(Reading Roundup: Saarelma, Miettinen, Witte on Activity Theory)

Originally posted: Thu, 05 May 2005 09:39:38

Although I have a shelf full of books to read, I've put them on hold for a bit so I can read some really interesting articles on activity theory. Here are three:

Saarelma, Osmo. "Descriptions of Subjective Networks as a Mediator of Developmental Dialogue."

This short article appeared in the Quarterly Newsletter of the Laboratory of Comparative Human Cognition in 1993, and is frequently cited as an early example of activity networks -- linked activity systems. The case itself isn't terribly illuminating -- a description of different fields that intersect in health care --- but we can see both the roots of current "third-generation" AT theorizing here and some of the problems that the "third generation" is running into. Insightfully, Saarelma says that "the meaning of the links may be multiple" (p.103) -- the contacts of the different activities may be seen quite differently. So Saarelma's picture of the "idealized network of activity systems" (courtesy Engestrom 1987) is followed by several diagrams depicting the network as it is described by workers in the different fields (pp.107-108). Most of the activities depicted are the same, but the descriptions vary widely, with different links and different types of links. It's a fairly interesting comparison, and Saarelma resists the urge to collapse the perspectives into a unified diagram (although Saarelma does end with the standard AT move of mapping out contradictions). Multiplicity is allowed to remain in the analysis.

In all, this is a nicely methodical way of handling networks. It still assumes that activities = fields and that fields = nodes, but that simplifying assumption seems worthwhile in the interest of exploring these different perceptions. I can see why later AT work still cites it.

Miettinen, Reijo. "Object Construction and Networks in Research Work: The Case of Research on Cellulose-Degrading Enzymes."

Miettinen's work also deals with activity networks, but it comes a few years later, in 1998 (Social Studies of Science 28.3). Miettinen's study is broader too: it examines the work of a research community, which is to say a network of activities formed around a shared research object: cellulose-degrading enzymes. And it covers a much longer time, 1972-1998. Miettinen is apparently trying to position AT as a superior contender to actor-network theory as a way to understand science, technology, and society (STS), as he has done elsewhere.

Unfortunately, he's hampered by the clunky language of AT. Take this sentence: "The concept of object transcends the duality between subject and object" (p.424). Or this one: "An object ... can be either a means or an object in research activity" (p.431). In both cases, Miettinen is using the word "object" in two different ways -- as a "thing" vs. the object(ive) of an activity -- and this confusion obscures the more interesting implications of his work. Miettinen is studying "object formation," which is to say, what a network of activities is working on, how it construes the object of its work, what it is trying to understand and produce.

When Miettinen speaks of networks of activity, he delineates an activity as "an object-oriented, culturally and socially mediated system, with division of labour and rules that regulate interaction between the participating individuals. A work community, or a community of practitioners with a common object, can be viewed as an activity system" (p.424). Scientific research is one such community. And it "depends on other societal activities that provide it with problems, tools and funds, and utilizes its results" (p.424). That is to say, Miettinen is portraying an activity network as a set of activities that provide inputs (problems, tools, funds) or receive outputs (utilizes results). Unlike Saarelma, Miettinen isn't so concerned with multiplicity and seems uninterested in examining closely overlapping activities and boundary work.

It is perhaps for this reason that he levels his critique against ANT: "an analysis of network formation in terms of interests, goals, persuasion and enrolment is not enough. I propose that, if we are to understand the content of collaboration and the motives of participants, we need an analysis of the historically formed objects of participating activity systems, and also of the objects constructed together by them" (p.426). That is to say, Miettinen wants an account of network formation that allows for cultural-historical development, which is not surprising given his AT grounding. But the problem, I think, has to do with AT's handling of activity systems. As Witte argues (see below), the level of activity is an abstraction without clear empirical grounding for determining and circumscribing an activity. ANT gets around the question in a way that is perhaps a bit facile, but nonetheless fairly clear: rather than assuming stable social structures (such as activity systems), assume that actors are defined and locked in through a process of translation in which interests, goals, persuasion, and enrollment play key parts. That is, AT assumes stable social formations that develop over time and have distinct, unitary identities; ANT assumes that actants are all made on the fly and have to be persuaded to hang together, not just to each other, but in themselves. (See my upcoming TCQ piece for more on this.)

