Eric Monteiro

Dept. of computer and information science,

Norwegian Univ. of Science and Technology (NTNU)

and

Dept. of Informatics, Univ. of Oslo

ericm@idi.ntnu.no

Keywords: research methodology, actor-network theory (ANT), interpreting information systems

1. Introduction

As we approach the closing of the millennium, it is well established that systems development extends beyond the purely technical domain comprising of coding, modeling, testing, verifying, integration and version control. During the last couple of decades a steadily growing body of scholarly work emphasizes – conceptually as well as empirically – that the development and use of information and communication technology (ICT) involves an array of non-technical issues as well. Scholars underscore how systems design includes political, organizational development, group- and communicative dynamics, ethical and cooperation and/or conflicts of interests (for an overview, see Dunlop and Kling 1991).

This opening up towards non-technical issues has resulted in a massive import of theories, frameworks, methodologies and concepts from other disciplines. The relationship between technology and society may be conceptualized in many ways. We embrace the fairly widespread belief that IT is a, perhaps the, crucial factor as it simultaneously enables and amplifies the currently dominating trends for restructuring of organizations (Applegate 1994; Orlikowski 1991). To be instructive in an inquiry concerning current organizational transformations, one has to supplement this with a grasp of the interplay between IT and organizations in more detail. We need to know more about how IT shapes, enables and constrains organizational changes.

Providing an extensive account of this import is prohibitive but relevant examples include: working out alternative, philosophical foundations for systems design (Hirschheim, Klein and Lyytinen 1996), borrowing from ethnography (Suchman and Trigg 1991), historical analysis (Mason, McKenney and Copeland 1997) or general social theories (Orlikowski and Robey 1991).

The resulting situation, then, is that there is a broad consensus about the need to extend systems design beyond technical considerations. But the subsequent import from other disciplines has produced a collection of notions, theories and perspectives that are far from self-explanatory. The pressing question, from the point of view of systems design, is how to operationalize and make use of it all. One attempt to translate this into practice in systems design is systems thinking, in particular Soft Systems Methodology (SSM) (Checkland and Scholes 1990). SSM underscores how technical and non-technical issues mesh. The problem, however, is that SSM has little to offer about the dynamics, that is, how the development process unfolds over time (see Guidelines 6 – 10 further below).

The aim of this chapter is to outline one, specific conceptualization of the socio-technical called actor-network theory (ANT). Based on an appropriation of ANT to systems design, the aim is to spell out a few methodological guidelines for systems design. It is necessary to elaborate this. ANT does not, as explained in more detail further below, present itself as a methodology. To tease out methodological guidelines of the present kind necessarily involves an appropriation. Firstly, a simplification where we focus on a selection of concepts as opposed to accounting for all of them, the modifications, revisions, extensions and controversies within ANT (Hanseth and Monteiro 1996; Walsham 1997). Secondly, these notions need to be operationalized into methodological guidelines related to the interpretation of information systems (Klein and Myers 1999).

2. Science and technology studies

ANT has not, of course, developed out of nowhere. To trace its roots in any detail would quickly take us beyond the scope of this chapter. Still, to map out at least crudely the background is helpful in getting under the skin of ANT.

ANT is born out of the interdisciplinary field of science and technology studies (STS). This field, which is only a couple of decades old, is grounded in a fairly simple observation. The way science is actually done – not how it is normatively supposed to be – and the way technological artifacts are actually designed – not how textbooks in engineering instruct us – have largely been ignored by the social sciences and the humanities. The essence of STS, then, is to ask the following question: if we bracket all espoused theories about this realm and instead bring along the critical and empirically underpinned apparatus from sociology, history and anthropology, what does the process of producing science and technology look like?

Looking back, a handful of approaches can be identified under the general label of STS (Bijker, Hughes and Pinch 1987; Bijker and Law 1992; Law 1991; Williams and Edge1996):

· The systems thinking developed by Hughes (1983, 1987, 1994) which describes the historical development of technological infrastructure by emphasizing the issues of inertia and heterogeneity.

· The social construction of technology (SCOT) which emphasizes the interpretative flexibility that the relevant actors have in ascribing meaning to a technological artifact. Key contributors are Bijker, Pinch and Woolgar.

· Actor-network theory as developed by Latour, Law, Callon and Akrich as presented further below.

· The sociology of scientific knowledge (SSK) which is devoted to unravelling the infights and maneuvering which go into the establishment of a scientific fact. Key contributors include Collins, MacKenzie and Traweek.

