Keywords: information infrastructure, actor-network theory, purity, fragmentation, Lotus Notes, rituals

1. Introduction

"Dirt is matter out of place" (M. Douglas)

The development, introduction and widespread use of a comprehensive information infrastructure is a resource consuming effort spanning years. It brings together and superimposes a range of technical, organizational, strategic and economic issues (Ciborra 1996; Kahin and Abbate 1995; Weill and Broadbent 1998). The infrastructure is negotiated across these issues rather than belonging properly to any one.

A commonly held ambition is to integrate the various applications, platforms and information systems that exist. The rhetorical thrust of the vision of tidying up — sweeping old mess aside — in order to avoid fragmentation and chaos is strong. It is one of the driving ideas behind the notion of a comprehensive and versatile infrastructure. But the fragmentation, of course, exist for good reasons and will not simply be swept away.

Our aim is to analyze an important but neglected aspect of the socio-technical process of establishing an information infrastructure. We empirically study and analyze how symbolic visions, images and icons of a non-fragmented infrastructure act as forceful "actors" that contribute substantially to the shaping of the technology (Pfaffenberger 1998; Swanson and Ramiller 1997). The necessarily open-ended nature of any information infrastructure — its generic character, or its ability to function as a boundary object (Star and Griesmer 1989) — implies that images and metaphors play a particularly productive role as a powerful ally (Boland and Greenberg 1992; Star and Ruhlender 1996). We are concerned with the following set of questions: how are our concepts of an orderly information infrastructure constructed; which mechanisms are at play in achieving the "natural" or taken for granted character of this order; what are the different forms of order in relation to information infrastructure; what are the implications of constructing order in a given way; who’s voices are silenced by the prevailing construction of order; and what are the implications for the development and use of information infrastructure?

The remainder of this paper is organized as follows. Section 2 outlines a theoretical approach to our study of order as an ongoing social, technical and symbolic process drawing primarily on actor-network theory (ANT) and supplemented with Douglas’ (1966, 1986) notion of "dirt" which we argue should be interpreted as "fragmented" in relation to information infrastructure. In section 3, the context of the case is outlined, namely the Norwegian based but internationally oriented oil company Statoil with about 17.000 employees. Section 4 describes and reflects upon the research methodology. Sections 5 through 8 contain the case. It is the historical reconstruction of the development and introduction of a Lotus Notes based infrastructure in Statoil during the year 1992 - 1998. We highlight how the fear for fragmentation was constructed, how it was circulated and reproduced and how this blocked for alternatives. In the analysis in section 9, we focus on three issues. Firstly, we discuss how a given sense of order came to be "obvious" or taken for granted. In social constructivist thinking, this is never obvious but rather a produced outcome of a process. In our case, an important element in this process was the rituals performed to confirm and re-establish the prevailing order. Secondly, we discuss how the constructed order failed to be neutral; it inscribed work routines and delegates roles to the organizational actors. Thirdly, we study the strategies at work in the punctuated episodes when the prevailing sense of order was challenged. This evoked, we argue, deep-seated sentiments about the alleged dangers of these alternatives. Section 10 offers a few concluding remarks about the more general implications of our study for the development and use of information infrastructure.

2. Theory

Developing information infrastructures has traditionally been regarded as a predominantly technical endeavor (Gunton 1989). This is no longer the case. It is a rapidly expanding body of literature addressing an array of issues of social, economical, institutional, political and strategic nature (Dutton 1997; Kahin and Abbate 1995). Most relevant to us is the subset of this literature focusing on how the development and use of information infrastructure is intervowen with social and strategic issues in business organizations (Weill and Broadbent 1998; Ciborra 1994, 2000; Davenport 1998; Earl 1996).

Our aim is to explore in some detail a key issue concerning information infrastructures, namely the background, forms and organizational responses to the fragmentation of information systems. As Davenport (1998, p. 123) underscores in relation to the particular instance of an information infrastructure of Enterprise Resource Planning (ERP) systems: "you first need to understand the problem [ERP systems] are designed to solve: the fragmentation of information in large business organizations".

This problem has been attempted framed as a technical one through the notion of data warehousing (Gardner 1998). The core idea here is to curb the increasing fragmentation by packaging it by presenting them through an uniform front end. There is, however, little attention paid to understanding the sources of this diversity. In much of the MIS literature (Broadbent and Weill 1998; see Ciborra 1997 for a critique), the problem of fragmentation is basically portrayed as the result of a lacking (adherence to a) IT strategy. The remedy, then, is a better "alignment" with the business strategy (Knights, Noble and Willmott 1997; Ciborra 1997). There is a tendency, however, to emphasize the planned and controlled aspect of such an effort vis a vis the more drifting (Berg, 1997; Ciborra, 1996, 1997), and improvised ones (Orlikowski 1996; Monteiro and Hepsø 2000). This downplays to the level of non-existence the organizational politics that account for why the problem of fragmentation is a real problem.

