To do a sociology of the invisible means to take on the erasing process as the central human behavior of concern, and then to track that comparatively across domains. This is, in the end, a profoundly political process, since so many modern forms of social control rely on the erasure or silencing of various workers, on deleting their work from representations of the work (Star, 1991, p. 291).

1While telecommunication networks keep widening and data flows are more and more important, it is a temptation to make the circulation of transparent information the main characteristic of a society deeply transformed by technical advances. In this picture of a world of fluidized exchanges, vocabulary of the virtual and the immaterial occupies a larger and larger place, up to the point of making “dematerialization” a massive reorientation impulse of productive activities in favor of intellectual work, services innovations and the production of knowledge.

2Nevertheless, this idyllic overview does not withstand analysis for long. The first ethnographical studies on the subject, carried out in highly intellectual work environments, have shown that significant work is needed in order to produce information, and have highlighted the sociotechnical density of such information. Instead of depicting geniuses knee-deep in original ideas, these studies have plunged us into an ocean of traces, inscriptions and documents of all types (Latour & Woolgar, 1979; Knorr-Cetina, 1981; Lynch & Woolgar, 1990). They have shown that intellectual matter indeed spreads, but strictly speaking: that is, by means of various technologies that have been meticulously articulated (Clarke & Fujimura, 1992). Above all, work is done collectively, by scientists, technicians and secretaries.

3Beyond the field of science and knowledge production, these studies highlighted, on the one hand, the role played by “pencil pushers” in the emergence of what would be later called “information society” and, on the other, general scorn for their kind (the stereotype being that in the sciences, the scholar thinks, reality speaks, and pencil pushers do nothing of grand importance...).

Sharing this scorn would be, however, a major mistake for us who wish to follow science in action up to the end. First, because what are seen as defects in the case of the paper-shufflers are considered noble qualities when considering these other paper-shufflers who are called scientists and engineers. […] It would be a mistake, second, because it is through bureaucracy and inside the files that the results of science travel the furthest. […] No, we should not overlook the administrative networks that produce, inside rooms in Wall Street, in the Pentagon, in university departments, fleeting or stable representations of what is the state of the forces, the nature of our society, the military balance, the health of the economy (Latour, 1987, pp. 255-257).

4Science, like law and administration (Yates, 1989), relies on armies of information workers who manipulate paper scraps, folders, and files essential for the world’s everyday performance. Now, computers and networks seem once again to remove these workers, their tools and materials from the map by tricks of transparence that have become commonplace. Today, information is accessed by keywords typed into search engines, even sometimes just by double-clicking. Information appears as a natural and objective entity available everywhere and effortlessly.

5Such immediacy is based on what Star and Strauss (1999) call the ecology of visible and invisible work. If knowledge and data reach visibility almost instantaneously, it is at the expense of concealing of the operations that make possible their creation, publication and circulation. Information society relies on sociotechnical infrastructures that effectively support it under the condition of staying invisible. Academic studies dealing with these information infrastructures are growing in number, becoming a true field of research in which systematic attention to processes of erasing certain tasks and making certain workers invisible are a key aspect. This research has focused on cooperative information systems (Star & Ruhleder, 1996; Bowker et al., 2010), classification systems (Bowker & Star, 1999) and standards (Lampland & Star, 2009). All show that the value of work and its recognition as an activity that counts are the result of the ecological distribution of what is present and available on one hand, and what is erased and invisible on the other. Assuming a political and moral position, these studies aim to restore invisible workers to visibility by means of an “infrastructural inversion” (Bowker, 1994) which documents their work conditions, the heterogeneity of their tasks, and their knowledge and skills.

6This special issue has several purposes. First, it aims to testify to the dynamism of such infrastructure studies, including in Francophone research, and to recall the importance of its contribution to an anthropology of knowledge. It also suggests widening the horizon of the activities studied until now, by linking the analysis of scientific databases and medical administration tools to various situations: emergency medical call centers, questionnaires and surveying, computer coding, even animation filming. This widening puts the analytical framework of infrastructure studies face to face with unusual technologies (though commonplace in the context of these activities) and with the forms of invisibility they entail.