Related to STS, but largely decoupled from it, there is a strand of thinking that has been influential, not the least within information systems research, namely socio-technical systems thinking. This approach has, however, seemed to fade away over the last couple of decades. During these years, socio-technical systems thinking has lost its ground and, ironically enough, even their old abbreviation, "STS". As we approach the end of the second millennium, the label "STS" is to an increasing degree associated with science and technology studies. From the point of view of ANT, Law (1991, p. 8, emphasis added) underscores the main problem with the socio-technical systems thinking, namely that "despite the pioneering work on sociotechnical systems by the Tavistock group in the 1960s, technology does not appear to productively integrated into large parts of the sociological imagination".

3. The basic move

The basic idea underlying ANT is quite simple. The problem is to delimit it in order to side-step or bracket a host of other, seemingly related issues that need not have implications for the present purposes.

A basic principle is connected to the ambition of breaking down disciplinary boundaries:

Guideline 1: Focus on the goal instead of the means

What this guideline says, in relation to systems design, is that traditionally there has been an unfortunate focus on how you accomplish things on behalf of what you want to achieve. Consider a simple example that illustrates how one and the same goal can be achieved by means from two different disciplines.

A few years ago a campaign to make car drivers use safety belts was launched. To change, that is design, the behavior of drivers was the goal. There where two different means or strategies that was attempted. One was through regulation and rules (with accompanying fines), the latter was by the car manufacturers themselves through blinking and beeping arrangements to signal the non-use of safety belts. The former of these belongs to the realm of political institutions and decision making, the latter to technological design in specific car factories (Volvo). What guideline 1 states, then, is that the separation of these two realms – who makes the decision, whether they are technological or not – is in a sense coincidental. When designing behavior, we should be open to any means of accomplishing our goal, regardless of what realm or discipline they belong to – this is subordinate. This of course transforms design into something pretty close to grad, social engineering but that is a matter we deal with in due course.

In terms of ANT, this guideline corresponds to granting both artifacts and actors the same explanatory status. The qualifier "explanatory" is essential. This core aspect of ANT has generated a considerable degree of controversy and misunderstanding that we will not address here (interested readers may consult Collins and Yearley 1992). It is not a statement to the effect that actors and artifacts are equal, that is, have the same kind of intentions, feelings, ethics or interests. It is rather the more modest claim that for our purposes – understanding why car drivers drive without safety belts, or what it would take to make them use belts – regulations or the blinking and beeping are basically two different ways to accomplish the same. They play (or rather: attempt to play; see more on "inscriptions" further below) the same role in shaping behavior. They accordingly have more in common (safety belts) that they differ (rules vs. technological development).

To underscore this similarity, the term "actant" is used to designate either a(n traditional) actor or a technical artifact. The term does not distinguish between the two cases, it is a "monster".

4. The context

Neither technology in general nor information systems in particular has a definite, well-defined or necessary meaning (Orlikowski and Robey 1991). Information systems may be conceived of and viewed in a variety of ways. They may accordingly be used in more than one way. As the conventional wisdom phrases it, ""information technology has both restricting and enabling implications" (Orlikowski and Robey 1991, p. 154).

What this entails is a need to interpret information systems. They need to be made sense of – how are they perceived and used, who views them in which ways, how do the different categories of users use it – in a given organizational setting. Information systems need to be interpreted relative to a context of use, we need to grasp computers in context.

This, of course, a well known argument (Kyng and Mathiassen 1997; Walsham 1993). The question, however, is what this "context" looks like, how should it be conceptualized and how should it be circumscribed?

According to ANT, all behavior including the use of an information system takes place within an actor-network. An actor-network (despite the term!) is simply the collection of the surrounding actants. Hence, there is a very straightforward ANT answer to what the opaque "context" is, namely:

Guideline 2: The context of use is the surrounding actor-network

Consider an example borrowed from the news agency described in chapter XXX KRISTINogTONE NEWS THIS VOLUME. The context of the work with the program schedules is the actor-network which includes the faxes, the shelves for paper, the desks, the information systems for manipulating the structured text and the users.

But the ANT notion of an actor-network contains more. Beyond simple collection, listing or enumeration of the actants, an actor-network underscores how the different actants tie in and mutually support each other.

Guideline 3: Artifacts do not exist in a vacuum but are "connected"; they belong to an encompassing actor-network.