With this in mind, Miettinen's critique is somewhat tautological. Miettinen gets around the problem by assuming that networks are composed of persistent activities linked through culturally and historically developing relationships that are primarily characterizable as inputs and outputs (i.e., border work that does not touch the basic structure of the component activity systems). When he says that interests, etc. are not enough, actor-network theorists might turn around and say: persistent social structures such as activity systems are not enough, and that's why we have pursued a relationist sociology in which the pursuit of interests becomes a central explanation in the first place!

Miettinen goes further as he gets into the analysis, arguing that the concept of "interest" in ANT begs the question of where goals and interests come from in the first place. Miettinen, of course, points back to the developmental view: "an explanation of the motives for participating in a network presupposes an analysis of the history and dilemmas associated with the participating activity systems" (p.446). The ANT account lacks the stability required of a developmental account. "The motive for participating in the construction of an object can be understood in reference to the participating activity system's attempts to expand, redirect or transform its historically formed basic activity. The emerging new motive can be characterized as an attempt to solve dilemmas and problems in its basic activity, or to find a new possibility for extending that activity" (pp.445-446). That is, activities are transformed and develop new motives by working out contradictions. "These new motives can be characterized as transformative interests, because they are related to the transformation of a collective activity system. Consequently, they cannot be analyzed in terms of the goals of individuals" (p.446). Here, Miettinen is assuming and ascribing the view of extreme asymmetry that he has previously accused ANT of holding; this is based on his understanding of Machiavelli (p.453). ANT scholars have typically argued for symmetry, in which humans, nonhumans, and collectives of these can be understood as actants; interests are not necessarily related to individuals, but can be applied and have been applied to organizations, scallops, variable-reluctance motors, and various other actants (and not just as "a rhetorical resource used by human actors in their controversies," as Miettinen says (p.453)!).

But the symmetrical, interest-based analysis provided by ANT simply doesn't allow the developmental perspective that interests Miettinen. "These networks are important arenas of learning, boundary-crossing and the creation of new ideas and artifacts. They are also a central mechanism for connecting research work to other societal activities -- an unavoidable issue in studying the relationship of science to society" (p.451). And:

To understand the motives of the participants of these innovation networks, one needs to analyze each participating activity system and its relation to the object to be constructed in the network. Participation in the construction can be understood as an attempt to extend, redirect or transform the participants' historically formed activity systems. Interests may be characterized as an attempt to solve emerging problems and contradictions of the activity, to extend that activity. (p.451)

Whereas ANT is totally unconcerned with learning and cognition, AT is quite concerned with both. So Miettinen takes it as a given that any proper account of science take into account learning, cognition, development, etc.

I've concentrated mostly on Miettinen's critique of ANT, which frustrates me because Miettinen is obviously smart and well read in the ANT literature, but doesn't quite seem to get what the project is about, and therefore ends up criticizing ANT in AT's terms. Of course ANT will not compare well with AT when AT serves as its own exemplar! But Miettinen's analysis itself is quite sharp and displays some of the real advantages of an AT approach. Miettinen persuasively argues that the scientific work in the case passes through several models of artifacts: from "where-to" to "why" to "where-to," in Wartofsky's terms. Whatever the merits of the ANT critique, the AT work is interesting and valuable.

Witte, Stephen. "Research in Activity: An Analysis of Speed Bumps as Mediational Means"

Several days ago, I described Bracewell & Witte's article critiquing activity theory. Witte recently passed away, so some of the work he produced in his last days has been edited and published; this piece, like the Bracewell & Witte piece, appears in Written Communication, of which he was co-editor. Based on an invited lecture, this piece lays out Witte's sympathetic critique of activity theory in broad terms. Witte was one of the most gifted thinkers in rhetoric, and he has some very interesting things to say here; I wish we were able to have a conversation about this piece, which is both perceptive and sometimes a little dismissive.

