Christopher Kelty

1. Reproduce vs. Emerge

The ARC project interests me for a specific reason. I see one of my tasks here as an attempt to move discussion beyond oppositions such as digital/analog or bits/atoms or new media/old media and towards an approach that I find resonates with the problematic of reproduction/problematic of emergence namely towards a notion of "collective experimental technical practices and their modulation." 15 years of being a software programmer manqué has left me more convinced than ever that oppositions like digital/analog are completely meaningless. However, when confronting the kinds of changes we associate with new media and new digital tools it's also obvious things are not "just the same, only faster." There are qualitative and quantitative changes at work. As I interpret it, a problematic of reproduction would be wedded to determining the difference between new and old media, digital and analog, and a committed to a rhetoric of radical transformation balanced by one of lasting sameness. In this mode, the ultimately interesting question is always "but is it new?" and the answer is always something like: no, it's just the relentless power of capitalism/empire/ideology/technocratic dominance etc. By contrast, A problematic of emergence would open up the possibility for thinking about how organized and settled assemblages of human practices, technical objects, legal regimes and ideologies are modulated over time until, through quasi-controlled collective experiment, something unfamiliar emerges" something surprising. Novelty in this case is not radical transformation, but constant, though heterogeneous and stratified, change. In my own research the paradigm case is Free Software and my goal has been to explore it not in order to show how it is just one more example of the relentless dissolving power of technology (or capitalism) but how it emerges in response to a problematic of knowledge and governance that is itself emergent but at a considerably slower pace.

In terms of ARC and the idea of a novel form of collaboration, the question can be directed at the production of scholarly work, rather than the domain of software" that is to say, to explore how the transformations of technical, legal and organizational practices of knowledge production can be related to the theoretical questions of concept work. In this respect, I am very much in favor of thinking through experiment" namely thinking through the problems of setting up a clear experiment, keeping track of its project, verifying its outcome, using the result as the basis for the next experiment, and above all making such a system into a "generator of surprises." In my participation with ARC (i.e. especially in the domain of setting up, configuring, imagining and promoting certain kinds of technical-social solutions, from websites to wikis to blogs to content management systems), it is the notion of "keeping track of" that is central. These tools change the nature of scholarly work" not radically, but "emergently"" they present us with ways of doing what we have always done differently, and it is in making such modulations that we are in the midst of experimenting" and thus need to take stock of what we've changed, why, and what the outcome is. Such tools are a replacement infrastructure for a professional system that used to (and in many disciplines still does) keep track of the development of concepts and results. They are not merely "new media, not merely digital replacements for analog practices" they modulate rather than replace. The practices of arguing, writing, circulating, re-writing" and indeed, the practice of "concept work" itself" is modulated in the doing. Thus, I advocate a more careful attention to the "collective technical experiment" as the concrete environment that makes "concept work" now different than "concept work" then.

2. Discipline vs. Profession

One distinction that has been offered in ARC is that between a profession and a discipline. In the AAA Panel associated with this exchange, Collier answered a question from the audience by suggesting that anthropology is a "discipline without discipline" by which I suspect he meant that it was a professional organization" encompassing a rather radical diversity of possible topics and approaches" with no agreement about central problems, concepts or methods. It may not be the case that any discipline actually possesses firmly agreed on problems, concepts and methods, but there is certainly a spectrum, so that computer science, for instance, in its various subfields has very rich disciplinary (and professional) structures, as does economics. The downside of course is that the stronger the discipline is, the less likely heterodoxy and critique are to survive within the Profession.

One of the ways in which both the Discipline and the Profession organize and reproduce themselves is through the practices of publication that they adopt. Entrants learn how to recognize immediately (in an article, for instance) what the common problem is, whether a new concept is being proposed, or a proposed one tested; and learn, through time, not only how to judge and discriminate, but to make use of, re-use, test, build on work, in ways that are more complex than simply citing another work, or using another work as evidence. Concepts enter an "archive" that is not necessarily co-terminous with an archive of literature identified with a Profession. It's an "archive" in a notional sense only" it belongs to the practitioners who navigate the field and understand the structure of the discipline. The texts are the traces of the disciplinary structure. To borrow an overwrought image from Roland Barthes, concepts are the imaginary tail of the texts" the meteroid is the text, while the comet, that flash in the sky, is the concept. To the extent that an audience of scholars is looking, there are concepts which they all see in a particular body of texts, but which disappear once that audience disappears.

Thus, the claim at least by some observers within the Profession, that anthropology has no concepts. For example, the virtuoso text" the extravagantly written, the philosophically polymathic or simply the empirically hyper-detailed text" does not participate in a discipline so much as construct its own" a comet reflecting on its own entrance into a vacant atmosphere, or one peopled by a few friends and insider colleagues. By this caricature, some of the most widely respected writing in the Profession would also be some of the most unwilling to collaborate as a discipline of concepts. A bit further down the scale are sub-fields like science studies that seem to have a pretty widespread core of problems and concepts, loosely shared across many professions. Even further down the scale, one might put Meso-American archaeologists, who (if my recent tenure on a job search committee is any indication) seem to have an extremely robust set of disciplinary problems and concepts to which they refer" at the expense of being able to translate that to any other field, even seemingly close ones like Africanist archeology.

