The In/visibility of Theory
Ab-, Infra-, Meta-: What does a prefix fix and how can it be shaken loose?
The Skyscraper and the Grapefruit
Theory in digital infrastructure is not so much invisible as it is in/visible—but we’ll defer this fine point for now. We begin instead with a couple of basic questions about one of the most popular, and likely hegemonic, theories informing and validating qualitative data analysis, grounded theory: what could be meant by “theory,” and what could it mean to “ground” it?
To limit this discussion of such potentially limitless questions, we focus our attention on a recent and one of the most cogent and compelling articulations of grounded theory by two sociologists, Stefan Timmermans and Iddo Tavory (2012). And even though our discussion of their refiguration of grounded theory as “abductive analysis” will lead into a discussion of our own different take on qualitative data analysis, which we put forward under the tag “ab-use,” we want to begin by noting our admiration for, our respect for, and our near-complete agreement with their grounded theory of abductive qualitative data analysis. It is gratifying to see the venerable and productive tradition of grounded theory so well presented, so convincingly argued for, and also so honestly critiqued and then re-formulated and strengthened. One of us (M.F.) had some training [don’t like that word] in grounded theory in graduate school, including a memorable and inspiring session with Anselm Strauss, who with Barney Glaser initiated this way of doing and thinking about constructing (sociological) theory from qualitative data. And like many researchers in the interdisciplinary field of science and technology studies, we have often invoked “a grounded theory approach” in grant proposals to the National Science Foundation and other funding agencies—partly because it was an accurate enough characterization, and partly because it seemed pragmatically necessary for the context, for reasons likely to become more evident below. (I.e, grounded theory is the more or less hegemonic form of qualitative data analysis in these funding worlds.) The differences between what we are pursuing with PECE and what Timmermans and Tavory are after with their “abductive analysis” may be small but are, in our Bateson-styled view, differences that make a difference. In addition, these small differences may make even more of a difference when these different styles of analysis are performed on digital infrastructures.
One of these differences concerns how “theory” is itself theorized, and the purposes and values that are imagined to adhere to theory by members of a given analytic culture—here, in a not-too-violently simplified categorization scheme, sociologists and cultural anthropologists. [[Note on “empirical humanists”? Or use it here?]] Timmermans and Tavory “approach” theory “informed” by the American philosophical tradition of pragmatism:
Our approach to the definition of theory is…informed by pragmatism. Our definition posits theory construction as an attempt to make sense of the world through generalizations of empirical phenomena. Theory as a specific kind of generalization allows us to move between data instances as well as to expect certain phenomena to happen and retrodict why certain phenomena have happened. (Timmermans and Tavory 2012, 182)
Theory may indeed be conceptualized as “a kind of generalization” – but this seems particular to sociology or a sociological thought-style, and there are other ways to conceptualize it. There are other ways to “make sense of the world” that don’t center on generalizing from the empirical.
Mike Fischer, in his recapitulation of anthropology’s evolving understandings of language, helps us see how these different conceptualizations of theory also index different language ideologies, or more broadly, semiotic ideologies. :
Behaviorist models take words and symbols to be unproblematic tokens, combined and rearranged in meaningful chains of sentences or utterances, done in turn-taking, stimulus-response sequences. Analysts can thus build up models of culture based on sets of belief statements made by actors. Symbolist models recognize that symbols are not univocal simple tokens but have fans of meanings, and that more is exchanged in any speech act than either speaker or receiver comprehends. Nonetheless, in symbolist models, symbols are still but more complex sign tokens – like overly full bouquets or pockets of fertile sediment – richly polysemic yet discrete. Indeed, the richest symbols are black holes: the entire culture is said to be condensed there. Symbolist analysts organize their models of culture around key symbols, symbol clusters, and nodes of semantic networks, somewhat like a crystal structure. There is a reassuring sense of relative stasis or stability of the symbolic system. Structuralist, and particularly poststructuralist, models decompose symbols and metaphors into chains of metonyms or association that play out into disseminating, ramifying, transmuting dynamics, attempting to model, in the structuralist case, the semantic-symbolic parameters of variation and transformation, and, in the poststructuralist case, the transmuting ambivalences of meaning that keep texts and communication labile (unless forcibly controlled, in which case poststructuralist deconstructive sensibilities highlight the tensions and processes of alternative meanings subversive to those intended and authorized by the controls).
In our design and construction of digital infrastructure we’ve found it helpful and indeed necessary to do this kind of analysis of analytic style, rendering more visible the conceptual attachments usually left implicit or even invisible. This is where the semiotic work of cultural anthropologists verges into the work of information scientists, as each work to elicit how other analytic cultures (geologists, astronomers, genomicists, etc.) think about data, about what can and should be done with it, and what values and potentials it holds or contributes to. (Poirier et al. 2019)
Another thing about Timmerman and Tavory: to the extent that it mentions or dwells at all on the materiality of either grounded theory or abductive analysis, it would be in analog form. It’s not at all clear, in other words, the infrastructure on which abductive analysis depends: scholars just seem to do it. “Writing” and “analysis” are in the end, for them, immaterial abstractions, or free-floating generalizations (like their view of theory). SO although they mention briefly computer assisted grounded theory, they don’t discuss it directly or in depth. So they don’t consider the possibility that when grounded theory is instantiated in CAQDAS platforms, abductive analysis may become more difficult or less likely to occur. One critique of CAQDAS is that the coding becomes the analysis, and abduction isn’t even in the picture – or in the platform. I.e., in digital infrastructures, grounded theory becomes more rigid and more structured and more identified with coding.