7The issue invites us to confront the generalization of work-effacement, the different types of effacement in operation, and the multiplication of invisible workers who produce and maintain our information society and its services up to date. In different ways, the articles presented here reveal the very density of the information infrastructures where the data, documents and files that inhabit our world are daily produced. In other words, they try to “make the invisibility of administrative work visible, to comprehend which are the activities that our contemporary economies carry out and on which technical and cognitive infrastructures are based our so-called information societies” (Gardey, 2001, p. 13).

From hand to hand: this issue’s contributions

8The first two articles plunge us into two very different fields. Analyzing open source code, Stéphane Couture identifies a form of distributed production based on participatory logic. Numerous contributors from all over the world are able to propose additions or modifications of certain software features. The article shows that this collective process goes through specific writing acts such as “commits”, which are studied here in detail and which have data associated to each (contributor name, revision number, date, time, description of modified files). Like legal documents, these acts guarantee both the traceability and the performativity of revisions operated on the code source. A strict hierarchy of contributors with various levels of action and visibility works to approve these changes. Emmanuel Grimaud focuses on the film animation industry in a globalized context where certain digital operations are outsourced to India. The distribution of production takes the specific shape of subcontracting arrangements. The article shows that the appearance of computers in the cartoon industry meant not only the automatization of boring and repetitive tasks and new relations between directors and cartoonists, but that it also redefined interactions between workers and their production tools by means of a digital hand whose mechanics are forever compensated by unforeseen talents.

9The following three articles deal with a now-classic issue in information infrastructures studies: technological formalism and its aporias (Star, 1995; Berg, 1997). As we read them, we bear witness to the diversity of formalist projects and discover the modes of change and resistance they engender. Exploring the backstage world of questionnaires, Rémy Caveng reveals the tensions brought on by a strict respect of formalist demands made on behalf of scientific objectivity, which guarantee quality of information, and the radical strategies used by pollsters who go so far as to sometimes invent their data. Caveng shows how this resistance links a criticism of working conditions to the scientific value of survey standardization. Mélanie Hénault-Tessier and Sophie Dalle-Nazébi examine request processing at emergency medical call centers, focusing on the relation between employees and the forms they must fill out. Here, formalism is dedicated to coordinating actors (doctors, ambulance drivers...) and to tracing activities and decisions. Within a perspective close to the traditional sociology of work, the authors show that the slight adjustments agents make do not conflict with formalism but indeed facilitate it. Exploring the case of the implementation of electronic medical records in a hospital, Anne Mayère, Isabelle Bazet and Angélique Roux analyze a prescribed multifaceted formalism aiming to remove the use of paper. Such a framing of nursing work reveals how an organizational and regulating project linked specifically to the implementation of electronic medical records is based on an essentialist definition of both information and coordination. The nurses struggle against this definition in their daily work.

10The last three articles in this issue deal with databases, essential devices in contemporary information infrastructures. The role of databases is growing increasingly important in scientific and medical work (Bowker, 2000; Hine, 2006). Each contribution focuses on a specific moment in the life of a scientific database: creation (Heaton & Proulx), data management (Millerand), legitimization (Dagiral & Peerbaye). In the first two articles, databases aim to support collective and interdisciplinary research, distributed to several locations. Lorna Heaton and Serge Proulx study the scientific and political conditions of the production of a worldwide data library for the botany field. They show how the digitalization of specimens in herbariums, spread out and organized differently from one location to another, is a key operation in the development of such infrastructure. The value of these historical collections may be weakened not only by separating data production from reference location, but also by ignoring or disregarding gestures when handling crucial specimens. Florence Millerand studies metadata that accompany, document and make possible the circulation of data within an environmental sciences network involving geology, climatology, zoology, and vegetable biology. The data is distributed to twenty-seven sites. As it both coordinates and supports coordination, the production of this metadata represents a real challenge for interdisciplinary research. The author emphasizes the ambiguity of the role of information managers in the process: essential, as they must face various collecting practices, technical architectures and methods of classifying and filing. In spite of their importance, they remain confined to an intermediary position. Éric Dagiral and Ashveen Peerbaye focus on quite a different database. Theirs is dedicated not only to collective scientific work on rare diseases, but also to data availability for nurses, associations, doctors, researchers, industrialists, and public health organizations. The authors show that visibility is not only a matter of research but a concrete problem with direct and unintended effects on the institution. They highlight the variety of activities used to legitimate the institution’s existence, and they reveal the additional work needed to render visible database supply operations. The authors shed light on the close connections between work visibilization, data valorization and the legitimization process

