In 1994, full of graduate-student vim, I wrote a seminar paper, “Collective Knowledge, Inhuman Agency: Some Notes Toward Redeeming Transhumanization”. It was a bit of an eclectic exercise, pulling together Tom Collicott's photographs of ruined factories, Donald Norman's design ideas, Patricia McKillip's fantasy novels, Douglas Adams' non-fiction, some bits of postmodern and postcolonial theory, and a whole bunch of Donna Haraway. It tripped over a fair bit of ground in the process, but here's the first takeaway:
There are two observations here that I feel are crucial. First, some knowledges are collective, not situated in an individual but emerging only from a group of interacting entities. Second, these entities are heterogeneous.
So that's me, in 1994.
Theories agonistes. Who rules Bartertown?
In Network, Spinuzzi does a really quite nice job of explaining and mediating between Actor-Network Theory (ANT) and Activity Theory (AT), I think. In chapter 3 in particular he goes to quite some length to illuminate their differences and similarities.
Of course this rhetorical move — placing two competing theories side by side, then announcing it is not so much a contest as an attempt to elicit a productive tension, or let each highlight what the other obscures, or describe an explanatory field in the space between them, or what have you — is these days a familiar one. After the high-theory fandom period of the '80s became embarassing, many theorists were happy to adopt the role of Derrida's bricoleur, taking from this system and that. We do love our incommensurabilities.
And it's often a productive way to operate. I happily do it m'self.
Certainly it has its dangers. When you coax competing systems of ideas into interpenetration, you rupture their surface constraints that ensure their consistency, and discard the internal rigor that shapes their interpretations. Some worry that theoretical schemas are like bananas: once you skin 'em and throw away the bone, there's not much left to eat.
Fortunately, Spinuzzi's pretty careful. He respects the theories he's sent into Thunderdome. That'll be important here in a bit.
That bit of nostalgia I opened with has a point: I got a lot of mileage out of ANT back when I was a lit-theory student. It was one of my favored “structural meaning-making machines”, to paraphrase Donna H (from “The Promises of Monsters”). I read a bunch of Latour, and some Callon and Law. (I read some Serres, too, though frankly I was not impressed.) And it wasn't just for its explanatory power, either. As the subtitle of CKIA suggests, I saw a political project in ANT. I thought it represented a challenge to human exceptionalism that helped get us out of the Enlightenment mentality that undergirds modern imperialism.
And I mention all that because I'm going to pick a fight with it.
Sometimes ANT theorists aren't so thorough. I felt that way about Potts. Not a bad piece, but a bit thin on the theoretical side.
(Of course, at this historical moment it may be a bit harder to be sympathetic toward anything that smacks of celebrating the populist power of social media — if we're not all sick of that line by now, I can't imagine why not.)
Take Potts' one-page precis of ANT (6): where's the theory in your theory? Other than the rejection of human exceptionalism, in this summary it's basically a nomenclature for interactions and abstractions. Potts says “use this method”; but there isn't really a method there, per se.
A node is a node is a node. Except when it isn't.
In his first chapter, where Spinuzzi lets ANT and AT bloody one another's noses before he makes them talk it out in chapter 3, and things are nice and uncomplicated, it's pretty clear what they're fighting over: how to define the network.
And this is interesting, because here there's an (admittedly simplistic) version of the schism that doesn't quite return in the later parts of the book:
In actor-network theory, actor-networks are assemblages of
humans and nonhumans; any person, artifact, practice, or
assemblage of these is considered a node in the network and
indeed can be an actor-network in itself.
… Activity networks are linked activity systems … These activities themselves are the nodes, nodes that are constituted by, but transcend, the humans and nonhumans who participate in them. (7)
Actor network: nodes are actants, edges (connections, “links”) are translations between them. Activity network: nodes are activities, edges are causal flows between them.
It's a distortion, but perhaps not a tremendous one, to see the two definitions of networks as transpositions: the nodes of actor networks become the edges of activity networks, and so on. And that's interesting because network transposition is not an isomorphism. Actor networks and activity networks for the same domain are fundamentally, formally different. This isn't just a question of definition or perspective.
And that suggests that Spinuzzi's strategy here has the potential to produce a new hermeneutic, a kind of meaning that isn't available to either ANT or AT alone. They're not just two versions of the same story.
a rhizome is not amenable to any structural or generative model … The point is that a rhizome or multiplicity never allows itself to be overcoded, never has available a supplementary dimension over and above its number of lines, that is, over and above the multiplicity of numbers attached to the lines. … The second kind is very different, molecular and of the “rhizome” type. The diagonal frees itself, breaks or twists. The line no longer forms a contour, and instead passes between things, between points. … It draws a plane that has no more dimensions than that which crosses it.
That's Deluze & Guattari, A Thousand Plateaus, pages 12, 9, 505. What I want to point out here is that they're wrong. Provably, formally, specifically wrong, in what they say about “rhizomes” as abstract structures.
Much of what Deleuze & Guattari have to say about rhizomes is nonsense. That's going to be a bit of a problem.
If more actants were for actants …
For some, like Potts, ANT seems to boil down to “hey, things are actors too!". This is not a terrible starting place.