If it is the case that ARC sees anthropology in this way "a discipline without discipline" then along with the obvious requirement of convincing other like-minded scholars to join, it faces the "technical" problem of publication and circulation directly. If it cannot rely on a professional system that promotes the development of disciplinary (conceptual) collaboration, then how, practically speaking, can it develop this practice of proposal, counter-proposal and building? On the one hand, it strikes me as absurd to suggest that ARC can do any such thing" disciplines are evolved, not made, even if charismatic personalities sometimes inaugurate them. On the other hand, we have the technology today to constitute a space that is focused on evolving discipline, but which is very different from existing Professions. It is truly, a different kind of public and publication to be able to instantly circulate everything" it raises very hard questions about how to sort, sift, validate, refine, ensure quality, find quality, maintain discussion, etc" to say nothing of simply dealing with the information overload.

So, I return to the question of "collective experimental technical practices and their modulation" and propose that one way ARC should be approaching its activities is by revealing how the collective experimental technical practices of Profession-making and Discipline making differed in the past and how they differ now. I think this question is behind the urge to conjure forth a genealogy of collaboration, such as Rebecca Lemov's contribution to the panel, and the repeated reference to issues such as complicity and adjacency or to past failures like the New Nations project. Is it possible to explain past practices within anthropology through the lens of Profession-making/Discipline-making? Is it further possible to ask how exactly contemporary practices and technologies are modulating these past practices, transforming them into something that serves the goal of providing a space for concept work perceived to be absent from anthropology as it exists today?

3. Collaborate vs. Coordinate

Collaboration is hard to distinguish. The first distinction within an anthropological setting is usually collaboration with "informants" (complicity, adjacency, co-optation, earnest service, collegiality, critique and denunciation) versus collaboration with colleagues (co-writing, co-granting, pedagogy, circulation of written work, hierarchy and division of labor). The distinction breaks down in almost every case, for different reasons: we "co-write" with informants all the time, we develop complicity with our colleagues, or denounce and critique them. Especially those working amongst elites and high-tech situations, the idea of a distinction between informants and colleagues is usually a question of whether they are formally in the department or not (and sometimes not even that, pace Donna Haraway amongst primatologists, or ethnographers amongst the archaeologists). A recent American Ethnologist volume devoted to IRB politics raises these issues in a particularly stark way: Rena Lederman's attempt to study IRB's turns bathetic when the IRB itself tries to evaluate her research on human subjects, who are members of the IRB. More broadly, linguists and earnest anthropologists of the down-trodden and oppressed insist for political reasons on the use of "collaborator" instead of "informant." The breakdowns are endless, the issues of responsibility and accountability a thicket of conceptually bleak auto-ethnography.

By the same token, when collaboration works, it seems to work absolutely. The best ethnographies seem to have at their heart a really delicious interplay between the knowledge of a "native" informant and the critical and probing reactions of an anthropologists. Successful collaboration amongst academics can leverage multiple depths of expertise that no single anthropologist can bring" and at the limit produce genuinely surprising results for everyone involved. This is just to say that despite the inability to distinguish between collaboration with informants and collaboration with colleagues, collaboration nonetheless still happens, and kind of like pornography, "I can't define it, but I know it when I see it."

Thus a third distinction: between coordination and collaboration. Coordination includes all those collective experimental technical practices that bring people and things into a particular kind of order. Disciplines, journals and conferences coordinate; calendars and email and letters coordinate; issues of access are issues of coordination" so jobs, letters of recommendation/introduction, internships, and all those forms of insinuating oneself into a field are about coordination. There has been less reflection on this practice than one might think. Almost all coordination practices are in the realm of the informally taught, the "just do it" school of methods; almost all successful coordination disappears precisely because the goal of coordination is to become transparent to the action of what coordination facilitates: collaboration.

Collaboration, by contrast, would depend absolutely on the forms of coordination in place. Collaboration is not amenable to control or planning" but it is also not possible without some kind of structure and play through which it takes on its shape as a game or as a project. A complete lack of coordination (and hence no collaboration) is the situationist drive" wandering flâneur and cultural studies' idiosyncratic observations of public matters. Too much coordination is the natural sciences laboratory: a hierarchical division of labor completely subordinated to the creative genius of the scientist, or worse, to the demands and exigencies of funding, fashion or politics" here collaboration is stringently produced and enforced. But in the middle, an amazing heterogeneous field of possibility. How to think carefully about controlling coordination" whether between anthropologists and informants, or amongst anthropologists" in order to maximize the potential for collaboration?

Collaboration is constrained and enabled by the kind of coordination in place. Students' projects are often failures when they choose a form of coordination that does not incorporate them in the heart of things: watching a corporation from the outside, for instance, rather than figuring out how to get inside it; or sticking with public sources and public culture rather than risking intimacy with families, communities, or organizations.
A current Rice Graduate student, for instance, studying Iranian NGOs in Washington, captures this beautifully when she discusses how hard it was to resist becoming an "intern" at one NGO, so that she could still openly associate with another one" one that eventually tried to incorporate her as a family member and in turn angered the first NGO which demanded transparency. Such a situation requires careful analysis after the fact" and the student is gifted at doing so" but it might also have benefited from more careful coordination before the fact, so that she could become more comfortable with the possibilities and limits of collaboration in the field.

While students are expected to collaborate with informants, the idea of collaboration between two or more graduate students is seen as deadly" dangerous to their careers because the system expects individual virtuosity first. This implies that collaboration is a zero-sum game, rather than the surprising production of something neither individual could have done on their own. What forms of coordination might facilitate collaboration that is not zero-sum? Collaborations that both/all parties can claim ownership of without diminishing it in another case? I think that the ARC discussion about concept-work is focused on these Utopian questions" but to me they are not first theoretical questions about the nature of what a concept is, so much as they are practical questions about the experimental structure of knowledge production: keeping track, auditing even, and making public the results of this work in a form common to all participants, but available to each of them as their own.

My own thinking on this subject is heavily influenced by my understanding of Free Software" and in particular my understanding of how coordination works in Free Software; the Linux operating system kernel is the canonical case. One of the most well known claims about Free Software is that it involves the collaboration of hundreds, if not thousands of volunteers who work together on the creation of highly complex software projects, like Linux. Common wisdom suggests that this mode of collaboration (or "peer production," to use Yochai Benkler's term) is somewhat anarchic and free-wheeling" a bazaar, not a cathedral. Further research has revealed that it is not anarchic at all, but that it does proceed in strikingly different ways than a conventional software company might. Although it is driven by volunteerism, and individuals are free to work or not work on whatever they choose, it is nonetheless extremely highly coordinated.

The Linux project, for instance, consists of a leader (Linus Torvalds) with a hierarchy of lieutenants each responsible for different parts of the kernel; a mailing list on which all participants are encouraged to communicate in implicitly structured ways (The Linux Kernel Mailing List, or LKML); and a source code management system (SCM) that coordinates collaboration by keeping track of who writes what and when" SCMs allow for a certain degree of automatic management of asynchronous, distributed contributions from participants around the world. There are, however, no goals and no planning. The project privileges a particular form of adaptability at all costs" whatever someone creates, it can be incorporated so long as it passes a series of tests having to do with a largely unarticulated, but learned, intuition about technical elegance, functionality, and the structure of the kernel itself. Torvalds and lieutenants facilitate this kind of contribution, but do not direct it. As a result, the Linux kernel does a great many things, some of them relevant only to very obscure architectures or uses, some of them useful to every user" but it was never designed to do any of them.

Of course one should ask: How do people know what to do? In some ways, this is the role of pedagogy: the construction of a disciplinary structure within which it makes sense to pursue one kind of problem rather than another. Linux makes sense because generations of students have been taught what an operating system is and should look like by studying UNIX (in the 80s and 90s) and Linux (today). The coordination of contributions to Linux is largely automatic and invisible. People learn what to do and how to do it and they simply do so. What emerges, sometimes, but not always, are forms of collaboration: co-work, co-labor, co-thinking about how to identify problems and functions, and how to solve them. Much of this work takes place on the LKML, simply as a kind of question and answer discussion, often with flame-wars around controversial topics. As people settle into these collaborations, coordination sets the stage: the structure of lieutenants, the mailing list and the SCM set the constraints around how that collaboration will unfold, and more importantly, keep track of it and manage it as an experiment. The success of a collaboration is in the outcome, not in the justification or planning" higher risk, higher reward, less bureaucracy and planning mentality.

ARC has a similar potential" as do any coordinations of people and tools in the scholarly world" a system that encourages a free software-like approach to unplanned problem-seeking investigative research, in an open but coordinated space. But this also raises a tricky question: is it clear what the operating system is, so to speak? A radical absence of such an organizing discipline would result in nothing more than a chaotic production of ideas and questions. But in fact ARC possesses a background that is resolutely ensconced in cultural anthropology and social theory, even as it seeks to break free from it. The "operating system" kernel is implicitly clear. To end on a playful note, that perhaps only crossover geek/scholars might appreciate, the background of ARC can be named: Weber and Foucault signal the core function of process management, the CPU as it were; others like Dewey (a file system, I presume) and Luhmann (memory management) form other key components; outside this kernel are loadable modules: Wittgenstein (framebuffer), Deleuze (virtual memory), Bourdieu (multi-tasking), Geetrz (garbage collection) etc. But the activity of ARC is hardly interested in such names per se" instead I would hope that they form the implicit ideal of an operating system, controlling in part what kinds of projects and potential collaborations make sense to existing and potential collaborators. It is unlikely that this core set of (continental, male, dead) names will change radically" just as it is unlikely that Linux will ever be anything other than a UNIX operating system" but it is just as likely that they will change slowly, in response to creative tuning and testing.