Within the field of science and technology studies in general and actor-network theory (ANT) in particular, there exist rich descriptions of the mobilization that underlie real implementation projects. Support needs to be gained and sustained through a mobilization process of actively enrolling actors, forging compromises and making the technology act as a spokesperson on your behalf (Bijker and Law 1992; Hanseth and Monteiro 1996; Latour 1996, 1999; Monteiro and Hepsø 1998). Our approach to the study of information infrastructure leans heavily on the elaboration of ANT by Star and Bowker (Bowker and Star 1994, 1999; Timmermans, Bowker and Star 1995; Star and Ruhlender 1996). The very activity of classification is an effort to establish categories that help produce order (Douglas and Hull 1992). Fragmentation of information systems represents a specific kind of "non-order". To understand fragmentation, it is essential to grasp how order is established through the construction of categories. In their continuing series of studies of classification schemes and infrastructures Bowker and Star identify a number of issues which are highly relevant to our study. In their historic study of the evolution of the classification of diseases maintained by the World Health Organization, Bowker and Star (1994, 1999) illustrate how coding and classification are anything but neutral. Interests are inscribed into the material of coding schemes. Timmermans, Bowker and Star (1995) studies how some aspects of work is made more visible than other by inscribing them into a classification scheme.

We fully embrace this perspective that emphasizes how technical efforts like that of establishing a working information infrastructure requires an ongoing mobilization process. Projects and plans need to be performed; they do not unfold autonomously by themselves. The focus on the central role of achieving order through socially constructed categories is very relevant to our analysis of information infrastructures. Few have put this point as clear as Foucault (1970). His famous citation (ibid., p. xv; also cited in Bowker and Star, 1999) from an ancient Chinese classification of animals makes the socially constructed character of the categories rather evident. Relative to our own sense of order, classifying animals into categories including those "belonging to the Emperor", "stray dogs" and those "drawn with a very fine camelhair brush" strike us as ridiculous. The upshot of this is to realize that order is never given but the outcome of an ongoing, socio-technical mobilization process. Our aim is to question the prevailing order and "discover that these orders are perhaps not the only possible ones or the best ones" (ibid., p. xx).

In our analysis of information infrastructure, we do not merely want to reiterate the, by now familiar, theme that technology is socially constructed (Orlikowski 1996; Walsham 1993). We want to explore this further by inquiring into how order is performed: how does it manifest itself; how does it gain support and become productive; what are its different forms in connection with establishing a working information infrastructure?

The analysis of how social order gets constructed has, of course, deep roots in the social sciences. Durkheim’s (1915) agenda was to analyze how social order was possible, that is, how moral feelings and social attachments could develop in diverse, social groups? A key mechanism, according to Durkheim, was through a variety of rituals. Durkheim identified religion as an exemplary case of how social practices produced feelings of solidarity. Hence, Durkheim did not grant religion a special status; it employed the same, basic mechanisms as other cases. Of special interest, is the way he describes how groups tend to perform rituals of respect for its sacred objects. Rituals are, according to Durkheim, the group worshipping themselves.

Douglas (1966) contributes with a highly relevant description of how order gets constructed. The reason why order and categories are so entrenched, Douglas argues, is that they are intrinsically linked to fundamental aspects of social life and sense making. It is what makes our world a comprehensive one. Order in social life is the existence of symbols that mark borders and systems of classifications. The symbols that rituals select and interpret, filter experience by communicating cultural themes and excluding unknown themes. The structuring of our experience evolves through a system of binary oppositions, like clean/ dirty, good/ evil, black/ white, inside/ outside. These categories make it possible for us to understand when we are entering or leaving a symbolic border. We are socialized into thinking that things are dirty in themselves, that is, that dirt is an attribute of an object. But if this was really the case, Douglas asks, how may we then account for the fact that most people regard it as dirty to put your shoes on the table but not on the floor? Douglas’ account, in essence, is that it is a misconception to think of dirt as a given category and an attribute of an object. An object is dirty, according to Douglas, when it transgresses the socially constructed categories of order.

Having elaborated how Douglas and social studies of science pave the road for an empirically open notion of order, it remains to spell out how this ties in with the development of information infrastructure. The link, we argue, is through the status of fragmentation::

Douglas essentially opens up for a relative notion of dirt. Dirt is not an attribute of an object but rather relative to socially constructed categories as social studies of science and technology have since iterated so convincingly (Berg 1997; Bijker and Law 1992; Bowker and Star 1994; Latour 1996; Timmermans, Bowker and Star 1995;). Hence, neither categories nor conceptions of what represents order are neutral. Order is neither given nor neutral.

Fragmentation of an information infrastructure represents the opposite of order as "uncleaness or dirt is that which must not included if a pattern is to be maintained" (ibid., 40). Maintaining an orderly information infrastructure then translates into the ongoing effort to keep it pure, that is, tidy or clean, by preserving the given categories and avoiding the transgression fragmentation would produce.

In what follows, we will extend the insights of ANT where technological efforts are recognized as ongoing processes of mobilization by interpreting Douglas’ emphasis on mundane, everyday rituals as important elements of this mobilization process underlying the contruction of an orderly information infrastructure. In short, we extend the socio-technical perspective of ANT with stronger symbolic and ritual elements borrowed from Douglas’ analysis of dirt.

Statoil is a young company. Founded in 1972 with only one employee, it has since grown to a $ 0, 7 billion operations profit enterprise with over 17.000 employees in 25 countries. Still, the major activities are located in Scandinavia and Northern Europe. After Phillips found the substantial oil field Ekofisk on the Norwegian continental shelf in 1969, there was an ongoing political debate in Norway about how to organize oil production. The decision to establish a new company, Statoil, with the Norwegian state as sole share-holder, was reached in a consensus voting in Parliament in 1972. It was anything but obvious. There were strong lobbying, also by the Prime, let Norsk Hydro take charge (Internal newspaper, Status 16. September 1992, pp. 8 - 9). The argument for Norsk Hydro was safer to trust a company that already had proven capable and competitive in energy production than establish a new organization from scratch.