After all, it achieves the basic intervention of ANT, which is to pay attention to all those nonhumans we have delegated effectivity to and inscribed duties upon, and which prescribe definitions right back on to us.
But as Mol & Law point out, paying attention alone isn't sufficient to address the issues that ANT might help us intervene in.
Networks aren't the problem
with ANT or AT. And they aren't some cipher, a label to be draped
across any old object of study, the way “text”, say,
often is. There are actual material, objectively real networks.
There is a formal, logical, provably consistent theory of
networks: it's called graph theory, and it includes descriptions
of and operations upon abstractions of networks that have
material consequences. Eulerian circuits, Hamiltonian circuits,
algorithms like Spanning Tree and A
So it seems like, if you're going to talk about networks, it might be important to acknowledge that. And it might be important to distinguish this kind of specific, consistent, consequential understanding of networks, from vague handwaving guff about rhizomes.
Panic! at the network.
It's interesting that Spinuzzi, in his first chapter, identifies much of the tension between ANT and AT, and even within those theories, in the use of the word “network” itself:
The term network in the way I'm using it here — heavily influenced by actor-network theory and activity theory — is being abandoned right and left. … But what I want to emphasize here is that just as actor-network theorists have more or less jettisoned the term network because it had come to imply static structures, activity theorists are now beginning to question the term for the same reason — and imputing its structural connotations to actor-network theory itself! … So with scholars turning away from networks in different directions … why should we stick with the tired old notion of networks? … No wonder actor-network theory has a problem accounting for the stability of networks! Latour ruefully points this out … [Activity theorists] are grappling with the problem of too much stability in activity networks. (5, 6, 8, 23)
There are problems with the term “network”. It's in popular use, so there's scope for confusion; and for that matter it has various meanings in various technical fields. The simple fact that Spinuzzi has to spend half the book hashing out what ANT and AT mean by “network” indicates some of the difficulty here.
But that's not what causes the tensions Spinuzzi identifies between ANT and AT. Those arise from the consequences of what networks those theories identify in the domain of study, and how they describe them.
ANT's network becomes an alkahest, a universal solvent, dissolving everything into a grammar of relations among ciphers. The nodes of an ANT network become nothing but smaller networks themselves. This sort of system of empty terms, where all you can describe is the relationships among things, has been a familiar concept in European-derived thought at least since Saussure, and arguably at least since Heraclitus.
AT, on the other hand, seeks to identify networks where the nodes have the privileged content (activities). Links are reduced to flows of causality; effects are external to the network proper.
The problem isn't the concept of a network, much less identifying networks in the domain of study — since those networks do in fact materially exist. If there's a problem, it's in focusing exclusively on one kind of network, and putting all the explanatory weight on one aspect of a network (nodes or edges).
One solution is Spinuzzi's: bring two kinds of network analysis into dialog. What's not a solution is to turn to a model with less explanatory power, like the “rhizome”.
Bruno “Jim Johnson” Latour's elaboration of the relationships represented in actor networks is worth reviewing. Delegation, translation, prescription become, in graph-theoretical terms, “labels” for edges in the graph. They qualify how one actant node influences another along a particular line of effectivity.
Latour's operations are modal weights — indicating what kind of connection is made, what sort of effect it carries. And they're fuzzy arrows, showing the imbalance of power, the relative ease or difficulty with which an effect moves one way or the other along the edge. Actants aren't just connected; they're connected in contingently differing ways.
And this is one reason why the rhizome is the wrong kind of graph to represent actor networks. Edges in actor networks have distinguishing characteristics. They can't be substituted for each other.
The rise of the rhizome appears to be well underway in ANT and AT, judging from both sources Spinuzzi quotes (including Latour, though he notes how Latour has returned to “network” in Reassembling the Social), and his own recourse to it a number of times in the text. (For example, “Splicing is rhizomatic” (66) — as if that meant something, as if there were any sort of graph for which splicing was not an operation.)
The rhizome, as a theoretical construct, was introduced with great fanfare by Deleuze & Guattari in 1980's Mille Plateaux, which appeared in English in 1987, as A Thousand Plateaus: Capitalism and Schizophrenia. In the ensuing quarter-century it's become the sort of thing people gesture at uncritically: you know, it's that good ol' rhizome, straight outta D&G, a construct that needs no introduction and certainly no critique.
But if we take their substantive claims about the rhizome literally, we're confronted with the worrisome observation that they're wrong. Incorrect. No good. And if we excuse them by calling the rhizome a metaphor, a poetic conceit, then while it might be striking, it's a bit trite. Not much explanatory power there.
No, the rhizome has not become popular because it answers questions. It's become popular, I'll claim, because:
But whatever its rhetorical power, the rhizome isn't doing much intellectual work, as far as I can see. Not in Network, anyway. It's a distraction.
Ant primarily describes power, Spinuzzi says. He makes this claim in a few places: “actor-network theory is interested in how power works” (32); ANT, “of course, is preoccupied with how power works” (41). Of course!
That's not how I'd gloss ANT — certainly not the ANT work I've read. In my view, ANT is primarily concerned with effectivity, with having effects; and with responsibility, the expectation of having effects. Power is definitely important, but it's a secondary attribute, a function that describes the probability of successfully having an effect.
Still, I must admit that Mol & