ‘even if those who use them are sceptical of the model's virtues, unaware of its details, or even ignorant of its very existence’ (MacKenzie, Reference MacKenzie2006: 19). While formal procedures and rules can always – at least in theory – be worked around and dismissed, in practice they often play a role. Especially when embedded in artifacts or technologies such as software, they become pervasive, difficult to change or avoid, and can be more easily enforced. Thus, while possible in theory, mechanistic prescription and full interpretive flexibility are in practice two extreme outcomes; to the extent that a rule or procedure is entangled in a web of tools and organizational relationships, some level of Performativity is likely to be at play. Artifact-embedded SOPS and rules thus ‘do not simply describe, do not often prescribe, mostly they are performed’ (D'Adderio, Reference D'Adderio2008a: 784).

A further important implication of adopting a Performative approach is that representations, models, are not simply a description of something resting outside the market (reality), but constituent parts of it. In other words, they do not simply represent, they have an effect on the reality in which they are embedded (Geertz, Reference Geertz and Banton1966; Pickering, Reference Pickering1994; Ferraro et al., Reference Ferraro, Pfeffer and Sutton2005; Holm, Reference Holm, MacKenzie, Muniesa and Siu2007; D'Adderio, Reference D'Adderio2008a). They are not just passive ‘guiding principles’, setting the boundaries of what can be done and what cannot be done, as scholars have argued so far, but they contribute to shaping actual processes. Similarly, in D'Adderio (Reference D'Adderio2008a), procedures – as artifactual representations of routines – do not rest outside the routine as implied by contributions so far, but are important constituent parts of it (Figures 1 and 4c). This observation holds potential for further studies that capture the emergence of routines through the mutual shaping of artifacts and performances. Whilst this move is fully in accordance with Cohen et al.'s (Reference Cohen, Burkhart, Dosi, Egidi, Marengo, Warglien and Winter1996) focus on the interactions between procedures and performances, Performativity provides the theoretical tools and framework to characterize their mutual influence.

Figure 4. Artifacts from the periphery to the centre of routines

Artifacts at the centre of routines: co-evolution of artifacts, procedures, and performances through performative struggles

I propose that recent advances in Performativity Theory, as informed by both the Pragmatist turn (Schatzki et al., Reference Schatzki, Cetina and von Savigny2001; Callon, Reference Callon, MacKenzie, Muniesa and Siu2007) and the Distributed Cognition turn (Hutchins, Reference Hutchins1995; Callon and Muniesa, Reference Callon and Muniesa2005; MacKenzie, Reference MacKenzie2006) in Actor Network Theory,Footnote ⁵ and as recently imported into Routines Theory (D'Adderio, Reference D'Adderio2008a, Reference D'Adderiob), can provide novel and valuable insights into the role of artifacts in routines evolution. In particular, the new approach can provide the key to unlock the dynamics of interaction and mutual shaping between formal procedures (routines-as-artifacts) and actual performances (routines-in-practice). In order to achieve this goal, however, we need to go one final step further to account fully for the role of artifacts and materiality in routines.Footnote ⁶

A first observation is that artifacts do not simply provide information, they construct information and they classify it according to specific worldviews that belong to the agencies that have created or used them (Bowker and Star, Reference Bowker and Star1999). In other words, artifacts embody the ways in which specific agencies understand and organize their world (D'Adderio, Reference D'Adderio2001, Reference D'Adderio2003). The fact that certain agencies are more successful at this task than others, leads to asymmetric power distribution among agencies and their performative programmes (Callon and Muniesa, Reference Callon and Muniesa2005). According to Callon (Reference Callon and Callon1998), the more successful performative programmes are those that manage to enrol an array of materials and tools to create a world in which they can function. Within these performative struggles among competing agencies, artifacts thus play a key role.