Immediately after Statoil was established, the oil crises struck. For Statoil, the crises had two important consequences. First, it shifted power and control out of the hands of the "seven sisters", the dominating oil companies, and in the direction of the oil producers, primarily OPEC, but also Statoil and Norway who preferred to stay out of OPEC but with the status of an observer. Second, and of crucial importance, it paved the road for a significant rise in oil prices throughout the 70s and into the 80s. Without this rise, the development of the inaccessible oil fields on the Norwegian shelf, deep under the turbulent Northern Sea, would simply not be cost-effective.

Statoil was the product of negotiations in Norwegian politics. From the outset, Statoil relied heavily on an array of different favorable measurements aimed at tilting the competition. This reliance on political negotiations has made Statoil particularly responsive and sensitive to signals in the political environment.

The broad, political obligation in Norway to systematically favor Statoil has gradually faded away as the tide of liberalism rose. This has step by step forced Statoil through a metamorphosis. It has through a sequence of small steps been transformed into an internationally competitive oil and gas producer.

There are a number of organizational actors involved in our case study. Their actions and interests will be elaborated further below. To assist the reader in keeping track of these actors, we provide a crude description of them in table 1 that may be consulted at need.

The methodology employed here is interpretive. Following Walsham’s (1993, pp. 4-5) definition, we were "aiming at producing an understanding of the context of the information system, and the processes whereby the information system influences and is influenced by the context". The set of seven principles outlined by Klein and Myers (1999) for assessing interpretative field studies of our kind is instructive in making explicit our approach and reflecting upon it’s strengths and weaknesses. We discuss these principles by elaborating those most relevant to our study.

The first principle deals with the hermeneutic circle, that is, how our understanding of the whole are linked to our understanding of the parts and the second principle deals with the historical background. In our study, we have combined these two in a manner that needs to be explained. We have worked out a historical reconstruction of the whole process around the introduction of Lotus Notes in Statoil during the years 1992 - 1998. This approach is strongly motivated by the assessment by Mason, McKenney and Copeland (1997, p. 258) that "MIS researchers, for the most part, have not sought to identify fully the broad socio-economic conditions of continuity and change that accompany the use of information technology". Rather than employing their phase-oriented "cascade" model of technical innovation, we lean heavily towards ANT influenced analysis of socio-technical mobilization processes as outlined in section 2. This implies relaxing the notion of phases in innovation and allowing a wider spectrum of organizational actors. As a vehicle in this historical reconstruction, we developed a scheme where we added a number of categories with dated episodes and trends during the years 1992 – 1998. These categories are: external conditions, prevailing management strategies, major IT/IS projects, the rise and fall of key organizational actors, important organizational development projects in addition to the dates of important events in the technological solution directly connected to the establishment of the Notes based infrastructure. By using the time dimension as the "anchor" is was possible to see how the development of the Notes infrastructure was connected to a number of other company efforts. For instance, spirits of the time changed considerably during this period including: Total Quality Management, ISO 9000, Business Process Reengineering and knowledge management. Each of these gave the development of the Notes infrastructure support from different rhetorical sources. This overall map functioned as a navigating and sensitizing device for making sense of the smaller episodes in the hermeneutic circle.

The third principle is that of the interaction between the researchers and the subjects. Embedded in this principle is the critical reflection about how the data was socially constructed through the interaction between the researchers and participants. Our access has been facilitated by our relation to Statoil. One of the authors (the latter) has worked for Statoil the last seven years, he has worked with Notes implementation from 1994-1998 outside the official domain of SData. This has given him detailed information about the issues, people, data sources and the context under investigation. The fact that he has worked in a research unit (KOT) that several of the informers historically describe as a competitor to SData, makes him biased. Still, it would be difficult for an outsider to develop the same depth of understanding. We have dealt with this bias in two ways. First, the relation between the Statoil internal and the external (former) author must be seen as dialogical, in the sense that the external author played the role of the "devil’s advocate". This Statoil external author was granted an office space, an access badge and a Lotus Notes account and spent on average two days a week in Statoil over a period of five months. Statoil has traditionally been relatively protective towards outsiders. The fact that the authors were free to wander about and make appointments — symbolically gestured by the existence of a Statoil based e-mail address — have greatly facilitated our ability to select and identify interesting sources of data rather than being closely steered. Second, since one of the authors knew the history as a direct participant in the setting he might have unconsciously, based upon his pre-conceptions looked for data that supported his "prejudices". We tried to address this by seeking to validate our ‘findings’ and discuss our account of the case with involved actors, and partly by relying on varied and independent sources of data that the external author collected and analyzed. Another principle we followed was that nothing of the internal author’s memory or past field notes should be used if it was not guaranteed by additional source material. Digital data sources related to the issue under investigation was considerable (see table 2). There is an extensive electronic achieve (Elark) which contains all official Statoil reports in addition to selected contracts, e-mail discussions, memos and project documentation. There are also a large number of Lotus Notes discussion data bases, newsletters, detailed project archives, budgets and various forms of corporate presentations (slides, brochures and folders). We have presented our findings in 3 meetings with major key organizational actors (SData, KOT, INF/ KIT) with about 30 participants from several organizational actors including the research center, Exploration & Production and SData. Written reports have also been circulated and resulted in both e-mail and oral feedback. In addition to this we have been engaged in participatory observation by taking part in project meetings, informal discussions and coffee breaks. We have conducted 20 semi- and unstructured interviews lasting one and a half to two and a half hours (for more details and categorization of the information, see table 2).

The fourth principle deals with abstraction and generalization. This is already presented in section 2 where we outlined our use of ANT and Douglas’ work on order and categorization. The fifth principle is that of dialogical reasoning, with requires openness to potential contradictions between the dialogical preconceptions guiding the research design and actual findings (the story the data tell) with subsequent cycles of revision. This is covered in our reflections immediately above on principle 3. The sixth principle is that of multiple interpretations. We presented our interpretations to key organizational actors in meetings and on informal occasions during fieldwork. Most actors agreed with our interpretations, though less so for SData. To meet this ambiguity, parts of the story were rewritten to incorporate SData perspectives. In other instances, we backed our interpretations with more detailed data that could better support our interpretation. The seventh and final principle is that of suspicion. It requires sensitivity to possible biases and systematic distortions in the narrative collected from the participants. ANT gives a good framework to encourage critical thinking and reveal taken for granted assumptions (social facts). The digital material that was archived in Statoil’s Notes databases provided "raw" material that could be interpreted as texts. Of special importance was the archived information from the electronic communication in Notes related projects (see table 2). This information was of importance in validating the feedback of the informers.

During its relative short history of existence, the revenues of Statoil have fluctuated considerable. This has created a shifting environment for how to conceive of the role and importance of IT. At the point just prior to the early establishment of a Lotus Notes based communication infrastructure in 1992, Statoil experienced (relative) economic hardship. The post golf war period after 1990 led to a recession in the oil industry with falling oil prices and dollar exchange rates. In Statoil, the Exploration & Production unit, the key contributor to company profits, painted a dark picture of the future.

The governing principles of the sequence of re-organizational efforts that were spawned in response to the sense of hardship identifies "standardization of components and systems" together with the "use of vendor and industry standards (...) and a minimum of in-house development" (Internal slide from presentation of one of the reorganization projects). Hence, investment into IT was largely viewed as any other source of cost generation and hence a liable candidate for rationalizing and cost-cutting because at this time around 1992 "IT was only considered an expense" (Intro1).

The IT infrastructure in Statoil at this time was highly fragmented and diversified. There was no wide-area network accessible to PCs. In the early 90s, there was a long discussion about LAN platforms in Statoil. There were in total seven different LAN solutions including IBM’s Token ring and an implementation of Ethernet by Novell (Manager2). In fact, corporate IT had in 1990 advocated Token ring as a corporate standard without achieving much impact. Similarly, there was a later decision to make LAN Manager OS/2 a common platform. What tilted the balance in the end towards a Novell based Ethernet, was that the core part of the IT department located at the corporate headquarter lobbied for this (Manager2).

The key mechanism for achieving a more cost-effective use of IT was through standardization of applications, it was "the thing to do" (Intro2). There was a widespread and general consensus about the need to standardize the existing jungle of office tools as "there existed more than ten different word processors at the time. (...) Everybody [out in the divisions] were happy that there would be only one tool (...) everybody applauded the decision."(Intro1).

This lack of uniformity, this frustration with fragmentation prompted a quick decision on a uniform office tool. The choice of the Lotus suite was made on price solely. Lotus, in a desperate move to carve a niche vis a vis Microsoft, made Statoil an offer they could not refuse.

The pressure outlined above for rationalizing operations, increasingly efficiency and cutting back on costs was immediately translated into a centralization of the organization of the IT services. The argument for the centralization was economy of scale arguments — or, in the language of earlier days — to "harvest the benefits of mainframes" (Manager1). In other words, the task of uniforming the technology by settling for Lotus was regarded as a symptom of the more general problem of a too decentralized decision and budgeting responsibility for IT investments (Manager1). Hence, the centralized SData (see immediately below) was expected to function as an arena for broad consensus for IT investments, a consensus unattainable before due to the decentralized organizational structure because "without the centralization of SData, it would be impossible to reach decisions" (Manager1).

The subsequent centralization of IT that established Statoil Data (SData for short) in April 1993 was neither smooth nor straightforward. Beyond the turbulence within SData itself, the real issue was a fundamental reorientation of SData. The real challenge was to transform SData into a business and market oriented organization, to change from "plan economy to a market economy" (Manager2).

SData needed to change substantially. They were vulnerable and had to re-establish trust with their biggest and most important customer, Exploration & Production. They were challenged by the newly formed group, KOT, dealing with communicative aspects of IT. SData was in bad need of an icon to symbolize the new era. SData worked hard to make Lotus Notes such a concerted effort. This fitted well with the challenges identified in a strategy document by corporate IT strategy from 1994 as:

SData lobbied for Lotus Notes as a solution. For this to work, SData needed to establish an accompanying wide-area network as a communication platform.

The Lotus Notes infrastructure that SData attempted to establish was packaged together with two other components, namely the standardized suite of office tools from Lotus and a PC based, wide area network that allowed the PC to communicate across the geographical locations of Statoil in and outside of Norway. This PC based wide-area network was called I-net. I-net represented a massive investment for SData. In combination with Notes, I-net was — and still is — the gem of SData in the sense that it is a vital, corporate asset entirely under the control of SData. The control over I-net allows SData to act as a "gatekeeper" (Latour 1987).

Hence, Notes was not introduced as a more or less isolated artifact, it was "bundled", packaged or aligned with existing and new elements such as I-net. The establishment of an information infrastructure always requires this kind of careful alignment. This alignment functions as a confirmation of the ordering that is already in place. The prevailing order needs to be kept alive.

By the end of 1994, there were 10.390 Lotus users but only about 4000 Notes users (Notes newsletter, TeamIT 221294).

The facade of Notes was undoubtedly e-mail, but in the background a more versatile use was prepared and encouraged as "we had a clear policy about allowing the users to develop their own Notes applications" (Intro2). The Notes introduction project team lobbied hard towards Exploration & Production, their most prominent customer, for a wider diffusion of Notes. Exploration & Production kept a short list of core systems that they paid well for. The Notes introduction was becoming increasingly important to SData, in commercial terms both even more so as a sign of acknowledgment. In a situation were SData still was working to acquire a sense of confidence, this was vital because "for SData, Lotus Notes was important, very important" (Intro2). After a series of rejections, Exploration & Production in September 1994 finally agreed to include Lotus Notes into their core portfolio of systems thus financially securing the situation of SData.

SData was then able to turn to the vast number of small details that were needed to glue Notes together, to facilitate further spreading. During this period from 1994 – 1996, the number of Lotus Notes users rose from about 4000 to about 14000. SData focused on supporting the management of users, that is, creating, deleting and moving of users, changing names and administration of the mailboxes. Filling in the gaps in the infrastructure, providing the invisible but necessary parts of the infrastructure is a recurring pattern in the development of an infrastructure (Latour 1996; Monteiro 1998).

The mobilizing vision behind the Lotus Notes infrastructure in Statoil was to provide a smooth medium of communication where information could flow effortless. A taken for granted assumption, then, was a perfectly uniform medium of communication. More specifically, the Lotus Notes applications in Statoil were not tailor-made. SData instead promoted six standard applications including calendar, meeting room reservation and archiving. It was an ongoing effort by SData to keep the threatening fragmentation of Lotus Notes applications at a distance which tailor-made Notes applications represented. An illustration of this is the attempt made by a member of one project within Exploration & Production to use a Notes application not among these six, standard ones. Instead, he wanted to use a Notes application he knew about from colleagues working in a different project because:

In their response to this inquiry, the SData Lotus Notes introduction team attempted to persuade them to reconsider using the relevant, standard application as otherwise "the project must cover the expenses of this adoption themselves" (email 280995 archived in Elark).

The rapidly growing Lotus Notes infrastructure that emerged which comprised also the Lotus office tools suite and the I-net PC network had an administrative bias. The actual use of Notes to support communication and coordination during the early phases of oil production was very slow. The people involved here are geo-physicist of various backgrounds (see table 1). The work consists of a variety of tasks including drilling, seismic exploration, analyzing and visualizing drilling data, simulations and modeling of the reservoir.

Given the nature of oil production, one might perhaps imagine that the competence of geo-scientists was highly valued, that their requirements and needs were swiftly acted upon. This, however, is clearly not the case. The geo-scientists are surprisingly low in the hierarchy and have more difficulties than one might expect in getting their voices heard. In the words of one of them, "it is the business students [Norwegian: blårussen] that govern" (User2). The largest union for geo-scientists (Norwegian: Norsk Sivilingeniør Forening, NIF) has lobbied internally in Statoil for upgrading their esteem (Status newsletter, Status, 5^th of January 1992). Still, "[T]here are quite distinctly two ladders of carriers, one for geo-scientists and one for management" (Manager5). Without exaggerating the level of conflict, a NIF union leader still points out the "intrinsic conflict of interests between the geo-scientists and management" (User9).

In relation to the emerging and prevailing sense of an orderly Notes infrastructure, the geo-scientists’ applications, work-routines and competence simply did not fit within the constructed categories. The tools and applications needed to support their work have throughout the 90s been Unix. In order to perform their work, these Unix based tools are essential. Even today, "it is unthinkable" to migrate these tools to PC due to performance considerations (User1). The current ratio between Unix users and PC users in Statoil is about 1:15.

The introduction of the Lotus suite and later Notes was made smooth by "promising that Unix clients would also be available" (Manager3). Hence, early in 1994 a small group of SData people was to explore the possibilities of finding or developing a Unix emulator because "in Statoil there is a pressing need to run PC applications from Unix work stations" (Notes introduction newsletter, TeamIT 280194). Only later did it become clear that these promises would never be kept, that Lotus simply was not going to develop Unix versions.