11The articles gathered here show the wide range of fields in which invisible information workers play an important role. As such, they broaden the scope of infrastructure studies, which have until now concentrated on scientific activities, and they bring into focus new angles of study. This special issue also pleads for the movement to continue, and for new research to push these concerns even further, to yet wider fields: journalism, commercial services and activities, law, web infrastructures (from social network sites to search engines, including e-sales and video games).

12Among the discussions in this issue, the question of the connections between visibility and invisibility are undoubtedly the deepest and most central. Of course, the question is highly pertinent with regards to interface production, but the various types of invisibility also reveal the different ways in which workers of information may be considered unimportant.

Invisibilities

13From labeling theory (Becker, 1963) to articulation work (Strauss, 1985), including “tie-signs” and “discrepant roles” (Goffman, 1959), the issue of relations between the visible and the invisible is present throughout the classic works of interactionist and even general sociology (Brighenti, 2010). New momentum arrived with research on scientific work (Clarke & Fujimura, 1992), information infrastructures (Bowker, 1994; Bowker & Star, 1999) and representations of work in categories and formalisms (Star, 1995; Suchman, 1995). This renewed focus on the ecology of the visible and the invisible revives the central sociological issue of the division of labor. But this traditional problem is somehow displaced. Here the division is seen as a process in itself, involving dedicated operations (Strauss, 1985). This perspective therefore allows to document a great variety of aspects: task hierarchy, distribution of work, and the link between the distribution of rights that may be claimed, negotiated or even imposed and the division of labor, including ways of accounting for completed work in the final product. Because of the extreme division of work that characterizes our societies, some activities remain inevitably invisible to others (Nardi & Engeström, 1999).

14For information workers, three main kinds of invisibility may be identified. They create a specific link between the type of task in question, the relative invisibility of the person or persons completing it, and the position of said persons.

15The first kind of invisibility is the direct product of the lower status granted to certain people in work organization. Maids are the archetype of such a position (Rollins, 1985), the visible signs of their work being systematically linked to a lowering of their condition as a person in interactions with employers. Taking up with this matter in science, Shapin (1989) in a well-known article spoke of the status of lab technicians (“servants” as they were called at those times). He pointed out three particular aspects that led to their invisibility in scientific reports of the 17^th century: their work was defined as a manual activity that required physical energy and skills, but not reflexive abilities. They worked under the authority of a master who supervised the site, ordered employees and saw the execution of necessary tasks as an extension of his personal wishes. Such tasks were performed under a financial dependence that contrasted with the freedom of thought and action that otherwise prevailed at the time.

16This form of invisibility is still present in some fields. For instance, in medical work, Acker (1997) documented the demands and tools used by nurses to make their work visible and gain recognition for their status. However, this recognition may be partially questioned in light of new equipment on the job. Digitalizing medical records raises the risk of a potential redefinition of worker status. Nurses perceive this threat and resist, opposing the computerization of a part of their activities and a too limited definition of their role (Mayère, Bazet & Roux, current issue). In scientific work, the status of technicians, often reduced to the translation of physical or natural phenomena into signs, symbols or inscriptions (Barley & Bechky, 1994), also risks transformation in light of new tools for production and data processing (Bowker, 1994; Hine, 2006; Millerand, current issue).