To illustrate this guideline, let us expand on the example used above from the news agency. The different actants – the faxes, the desks etc. – are not isolated by are connected in the sense of mutually confirming each other. For instance, the way paper faxes are inserted, sorted and stacked in the shelves serve to support the work routines connected to the production of the schedule. Faxes and shelf mutually support each other; they are aligned in the same actor-network.

5. Scaling

All interpretation, also of information systems, is about making reasonable judgements. Such interpretations are never neutral but are constructed with a purpose in mind. This implies that in order to make sense of an information system in context the contexts needs to be circumscribed, that is, delimited. In terms of ANT, this takes on two forms.

First, ANT allows you to zoom in and out of a context as suits your purpose. This means that the granularity, the unit of analysis, is not given: it is tuned to your needs. In other words:

Guideline 4: Unpack complexity by zooming in - or collapse complexity by zooming out - as suits your purpose.

What this means is that for one analysis (that is, one description of an actor-network) an organizational unit, say, is treated as one node in the network (an actant) while it is unpacked into a set of different persons, sub-units, information systems, work routines – in short, unfolded into a complete actor-network – in another analysis. The decision about whether or not to unpack it is done by the researcher who describes the actor-network for a given purpose.

This ability to zoom in and out is especially important when analyzing large contexts with a large number of actors, interests, information systems and so forth. To make sense of such a complex context, it is absolutely essential to simplify, that is, collapse complexity by zooming out, by treating comprehensive actor-networks as simple actants. Hence, we talk about the interests of whole organizations, governmental agencies etc. even though it is clear that this is but a short-hand.

A consequence of this is the possibility of linking "large" issues with "small" details. To illustrate, the "large" issues of knowledge competence planning in the Swedish industrial company described in chapter XXXXMagnus og Janne, got linked to and given a material form in the maps and indexing functionality offered by the Intranet solution.

In much the same way as ANT refuses to a priori distinguish between humans and non-humans, it refuses to a priori distinguish between small and big networks. There is, in other words, no a priori distinction between the micro, meso and macro level. ANT offers a uniform framework regardless of the unit of analysis.

What, then, does this imply in an analysis of information infrastructures, what does this zooming in and out of actor-networks look like? It entails that the "actors" of an analysis is of the "size" that the researcher chooses as most convenient relative to the direction of the analysis. Hence, in one case an actor is a given person, sometimes a whole group or community of practice including their working technology, a whole organization or yet another time a profession. Similarly, you would vary between a focus on a technological platform as a whole including the aligned, administrative routines, or some of its applications and patterns of use or down to the details of integration mechanisms, functionality and protocols.

In addition to zooming in and out, it is necessary to delimit the phenomenon in the foreground from those in the background. This is a notorious (and unsolvable) problem in any interpretative analysis (Bijker 1993). But it is especially pressing due to the way actor-networks literally are never-ending. Everything is connected to everything else – if you follow the links far enough. This implies, if you would to read ANT in a square-headed sense, that regardless of what you set out to analyze, you end up with an impossibly large actor-network in your lap.

Hence, you need to delimit background from foreground, main topic from subordinate issues. ANT as such, of course, cannot tell you anything about where to draw this line. That would be naïve to expect. That line is drawn based on a judgement beyond ANT.

Guideline 5: Any analysis needs to delimit the foreground phenomena from the background ones – but ANT cannot help you do this.

6. The dynamics

The process of designing and introducing an information system is an involved socio-technical negotiation. Keeping guideline 1 in the back of our heads, it is essential to recognize the development process as also a social process that needs to be interpreted. Not only the context of use but also the historical trajectory of an information systems project needs to be interpreted.

At the heart of such an interpretation, ANT instructs us, is the insight that there are more than one way to accomplish a goal. Hence, two different interests may be translated and subsequently inscribed into two different materials but still have the same intention. The governmental interest with car driving is to minimize health risks (and costs). This interest is translated and inscribed into regulations and rules. Car manufacturers are interested in selling as many cars as possible and translate and inscribe this into blinking lights.

Tracking such dynamic processes of translations and inscriptions across the development process is the contents of an ANT based longitudinal case study.

 

Guideline 6: The basis for interpreting the process is the tracing over time of the different translations that go into the different inscriptions.

An illustration of such a translation is the example cited in chapter XXXMagnus og Janne where the hierarchical organizational structure got translated and inscribed into the way distribution lists were defined.