The analysis, simulation and modeling based on the drilling data take place in project teams. All the drilling data for all the oil fields of Statoil originate from a Unix based data base which has been revised and upgraded a number of times during the last 10-15 years. Statoil’s competitive edge is closely linked to its ability to exploit and continue to refine this database. The tools extract data from the data base but these tables, graphics and models, then "have to exported from Unix" for further elaboration into reports, slide presentations and archives (User4). The tools for producing this documentation are not Unix based but PC based implying that everything has to be move across the two platforms using file transfer services. This moving around of data is not always as smooth as it is supposed to be as "it is a lot of fuss with Postscript files, they cannot be read on a PC when exported from Unix" (User4). This prompts an extensive repertoire of work-arounds to cope with these problems. Some keep a bag of tricks that they have accumulated. To illustrate, a file "washing" program was picked up by one engineer from a friend and colleague working for another oil company, Norsk Hydro; another asks favors from people he knows in the graphics group and yet others use "the people I know" (User1).

In a number of ways, the pressure for opening up and orienting Statoil more towards the outside world was building up from the mid 90s. This came partly from fairly general trends like the rapid folklorization of Internet and Web in media. In addition, the oil industry underwent important restructuring during the period 1995 - 1997. There was a growing awareness about the need to communicate with external partners and subcontractors.

In terms of technological infrastructure, the situation in Statoil was still characterized by fragmentation, by multiple communication standards and platforms (Internal Statil report: Statoil’s integrated network in year 2000, April 1997, pp. 35 - 37). There existed a FDDI fiber optical network at the corporate headquarter. Between major sites, a number of different wide-area network solutions were used including: leased lines, ISDN, Frame relay and ATM. Communication with Statoil sites outside the Norwegian main land was by Frame relay or satellite. LAN communication was dominated by 10 or 100 Mbps Ethernet running Novell IPX, TCP/IP and some Apple talk. There still existed Token ring segments in some locations.

With regards to the evolving Notes infrastructure, the over-arching trends got translated into a simple question: was Notes an appropriate infrastructure to meet these challenges? There was at this time no obvious way to align the new requirements concerning opening up towards the outside world with the existing Notes infrastructure.

The strategy used by SData was one of marginalization. The proponents and arguments behind, for instance, Internet and Web were attempted sidelined by presenting them as misguided. Hence, the proponents of Notes tended to downplay the significance and substance of the objections towards Notes because "the advocates of Internet are those who do not know how good Notes is with regards to Internet" (Intro1). The heart of the problem, the accusation that Notes was a closed system and hence inappropriate when Statoil was to open up to the world, evaporated with Lotus’ Domino servers capable of gatewaying between Notes and Web. It was accordingly presented as a "misunderstanding" that Notes was closed as "Notes has tools for SQL queries together with the new Domino servers" (Intro1). And as a consequence, "the controversy have died out" (Intro2).

Statoil has traditionally been fairly closed towards the outside world. Unix users have had access to e-mail communication with external partners from the early 90s. Memo, the corporate wide e-mail system introduced in Statoil in the 80s, was only for internal communication. With the establishment of an X.400 email gateway, Memo and Notes mail was able to be used for external communication from 1995. In 1996, a Notes based simple mail transfer protocol (SMTP) server made Internet mail directly available from Notes. Non e-mail communication with the outside world, however, arrived rather late in Statoil, that is, for PC users. Unix users had access to a broad range of services such as archie, ftp, telnet and Web. Web was for this reason shrugged off as "a Unix thing". PC based Web browser were allowed only from late 1995, and only in response to a formal application. Only in January 1998 was it allowed to browse from a PC without special permission.

9. Analysis: establishing and maintaining order

The most fundamental, shared insight provided by (among others) the discursive analysis of Foucault (1970), Douglas’ (1966) analysis of the socially constructed categories underpinning our sense of order and purity and actor-network theory (Bowker and Star 1994; Latour 1996; Law 1991), is that although we take our notions, beliefs and knowledge very much for granted, they are contingent. But this contingency is not apparent, hence requires an effort to unpack. In relation to images of order in information infrastructure, it is striking how seemingly self-evident this is. It is common wisdom that collections of information systems in general and an information infrastructure in particular need to be tidy, well-defined and with clean interfaces to surrounding services and modules (Gardner 1998; Weill and Broadbent 1998). When the Lotus suite of office tools was introduced in Statoil, it was simply so obvious that order in the form of "standardization" was the answer that critical remarks appeared laughable. There are, of course, a number of good reasons to strive for this form of order in an information infrastructure: it could make it more transparent, easier to grasp, simplifying maintenance and further development and so forth. Empirical evidence (Ciborra 1996; Hanseth, Monteiro and Hatling 1996), on the other hand, strongly suggests that this level of order and stability is unattainable — it is a beautiful dream — and probably neither functional nor cost-effective (Berg 1997; Hanseth and Monteiro 1997; Williams 1998).

Hence, we want to pose the following set of question: how was it that order got constructed in Statoil in the form of a standardized Notes solution; how is it that it is possible to mobilize such a strong support for avoiding fragmentation of the information infrastructure; what are the mechanisms at play which produce such an effect and how does order get socially constructed as self-evident? There are, as we see it, several sources which we discuss in turn.

Clearly, the image of tidying up, for instance, the "jungle of office tools" (User1) acted as a forceful organizing vision in the sense of bringing the community of users together and developing a sense of shared purpose and destiny (Swanson and Ramiller 1997, pp. 460 – 461). The portrayed economic hardship in Statoil in the early 90s also had a strong disciplinary effect. Hence, when the request cited above in section 7 about customizing a Notes application was answered by insisting that they then had to bear the full costs themselves, this was a fairly effective way of saying no.