An example of this is Pierre Bourdieu's analysis of the real estate market where the encounter between the seller and the potential buyer becomes a tug of war in which the former tries to impose his or her own calculative tools on the latter – often with success (Bourdieu, Reference Bourdieu2005). In these encounters, ‘whether it is the consumer hesitating between two packets of smoked ham or a couple anxiously following the real estate agent's calculations to assess their debt capacity, radically different values are confronted. When a compromise is reached, it has to be interpreted as a compromise not on values but on the instruments that calculate values’ (Callon and Muniesa, Reference Callon and Muniesa2005: 1239). Similarly, this author (2008a) has shown how Programme Management and Manufacturing functions at a vehicle manufacturer manage to enrol the software, and software-embedded SOPs, and thus to impose (at least partially) the culture, goals, and priorities that belong to their occupational communities, including their idiosyncratic ‘languages’ and worldviews (Dougherty, Reference Dougherty1992; Galison, Reference Galison and Biagioli1999; D'Adderio, Reference D'Adderio2001, Reference D'Adderio2003; Carlile, Reference Carlile2004). These examples demonstrate that, while a Performative approach brings artifacts to the centre of attention in Routines Theory, it does not do so at the expense of agency. Rather, it provides a means to capture the reciprocal influence of agency and artifacts. This is where the above-mentioned notion of ‘inscription’ becomes especially useful: once we accept that artifacts in general, and artifactual representations of routines in particular, as inscriptions, are not neutral, there are important implications for routines and routine-following.

In the examples above, agencies are able to use artifacts to impose their own classification and ordering of data, and process and decide what is and what is not important, what should be allowed or disallowed. Agencies thus can enrol arrays of artifacts and tools to help strengthen their performative programme. Artifacts, in the case of routines, can include procedures as well as a wide range of objects, tools, and technologies that can be harnessed to actualize their course or ‘make them happen’. As argued earlier, technologies and tools can help stabilize statements and classifications, thus giving them solidity – just like the uploading of the Virtual Prototype and the Engineering Parts List in software (in D'Adderio, Reference D'Adderio2001, Reference D'Adderio2003 respectively) and the tracing of price variation curves in Preda (Reference Preda, Callon, Millo and Muniesa2007). In these examples, software-embedded SOPs and rules, as inscriptions, ‘impose a principle of reality; they constitute an obligatory point of passage, a perfectly material reality to take into account . . . They are articulated to socio-technical agencements that produce the traces that they use to inscribe the world in which they are participants and on which they will, in turn, make possible to act’ (pp. 39–40).

While the above evidence could still be interpreted as a mere clash between conflicting interests, there is, however, more to it. To paraphrase Callon (Reference Callon, MacKenzie, Muniesa and Siu2007), the various organizational actors attempt to construct the world (or socio-technical agencement) they believe to resemble their own assumptions, views, and aims. To do so, they enrol a number of artifacts and tools to strengthen their influence. Confrontation therefore takes place not simply between different agencies but between different worlds that are struggling to exist, one at the expense of the other. The result of these struggles is that often none of the actors is able to take their programme to its conclusion, since often no one function is able to exclusively frame the process, as in the ‘engineering freeze’ example (D'Adderio, Reference D'Adderio2008a). Here Programme Management, IT, and Manufacturing partially succeeded in disciplining actions according to the software freeze rule – therefore enforcing the deadline – and yet the Engineering function retained some of their discretion to make substantial and late alterations. In the end, the evidence showed that each agency had to compromise and at least partially accept the other's programme, meaning that only portions of their world were realized. The world that ended up existing was thus, as in Callon's discussion (2007), a compromise, a patchwork containing elements from competing worlds.

An important implication of acknowledging the struggles among competitive performative programmes is that this lens can be used to uncover the dynamics of convergence and divergence between routines’ constituent parts or aspects, such as procedures and performances. In D'Adderio (Reference D'Adderio2008a, Reference D'Adderiob), evidence drawn from the ethnographic observation of the mutual influence between SOPs as artifactual representations and actual performances, showed how rules and SOPs contain abstracted, filtered, and reconfigured traces of actors’ knowledge and intent and therefore provide a strong example of distributed cognition in the Hutchins sense (MacKenzie, Reference MacKenzie2006; D'Adderio, Reference D'Adderio2008a). Thus, by focusing on SOPs, this author was able to show how performative struggles influence the direction of their interactions, i.e. performativity vs. counter-performativity), the intensity of these interactions (i.e. weak vs. strong performativity), and, ultimately, the persistence of the routinized pattern or deviation from its course (depending upon which programme – or configuration of programmes – manages to prevail).

The above considerations led to important conclusions, namely that, in order to understand routines evolution, we need to study how artifacts – and artifactual representations – themselves evolve. From the above discussion, it emerges that artifacts have a history which reflects the performative struggles in which they have been involved and from which they have emerged and evolve. In D'Adderio (Reference D'Adderio2008b), we see how the computer model, containing SOPs, evolves as a consequence of being incorporated in the struggles among different agencies that want to prevail. The artifact, as a result, comes to reflect the stratification of different (and possibly contrasting) organizational logics or rationales. Focusing on artifacts, and their evolution, thus can provide us with important new insights into routines and wider organizational dynamics.