During this process, focus on the material of inscriptions, whether driving regulations or blinking lights.

Guideline 7: Focus on the four aspects of inscriptions, namely (i) what is the aim, (ii) who inscribes, (iii) into what material is the inscription made and (iv) how strong is the inscription

An example of a material inscription is the CheckRite application described in chapter XXXX Fredrik og Steinar. The scenario for the new work routine for users was inscribed into functions for extracting predefined templates, retrieving and storing checklists. Another example is the discussion in chapter XXXXKristinogTone on the different strategies for standardization, i.e. soft, guided or enforced. The success of these strategies assumes successively stronger inscriptions to work.

A key issue when interpreting how information systems come into being is to make sense of how certain design solutions more or less mysteriously succeed in mobilizing support while others fail to gather support. The flip side of abandoning systems design as a technical discipline is to abandon the corresponding assumption that the "best" (technically, in terms of efficiency) solution wins out. If this is no longer the case, how is it then that some solutions win out and others loose?

The stability or success of any given solution, ANT holds, is a socio-technical negotiation where the solution with the "heaviest" backing wins out. And backing is the net result of having mobilized support and accumulated inscriptions in favor of it. As Latour (1999, p. 158) phrases it:

"An entity gains in reality if it is associated with many others that are viewed as collaborating with it. It loses in reality if, on the contrary, it has to shed associates or collaborators (human and nonhuman)... [S]tates of affairs become matters of fact, and then matters of course."

The stability of a solution is a measure of how influential the inscriptions backing it are together with the extent to which these inscriptions are aligned, that is, point in the same direction.

Guideline 8: Stability is a measure of the degree of alignment

The example outlined earlier from the news agency may illustrate stability and alignment as well. The way the paper faxes and shelves serve to support the present work practice is may be rephrased as follows. The stability of the present way of organizing work is not given; it is produced. It is not given a priori but emerges ex posteriori. It is the result of the successful alignment of the tangible and visible property of paper faxes and the way shelves lend themselves to checking, sorting and stacking of paper.

A systems development project, after mobilizing support for one of the solutions, cannot rest. The seemingly self-perpetuating process whereby a solution spreads, gains inertia, irreversibility, gathers momentum or picks up speed is neither automatic nor necessary. It is an ongoing effort to keep a decision alive, to conquer opponents, to coopt opposition, to seize opportunities for support and improvise on "surprises" and unintended consequences. The scenarios and plans inscribed in materials never unfold exactly as intended. A project is fragile. The process of alignment is a process also of reconfirmation, of re-appropriation. Aligning the many small, gray and dull details is what keeps the motion going. Without it, everything would fall to pieces.

A fact, an agreement or a design is never "done". Decisions can be undone. A project needs to breathe life into it – to re-confirm it – in order for it not to fall to pieces. In Latour (1996, p. 86) words, this amounts to:

Guideline 9: "For technology, every day is a working day"

This implies more than the fact that there has to be actors actively lobbying or promoting a project. It entails that, especially larger projects, need to find ways to define the project that fits the changing environment. If the definition remains stable it is likely to die.

In opposition to this principle, ANT also contains the notion of irreversibility, a measure for how entrenched the inertia is (Callon 1991).

Guideline 10: Irreversibility is a measure of (i) how difficult it is to undo decisions and (ii) the extent to which these determine subsequent ones

Live projects, especially large, infrastructural ones, are condemned to balance between the contradictory guidelines 9 and 10.

The degree of irreversibility of a network may be regarded as a process of institutionalizing. This operates both ways: an increased degree of irreversibility is signaled by a firmer institutionalization and, the other way around, the construction of institutions functions as a way to align the network and make it increasingly irreversible.

7. Conclusion

The basic message of ANT related to information systems design is a cautionary one. ANT is a strategy for unpacking the complexity of information systems development, introduction and use. Abbreviations, short-circuits and simplifications are always produced. They are the (up till now, successful) result of a mobilization process with black-boxing effects. The ordering these simplifications produce is neither neutral nor "obvious". They are made obvious or "natural" in order to achieve an effect, namely to curb opposition or alternatives.

The way the ANT based guidelines for research methodology outlined above address key issues– spelling out the "context" and tracing the dynamics of a socio-technical process – is a fruitful platform for conducting interpretative research into information systems. The problems and dangers related to the appropriation of ANT, the simplifications, are real but not prohibitive.