What takes place, then, was that the one and same problem was addressed from three different angles: cutting costs (external pressure) leading to curbing the mess IT (the technical side) which subsequently led to the centralization of the IT department (the organizational aspect). This trinity of cost-cutting, tidy technology and centralized organization was extremely strong. So strong, in fact, that no opposition was voiced. The three of them were mutually reinforcing, or in an ANT vocabulary, they were aligned to form a stabilized network.

In the work of Douglas (1966), there is no question why efforts of tidying up touch upon such deep feelings: preserving a sense of our order is intrinsically linked to how we make sense of our (from the outset chaotic) everyday world. Avoiding chaos in this way is a basic, human instinct according to Douglas. Hence, we have to turn to ways in which order is performed. Mundane, everyday rituals play an important role in re-establishing and re-confirming the prevailing order. They keep it alive and vivid. Examples of such rituals in our case are project meetings in SData discussing the further development of Notes, the launching of campaigns for disciplining the use of email and the upgrading to newer versions of Notes and I-net. These rituals functioned as a stabilizing principle as they naturalized order and classification (Douglas 1986).

Other scholars have argued that there is a strong, rhetorical device at work which explain why organizational members develop taken-for-granted beliefs (Meyer and Rowan 1977). These beliefs function as myths and exercise a strong influence. Alvesson (1993) echoes this and suggests that organizations cultivate symbols to construct positive images of themselves. In Statoil, it is reasonable to read SData’s commitment around 1993 – 4 to Lotus Notes as, at least in part, an expression of their attempt to construct a more fast-moving, responsive image of themselves vis a vis KOT.

Moving forward from the analysis above about how the given order cast as a standardization regime got constructed as "obvious" or "natural, let us now turn to the implications. Actor-network theory, as outlined in section 2, instructs us that no given order is neutral. It "embodies and inscribes work" by privileging one way of working on behalf of another (Bowker and Star 1994, p. 187), it makes one set of organizational actors more visible and influential on behalf of others and it acknowledges one kind of knowledge and practical abilities over others. This is, of course, the reason why an analysis of the social construction of order in an information infrastructure is interesting. To merely document that order is not given is neither original nor too difficult. This is but a stepping stone to analyze the implications of a given order and, to reiterate Foucault (1970, p. xx), that it is "perhaps not the only possible [one] or the best [one]". What, then, were the implications in Statoil of constructing an orderly infrastructure as a non-fragmented Lotus Notes based one: what kind of work-routines had to be augmented and needed work-arounds to fit (Gasser 1986), who became invisible and what type of work and knowledge was less valued?

We suggest that the geo-scientists is a group in point. Their existing information infrastructure was down-played, their work had to be extended with a rich set of work-arounds and their knowledge and professional expertise was challenged. We have already illustrated the former two of these, namely how the Unix based infrastructure of the engineers was poorly supported and how the fragmentation across the Unix/ Notes border spawned a set of work-arounds. To illustrate the latter, how their knowledge and expertise was challenged, consider the controversy over the interpretation of drilling data.

Generating, analyzing and interpreting the drilling data is, of course, central to the operations of an oil company like Statoil. There is a continuous negotiation over who is to be in charge of this operation. The controversy circled around who should rightly perform this task and what kind of skills were required to do it. In short, the geo-scientists argued that their competence was needed to makes sense of the data, that the data could not be interpreted out of context nor by persons without a sufficient technical background. Management, on the other hand, argued that what was needed was assessment, prioritizing and decisions that cut across the details of the single drilling. In other words, that the geo-scientists focused too much on the details and forgot the larger picture. This negotiation over the influence and gatekeeping function over a key activity for Statoil unfolded largely as a controversy over technology as illustrated below.

The core data base, Sphinx, which contained all Statoil’s drilling data was an old data base which had been migrated and revised on several occasions: from an IBM platform to Unix, from a text based interface to X-windows, from X-windows to Motif. The historical heritage of oil production was in important ways inscribed into its core. In the early years of Norwegian oil production during the 60s and early 70s, there was a massive import of US based oil competence, workers and technology. This is in numerous ways evident even today. Sphinx was an example of this import. The core of Sphinx, its organizing principle, that is, its indexing, inscribed the traditional way oil searching and production was done: by drilling strictly vertical, land based holes (Bowker 1994). This implied that for every hole, the length uniquely identified a location. This inscription had been challenged by the searching on the Norwegian, sea based shelf. Over the last years, the ability to steer the drilling sideways, even upwards, undermined this design assumption inscribed into Sphinx. It implied, among other things, that there could be two locations at the same vertical depth. This reinforced the need of making sense of the data in Sphinx as "it is important to be technically skilled when interpreting the data" (User4). This skill was supplemented by the exploitation of the tightly knit informal network which existed among the different clans of geo-scientists. Whenever there was problems in interpreting the data, "I call the people I know" (User1) or consulted their own, private copies and note pads from earlier projects (User4). Accordingly, the data in Sphinx was but the espoused version of the interpretations that need to be supplemented by locally informed insights as "I hang on to my local copies as well" (User 4). Hence, the way the documents were collectively produced left a much richer material than what got "dumped into the electronic [Notes] archive" in the end (User4). This gave rise to the forceful myth about the poor data quality of Sphinx, namely that it was basically sloppiness by the geo-scientists that account for the difficulties in extracting and interpreting data from Sphinx (Garfinkel 1967).