17The second kind of invisibility arises from the position of workers within a sociotechnical network. It is understood from a semiotic point of view and on the basis of actor-network theory, which establishes that human and non-human entities take on a more or less stabilized meaning and shape from their relationship with others. For instance, there are certain tools in a laboratory whose only function is transforming matter from one state to another (e.g. a centrifuge or a mill). Yet, such a tool (even something as simple as a scale) can play a very different role depending on its use, whether for example it is used to weigh a known substance or to obtain information about a new compound. In the first case the tool is handled as a “machine”, a daily partner in the activity; in the second case, it is considered as an “inscription device” used to translate matter into signs or figures and thus producing significant traces (Latour & Woolgar, 1979, chapter 2). This distinction sheds light on the differential role that each entity plays in the production of reliable information that may be useful in scientific literature.

18Nevertheless, most entities involved in scientific production (appliances and machines, technicians, materials, secretaries, specialized literature, databases, cleaning staff...) are absent from reports. Within the actor-network framework, it is by means of this process of deletion, of placing in black boxes, that utterances achieve their value as truth and as objectivity. Invisibility is in the relative position occupied by entities in a sociotechnical agencement where threshold effects between the laboratory and the world of publications and knowledge enhance certain contributions to the detriment of others. These neglected contributions are confined to the backstage spaces of the public result. This process is particularly delicate, even problematic, as it may concern the activities of an entire organization, such as a website that produces and provides information in five languages about thousands of rare diseases (Dagiral & Peerbaye, current issue). As a result, it is sometimes the case that incessant visibilization work is needed in order to restore legitimacy to a plethora of activities accomplished daily in the backrooms of the publication.

19The third form of invisibility is less related to workers themselves and their tools as it is to the nature of the tasks they carry out, which are considered barely honorable or respectful. As Hughes (1962) has showed, professions are not just organized by the technical distribution of tasks, but are also affected by a moral division of labor. Every profession, even the most prestigious, requires some “dirty work”. Tasks seen as less noble, demeaning and sometimes even disgusting are considered to be lower, and people in charge of them are not part of the “real” professional circle (ex. taking care of the laundry or cleaning patients in medical work, applying laws in the legal field). Thus the moral definition of dirty work allows the exercise of a monopoly on activities that are valued in workplaces, and this monopoly defines “what counts” as a job in a particular professional world (Strauss, 1985; Star & Strauss, 1999).

20Of course, occupations dedicated to the production of information are not exempt from a moral division of labor. Disregard for punctilious bureaucrats, pencil pushers and paperwork haunts administrative work. With access to computers and new forms of management, this moral division also has another side: tools for description and assessment objectify and rigidify this work, fixing it into temporary categories that divide important work from the non-important. Such processes of abstraction and reification coexist with exclusion phenomena; every set of categories, as rich as it may be, generates its own fringe workers and “monsters” who do not fit inside any box (Bowker & Star, 1999). These dilemmas are part of the job of pollsters in charge of delivering questionnaires to survey institutes (Roth, 1966; Caveng, current issue). The latter are constantly faced with the standardization of categories and procedures, imagined as guarantees of data quality. They must adapt to constraints, and even sometimes must openly “fudge” their operations in order to meet their survey quota.

21These three forms of invisibility produce “non-persons” (Goffman, 1959) in their own way; that is, beings considered as absent from interactions despite their presence and effective participation. Each of them emphasizes a different aspect of invisibility. It is important to differentiate them so as to understand their modalities better, all the more so as these modalities are often concomitant. For example, the appearance of a new technology such as the typewriter radically transformed typists’ work: aside from re-organizing the sociotechnical network, the mechanization of writing brought upon a devaluation of tasks and the declassification of operator status, leading to the mass feminization of the activity (Gardey, 2001). The distinction between different forms of invisibility invites us to examine the conditions under which they emerge, how they are perpetuated, and also their possible combinations. It is a way of “surfacing” the ecology of work (Star, 1999), and of exploring the always relational and recursive production of the visible and the invisible. The same task sequence can indeed be exhibited for a while in certain places and be hidden at a different moment or in other places. Far from being intrinsic properties, visibility and invisibility are constantly generated one with respect to the other.