PECE’s response is: light structure, coding as question. And our goal is not to generalize the empirical into theory or even multiple theories, but: ? To create concepts (a deleuzian response)? To deepen and thicken the particular? To provoke? To “highlight the tensions and processes of alternative meanings subversive to those intended and authorized by the controls” (Fischer)? That is, to act as a driver of the “transmuting ambivalences of meaning,” rather than stabilize into a generalized or generalizable theory. And that may be the fundamental difference in style: stabilization, settling, consolidating, versus transmuting, disseminating, opening, multiplying. Abduction versus ab-use. “expecting certain phenomena to happen” versus creating a future anterior.
And another thing: they don't recognize or analyze differences in logical levels, or what happens at the transition. this is the other thing supposedly "designed into" PECE: the double bind [slash] set theory that says contradictions, misfires, etc happen when you change or move across logical levels
From a previous start:
We don’t disagree, we share it. We want something else, more. We’re not here to critique grounded theory, except in Spivak’s sense of critiquing something you cannot not want.
We start our discussion by noting the subhead: we’ve inserted the “/” to signify, as many before us have, an interruption to or swerve from a binary system of coding which presumes “invisible” is entirely distinct from and indeed opposed to “visible.” In/visibility is more of a flicker than a difference—with “flicker” itself referencing an additional in-between possibility in a scopic regime usually conceived as digital: lit or unlit, visible or invisible, present or absent, on or off. In/visibility momentarily unsettles habits of reading that naturalize meaning into stable and definitive patterns, figuring levels of visibility as an instability, a question to hesitate over, something begging to be further worked out in practices of inquiry.
The grapheme in/visibility (or “(in)visibility”) is also a sign of a body of theory, concepts, and scholarly ways of reading and writing that follow Gayatri Spivak’s dictum to “honor the slash” (Spivak XXXX:yy). Honoring the slash (for us) means to diligently signify the internal differences and disseminations that inhabit any sign, concept, code, system, phenomenon, or object, at the risk of alienating some readers averse to typographical innovation. To honor the slash is to engage, in other words, in “deconstruction.” To reference Spivak again: “Deconstruction, if one wants a formula, is among other things, a persistent critique of what one cannot not want.” One cannot not want the difference between visibility and invisibility, for example. One cannot not want signs and meanings to be stable and reliably consistent.
Or more pertinent to our analysis here: one cannot not code. (“Code” here is code for translating from one sign system to another, without remainder.) That is most obviously true when one has to write the computer code that is the software on and through which any digital infrastructure operates. But precisely what the sign “in/visibility” signifies with its slashed interiority are the limits of a system of binary coding for handling something that is neither this or that, and/or this and that. Writing computer code that can handle multiplicities and ambiguities is far from straightforward. A system predicated on decidability does not easily tolerate, let alone welcome, the ambiguous, the ambivalent, or the undecidable.
It’s also true that one cannot not code when one “codes” one’s (qualitative) data for (qualitative) analysis, where the “code” in marked quotes is and is not the same as the unmarked “code” of computer code. As designers and builders of digital infrastructure for the empirical humanities The intertwinings of the practices and theories of these two coding regimes—these two semiotic systems--are fundamental to the experimental infrastructural work we have been engaged with for many years now and which we are writing about here. The overarching question our experimental inquiry pursues may be summed up as: can one honor the slash in a system that has to disavow its existence in order to operate?
This is a good place to allow anthropologist Brian Larkin to remind us about the limits of another invisibility, in an often-repeated contention about the “invisibility of infrastructure” is at best a “partial truth;” “flatly untenable” and “fundamentally inaccurate” are other ways he codes such assertions. “Invisibility is certainly one aspect of infrastructure,” he acknowledges, “but it is only one and at the extreme edge of a range of visibilities that move from unseen to grand spectacles and everything in between.” (Larkin 2013:336) Mapping “everything in between” the conceptual limits of any system—limits that one cannot not want, even if one cannot actually have them—is another way of naming deconstructive critique.
Since we cannot not code, the question driving the development of PECE for us became: how might we code otherwise? How might we code within the limits of coding while questioning and pushing beyond those limits—limits for which, technically, there is no beyond?
Brown, S., and J. Simpson. 2013. “The Curious Identity of Michael Field and Its Implications for Humanities Research with the Semantic Web.” In 2013 IEEE International Conference on Big Data, 77–85. https://doi.org/10.1109/BigData.2013.6691674.
Timmermans, Stefan, and Iddo Tavory. 2012. “Theory Construction in Qualitative Research: From Grounded Theory to Abductive Analysis.” Sociological Theory 30 (3): 167–86. https://doi.org/10.1177/0735275112457914.
 See Brown and Simpson (2013) for a discussion, formative for our own attempts to think through these questions, of their design of digital humanities infrastructure that could handle multiple or ambiguous identities.