There were moments when the very construction of order got challenged and the seemingly "natural" order was questioned. When it no longer was obvious, how was the prevailing order defended and how was opposition met? In addressing this, we analyze the controversy which surfaced 1995 – 1996 around Web.

The threat to the Notes infrastructure from Web was quite real. What the outcome would have been had not Statoil been saved by the bell through the introduction of the Lotus Domino servers which acted as Notes/ Web gateways was uncertain:

Internet had been of marginal importance to the company up till this point in time. Unix-based specialists and a few people at the research center or SData had used it regularly since 1993. But the general potential of this new phenomenon was first realized via the media from late 1995, increasing steadily in 1996 with the folklorization of Internet. With this, a general change of spirit occurred focusing on IT as an enabler.

The mobilization of Web as an alternative to Notes was not merely in the form of "pure" technology. Also organizational actors moved in order to enroll the Web as an ally. Especially the media and information unit (INF) was active here. As they were delegated the new responsibility of Web editing, they felt a special need to become visible. As is the case in many places, the most enthusiastic proponents of Web are initially found outside the traditional IT department. In a memo outlining a new project, a project leader in the media and information unit described the situation as follows:

There were distinct and conflicting views — and still are — about whether Domino represented a sufficient strategy to address the requirements on openness as "many are still very skeptical to whether Domino is sufficient" (Manager1).

Having worked hard for the emerging Notes based infrastructure for more than four years, SData had invested a substantial amount of sweat and prestige into it. The Notes infrastructure had, to use actor-network theory (Callon 1991; Hanseth and Monteiro 1996), acquired a certain irreversibility. It had become difficult to undo it, as Douglas (1966, p. 36) explains, because "as time goes on … we make a greater and greater investment into our system … [and] uncomfortable facts which refuse to be fitted in, we find ourselves ignoring or distorting so that they do not disturb these established assumptions."

Given this inertia (Bowker and Star, 1999; Star and Ruhlender, 1996; Hanseth, Monteiro and Hatling, 1996), what kind of strategy did SData (unintentionally, most likely) pursue? Again, Douglas’ analysis of purity contains the essential insight. To mobilize support for the orderly, standardized Notes solution, all you have to do is to construct your opponent as "dirty" because "a polluting person is always in the wrong …[and] unleashes danger for someone" (ibid., p. 113). We argue that this is exactly the (unintentional) strategy SData pursued when they forcefully argued that the existing Lotus Notes solution represented order, hence purity, and that the introduction of Web represented fragmentation, hence dirt, because:

This strategy of describing the alleged dangers of fragmentation was quite explicit. The citation above from a strategy document was made productive by circulating and emphasizing it in subsequent email discussions (email archived in Elark, 031296). This echoes Douglas’ analysis of the dangers of giving sacred objects (here: Notes) a profane status. Hence, any attack on the sacred will evoke strong reactions to defend it.

Still, the present situation (2000) is characterized by a certain stability in the Domino-based compromise. This, interestingly enough, is sustained by the increasing invisibility of the original Lotus Notes infrastructure. With the widespread use of Web browsers, the majority of users need not (and, indeed, do not) know whether the information originated from Notes data bases or not.

The basic moral of social constructivism is one of caution. Taken for granted beliefs, technological solutions and sense of order are not "given"; there is a process of naturalization that produces this effect. In this sense, our study has reiterated an old moral. Still, we believe that the challenges with information infrastructure require a grasp beyond the programmatic stipulation that they are "socially constructed". For users, policy-makers, managers, designers and researchers it is of vital importance and relevance to dig deeper into how this construction unfolds. Our study is aiming at contributing to one – neglected but essential – aspect of this process, namely the symbolic, social and technical mechanisms at play in constructing an orderly information infrastructure. This is difficult to get at due to the deep-seated nature of our sense of order. Still, there is no way around a critical assessment of order when establishing a working information infrastructure; it is cumbersome but necessary.

We have demonstrated how the taken for granted sense of order in the establishment of a working information infrastructure in Statoil – standard Notes applications, standardization on Lotus products, standard communication protocols – was in fact not obvious; it got constructed as obvious. The way this hampered alternative design and created work-arounds for some of the actors in Statoil, strongly suggest that designers, managers and users alike need to critically assess "obvious" assumptions about what constitutes an orderly infrastructure.

Eric Monteiro,

(corresponding author)

Dept. of computer and information science, Norwegian Univ. of Science and Technology (NTNU), N-7491 Trondheim, Norway

phone: +47 7359 6751, fax: +47 7359 1733 (eric.monteiro@idi.ntnu.no)

Vidar Hepsø,

Statoil Research Centre, Statoil, Norway (vihe@statoil.no)

We are grateful for the support and constructive criticism through our participation in the InfraGlob project led by C. Ciborra together with K. Braa, A. Cordella, B. Dahlbom, O. Hanseth, J. Ljungberg and K. Simon. The numerous comments and objections from Statoil employees have been essential. We have also benefited from comments from O. Ngwenyama and D. Greenwood. Elements of this work have been presented and discussed at the workshop Cultural politics of technology, the Information systems research in Scandinavia and the working conference IFIP WG 8.2 and 8.6. This work has been partly sponsored by the Norwegian Research Council’s program on social aspects of information infrastructure (SKIKT).