22Nevertheless, one must be careful on this point. Visibility is not necessarily desirable by everyone at all times. Surfacing work that escapes formal representations (Star, 1995; Suchman, 1995) contributes to its possible recognition, but it also offers support for its formalization and the production of criteria for its assessment and supervision (Timmermans, Bowker & Star, 1995). Making something visible, even for the sake of a social justice principle, is also providing means for its control.

24This is an important issue also because it does not only concern the situations studied here, whose complex relations between the visible and the invisible may be described from the outside. The issue is also epistemological. Though traces, measurements and new accounts are generated by actors in various fields, they also are the products of the researchers themselves (Baszanger & Dodier, 2004). As we stress one aspect or another of work, our own reports generate visibilities and invisibilities, whose consequences should be, if not fully anticipated, at least taken into account.

The problem, then, is both analytical and political. It is to try to work out – to make decisions about – how to interfere, or, at the very least, to be conscious of the fact that descriptions are performances, and that no description is ever entirely innocent (Law & Singleton, 2000, pp. 769-770).

25This is one of the strengths of the model: the ecology of the visible and the invisible is not a temporary specific phenomenon. It is an invariant of work itself and its description. From the researchers’ perspective, it requires a certain “methodological humility” (Law & Singleton, 2005, p. 350) – no complete and neutral description is possible – as well as epistemological and political reflexivity.

26Although connections between the visible and the invisible are essential in infrastructure studies, this special issue is not limited to this aspect. It also explores possible specificities of activities that are both omnipresent and invisible in the descriptions of what would be an “information society”. As such, it can be hypothesized that the abstract and sometimes essentialist definition of information on which these descriptions are based, and on which the organization of society itself relies tends to favor an increasing erasure of work, for the sake of more fluid services made available to an increasing number of users. This hypothesis of a vast “back-officization” of the world provides much pertinence to Goffman’s proposal (1959, p. 158), which encouraged a closer examination of the “specialists in verbal fronts” organizations ask “to collect and formulate the array of facts [with the purpose of] writing the ceremony for” their own public performance.

27To follow Goffman’s urging, and to pursue the discussion taken up more or less explicitly by the articles in this issue, we will now insist on two aspects which, because they allow us to re-specify the uses of the vague notion of information, point out useful avenues that may widen the current methods of analyzing information infrastructures. The first aspect refers to the scriptural nature of most information handled and produced backstage. The second aspect emphasizes how studies on invisible workers may be useful in reexamining the issue of information materiality.

The work of writing

28Invisible workers of information, whether in the backstage spaces of administrations, scientific productions or services, all live in a world full of writings. Information infrastructures are mostly scriptural infrastructures, as they consist of documents, forms, notebooks, handwriting notes, letters, etc. Since the 1920s, real “factories of information” (Gardey, 2001) have been introduced in administrations and large companies (Beniger, 1986; Yates, 1989) thanks to which some information moves around, some is compiled, some is validated, processed, categories are stabilized, formats are consolidated, documents are issued, charts are completed, etc.

29Several studies have already highlighted the importance of the scriptural part of work in different fields. They insist in particular on the role of varied formats and materials of writing in coordination (Berg, 1996; Heath & Luff, 1996; Vacher, 1998; Nomura, Hutchins & Holder, 2006). They have also broken with an exclusively formalist view of writing, which reduces its scope to pure prescription (Boutet, 1993; Fraenkel, 2001). Not only are reading and writing practices articulated with forms of oral communication (Grosjean & Lacoste, 1998), but they also get tangled with tensions between official (sometimes statutory) prescriptions and “real work” adjustments (Mayère, Bazet & Roux, current issue). Because this research perspective is linked to the concrete handling of writing and the variety of its instantiations, it is a way of beginning to question the essentialist definitions of information, born of cybernetics (Hayles, 1999).

30Interest in backstage work extends this movement, focusing on a different side of scriptural work: production and processing. It implies a reversal of perspective, however. From the analysis of writing at work we must switch to an analysis of the work of writing (Pontille, 2006; Denis, 2011) in order to shed the light on the concrete operations and skills upon which the daily maintenance and functioning of information infrastructures rely. Largely neglected in official documents dealing with information work, they are also overlooking in the social sciences. They deserve particular treatment.