Witte frames his claims as tentative:

The analysis I present in this article should be seen as preliminary and exploratory; it is not (yet) an argument but rather an argument in the making. However, should an argument eventually be developed from this preliminary analysis, it would include at least these three tentative claims:(a) literacy cannot be understood by focusing on literacy directly because that approach inevitably turns literacy into its own explanation; (b) literacy cannot be understood as activity because that procedure yields a fusion of part and whole such that activity becomes its own explanation; and (c) the unit-of-analysis problems that underlie accounts that explain literacy in terms of literacy or which explain literacy as activity will require the development of new units of analysis for studying literacy from an activity theoretic point of view. (p.128)

This seems like false modesty, since Witte then tears into AT with a vengeance. Or at least he strongly critiques its usefulness for questions of literacy, and intends to locate "some of the weaknesses in the thinking of those whose uses of activity theory resemble the antics of a kitten that has just discovered that a ball of twine can be a plaything" (p.130). (I can't help but feel implicated.)

As in the Bracewell and Witte piece, he argues that a disjuncture between Vygotskian-Lurian and Leontev'an strands of thought has muddied the waters considerably. In particular, Leont'ev's contribution of the activity system, further developed by Engestrom, really problematized things because activity systems are nearly impossible to apply as anything other than a theoretical concept. The activity system as unit of analysis has both focus and boundary problems, especially in research design. Witte asserts that the criteria for focus and boundary questions seem to come from outside the model rather than within it (p.140).

The boundary problem is particularly vexing at the level of actions "because in the context of overlapping activities, one can find no principled way of aligning a given action unambiguously with a given activity" (p.141). But the problem exists at all levels. Take for instance -- and this is my example here, not Witte's -- a customer service representative at a telephone company who is being shadowed by a new employee. When she uses her tools and talks to customers, is she training the employee or serving customers? And since she knows that the new employee may be asked about his training experience, is she thinking about a third activity involving self-promotion? If she speeds up her pace near the end of her shift, is it because she is thinking about her home activities and trying to get out the door in time? Each operation and action could potentially be multivalent, and there's no methodologically sound way to firmly separate them.

The unit-of-analysis problem is similarly vexing:

If every human engagement with the social or material world is part of an activity or an activity system(and all such engagements may well be), as a researcher I amleft with two equally unsatisfactory research strategies, given the two writers? theoretical models. Either I can offer a description of some activity around which I?ve drawn some arbitrary boundaries, knowing that my description is not in and of itself an explanation, or I can explain activity in terms of activity, in which case the object of analysis becomes indistinguishable from the analysis itself. What is missing in Engeström?s and Leont?ev?s theories of activity is, in effect, the specification of a unit or units of analysis that would allow me to account for or to explain activity in terms of something other than itself. Tautology is not just a problem in logic and in formal definition; it is also a problem when one tries to transform activity theory into a program of research. (p.141)

Activity, he charges, cannot function as a unit of analysis; it can be known only through observed effects or consequences, like gravity (p.142).

At this point, Witte goes into the example of the speed bump. This is familiar territory if you've read your Latour, who talks about the "sleeping policeman" as mediator. The examination is a good one, but didn't shed a lot of new light for me. But at the end of this discussion, he does get to the point that "any given activity never exists independently of other activities" and says that Edwin Hutchins' and Christina Haas'

analyses of mediational tools, technologies, and / or artifacts similarly discover a range of activities and activity systems informing, circulating about, and penetrating the mediational means they explore. This interdependence among activities may?if I can be so bold to reach beyond the boundaries of my own data and theirs?provide the objective basis for what might be termed the continuity of mental life and the continuity of social life. (p.154)

Yes, yes. Witte wants to press forward into the realm of networks -- we could even say activity networks -- and Witte certainly wants to explore them in terms of Vygotskian mediation (which he finds to be a much more congenial way to talk about activity than levels). He does not cite any of the "third generation" work on activity networks in particular, though I assume he was familiar with the relevant literature. My take is that he simply rejected it because he saw too much of the input/output school (see Miettinen above) and not enough consideration of the dialogic, multivalent nature of these linkages. Witte's main concern was literacy, and I find his work here to be congenial with Callon's discussions of texts as circulators as well as Star's discussions of texts as boundary objects.