31We shall insist here on two main dimensions: the importance of operators’ physical commitment and the role of judgments in the production and processing of information. Each of them illustrates the pragmatical depth of the work of writing.

32Document checking, database entry and form filling are generally degraded to the level of paperwork, and those who carry out these tasks are seen as mere office assistants. This conception comes from the notion that no special skills are required; that the tasks can be done almost automatically and so do not deserve to be called “real” work. Dirty work par excellence, the work of writing involves nevertheless much more than second nature reflexes.

33Certain studies on reading practices have, notably, demonstrated this. Reading can be based on very different forms of commitment on the part of the “reader”. For instance, in the legal field, while exhaustive reading is sometimes essential to ensure statements’ felicity conditions, rereading documents meant to be signed may be “skimmed”, wherein selective scanning focuses on certain zones of text. This practice is based on a widely distributed chain of writing and reading practices (Fraenkel et al., 2010, chapters 4 and 9). In the backrooms of commercial relations, checking also involves ways of reading that strongly contrast with a narrow view of “information processing”. Client record verification goes through several steps, some of which consist in reading the same document - sometimes skimmed, sometimes annotated, and sometimes carefully deciphered (Denis, 2011). These consecutive readings are also present in biomedicine during database updating. A particularly focused form of reading, dense and quasi-expert, is required in order to allow an operator to identify the writer of certain documents and thus assess its pertinence or refine it (Pontille, 2010).

34Beyond rich and varied reading formats, information work goes through multiple treatments. A series of documents may be re-ordered, or information sources may be gathered, their processing methods enriched via positions and movement in the workplace (Kirsh, 1995). Some of the treatment necessary for producing and maintaining information infrastructures may turn out to be particularly delicate. For instance, the creation of an international botany database is based on skills essential to avoid damaging the herbarium’s specimens during their digitalization (Heaton & Prouxl, current issue). Delicacy and even virtuosity can also be found in the work of animation studio technicians who, even when guided by an electronic hand, develop personalized information tools and create calibrated support mechanisms within the machine as to maximize fluidity and “naturalness” in animation images (Grimaud, current issue).

35Insisting on the variety of operations found in the backrooms of information infrastructures allows us to enrich existent thought on the theoretical and political shortfalls that a transparent and dematerialized definition of writing conveys (Grosjean & Lacoste, 1998). We are able to focus not just on crucial moments of elaboration (databases, categories, information systems, etc.) but also on situations of daily maintenance. This approach also incites us not to limit the invisibility issue just to workers. Studying these operations, it is necessary to ask systematically which receive recognition or of some kind of visibility, and which are “erased” from official work descriptions and moral hierarchies established as time goes by. Among several tasks that make up the writing of computer code, the command “commit”, for example, has become particularly visible. It has turned into an act of writing itself, which is a strong indicator of its contribution to a collective and continuous production process. At the same time, though, it masks the role of the numerous contributors involved in these hidden operations (Couture, current issue).

36The lack of recognition of information workers’ labor is also characterized by the refusal to see it as intellectual work. This is especially true in the backrooms of customer services, where the work done is considered by managers to be more or less “unthinking” (Boutet, 2001). Though it may require specific actions, for them this type of work is not based on cognitive skills. Thus it may be strictly confined to the realm of technical labor. This position is closely related to the definition of information as a vehicle of an unequivocal meaning capable of flowing seamlessly from one part of a system to another. Such a point of view completes, in a way, the model of dematerialization, intensifying the image of an increasingly fluid information society. In the humanities and social sciences it is easy to reject such a caricatured view. But rejections are frequently made on principle alone, and they rarely include a precise description of the intellectual commitments involved in the production and circulation of information.

37This question has been dealt with best in ergonomics and the sociology of work. In particular, several research projects have focused on the interactional dimension in service relations, highlighting the specific skill sets required from employees (Joseph & Jeannot, 1995; Cerf & Falzon, 2005). Others have shown that the success of relations largely depends on agents’ ability to make in situ adjustments to the requests received, articulating them with respect to the organization they represent (Eymard-Duvernay & Marchal, 1994; Weller, 1999).

38These tasks, which involve both translation and (sometimes sensitive) negotiation with users are based on complex cognitive operations, and always require the agent’s judgment, particularly when information production depends on the very specific act of qualification (of an object, a situation, or a person). Qualification involves subtle estimation and delicate adjustments, made all the more difficult in emergency situations (Hénaut-Tessier & Dalle-Nazébi, current issue). Information production, coding, and data processing are situated, like data elaboration, in advance of classification and metadata (Millerand, current issue), on a series of decisions – more or less collective, and with varying levels of importance – which ensure the quality of information infrastructures on a daily basis.

The materials of information

39Interest in the production and maintenance of information infrastructures is not limited to the analysis of invisible workers and the mechanisms that favor their invisibility. The studies mentioned here, and particularly the papers in this issue, all question essentialist and disembodied definitions of information. Some of them even struggle against them explicitly. In so doing, they oppose the view according to which immateriality and transparency are the foundations of information society, inviting instead a deeper analysis of the material properties of informational “objects” as they are produced, manipulated and transformed.

40This is a vast and complex matter, and we cannot address it fully here. Nevertheless, its previous explorations – under the pretext of interest in information “materiality” – have been so shaky, even dubious, that we would like to examine it more closely here. As Ingold (2007) pointed out in an article addressed to material culture studies scholars, the very notion of materiality is deeply questionable. Though it is generally mobilized in order to lay claim to the world’s physical and social dimensions, it is always based on an anthropological division between materiality on the one hand, and ideas, information, beliefs, etc., on the other.

41This reasoning can be found in the French usage of the notion of “support” (a medium, format, or material used to convey information). Many are the scholars who have shed light on the role of “supports” in writing. They have explored the active role of artifacts such as papers, whiteboards or software. Although these studies have the advantage of rejecting models of immateriality, they are nevertheless problematic in that they render the “support” autonomous, a kind of table upon which information would be “placed” (Vinck, 1997). They seek to describe the role of these mediums, namely in isolating their representational properties. As such, the “support” is given agency not only with regards to the world, but also to the text. While insisting on the close relationships between the two, but without even explicitly accepting the consequences, this approach assumes a fairly sharp distinction between “support” and text, between medium and message.

42As soon as we decide to focus on the instantiation rather than the representation of information – and thus to focus on production – this distinction no longer stands. The use of two tools (for example, an accounting book and a finance database) to record figures does not mobilize two separate representations of the “same” information, which would be only modified by its “support” or means of conveyance. Rather, it participates in the production of two different pieces of information that have neither the same form, the same value, nor the same scope. Data processing always consists in creating informational objects (Pontille, 2010; Denis, 2011).

43This is a vast and complex matter, and we cannot address it fully here. Nevertheless, its previous explorations – under the pretext of interest in information “materiality” – have been so shaky, even dubious, that we would like to examine it more closely here. As Ingold (2007) pointed out in an article addressed to material culture studies scholars, the very notion of materiality is deeply questionable. Though it is generally mobilized in order to lay claim to the world’s physical and social dimensions, it is always based on an anthropological division between materiality on the one hand, and ideas, information, beliefs, etc., on the other.

44This reasoning can be found in the French usage of the notion of “support” (a medium, format, or material used to convey information). Many are the scholars who have shed light on the role of “supports” in writing. They have explored the active role of artifacts such as papers, whiteboards or software. Although these studies have the advantage of rejecting models of immateriality, they are nevertheless problematic in that they render the “support” autonomous, a kind of table upon which information would be “placed” (Vinck, 1997). They seek to describe the role of these mediums, namely in isolating their representational properties. As such, the “support” is given agency not only with regards to the world, but also to the text. While insisting on the close relationships between the two, but without even explicitly accepting the consequences, this approach assumes a fairly sharp distinction between “support” and text, between medium and message.

45As soon as we decide to focus on the instantiation rather than the representation of information – and thus to focus on production – this distinction no longer stands. The use of two tools (for example, an accounting book and a finance database) to record figures does not mobilize two separate representations of the “same” information, which would be only modified by its “support” or means of conveyance. Rather, it participates in the production of two different pieces of information that have neither the same form, the same value, nor the same scope. Data processing always consists in creating informational objects (Pontille, 2010; Denis, 2011).

46Within information infrastructures, moving from one form to another, from one informational object to another, is common. These movements are at the heart of the idea of “processing”, essential in back-office work. Examples can be found in most of the papers in this issue, from the transcription of oral responses to a survey (Caveng), the digitalization of plant specimens (Heaton & Proulx), the creation of metadata used for database compatibility (Millerand), computerization of medical record (Mayère, Bazet & Roux), or even the publication of a printed yearbook displaying a large database of rare diseases (Dagiral & Peerbaye).

47Emergency calls (Hénault-Tessier & Dalle-Nazébi) are, in many aspects, an extreme case: they involve complex operations that allow a phone call to turn into (or not) an in-person visit, thanks to the production of several documents. According to the essentialist model, passages from one to another (including from voice to paper) are considered transparent, and each object is a more or less precise vehicle for information. In the materiality model, on the contrary, “supports” are involved (papers, post-its, screens, records), and the representational properties of each modify the text. From the perspective we are interested in, each object is a more or less stabilized result of production. Rather than transmission or representation, these objects carry out translations, in the sense of actor-network theory (Callon, 1986). In these situations, there is no such thing as information that would move from one “support” to another, changing in the process. Instead, several pieces of information, instantiated in different artifacts, are created progressively.

48The fuzzy notion of “materiality” is another problem. Influenced by a biased reading of Goody, Olson and Eisenstein’s works, many of the scholars who make use of this notion confuse it with tangibility. The materiality of writing consists in their view in insisting on the stability, solidity and durability of written artifacts. This point of view tends to forget that words written in the sand, those made using smoke in the sky, and those spoken aloud are “material” too. It risks neglecting the fragility and vulnerability of information artifacts, and therefore risks forgetting about the work that is necessary to maintain them (Denis & Pontille, 2010a). Such work is nevertheless essential to the existence of information infrastructures that, without it, may become flawed and may even break down.

49Maintenance activities are, on the other hand, useful in understanding the pragmatic depth of the work of writing, and in exploring the diversity of materials that make up information. Studying them allows us to accept Ingold’s invitation (2007) to abandon the vague vocabulary of materiality and instead to question the variety of materials, before or after they become artifacts with hermetic boundaries.

We see the building and not the plaster of its walls, the words and not the ink with which they were written. In reality, of course, the materials are still there and continue to mingle and react as they have always done, forever threatening the things they comprise with dissolution or even ‘dematerialization’ (Ingold, 2007, p. 9)

51Examining the production and maintenance of scriptural objects rather than their consumption and uses allows for a direct questioning of the material properties of information, revealing the multiplicity of materials they consist of, and the diversity of operations required for their assemblage, consistency, and maintenance. In contrast to what Ingold (2007) perhaps suggests in his critique of his anthropologist colleagues, though, it appears to us that this point of view should not prevent us from seriously considering situations when informational objects are considered as stable and solid entities. The attention that infrastructure studies pay to uncertain moments in the processes of reparation, shaping and assembly does not further a univocal definition of information. On the contrary, it is worth examining the dual quality of infrastructures. Files, documents, forms, databases, bits of computer code, pages of medical records: these all exist in two closely linked versions. They are at once stabilized and detached artifacts that present information upon which some people can rely, and they are fragile objects with multiple components and imprecise boundaries that need constant maintenance, configuring and reassembling.

52Instead of pretending that the “real” nature of information lies in the uncertain forms we can find backstage, the challenge is to understand how, and whose behalf, the latter get erased in the stabilized and solidified version. It is by documenting the invisible work of information, its materials and the forms of its invisibility, that we will be able to pursue a debate that truly challenges the model of disembodied information.

We are grateful to Jill McCoy for her translating assistance.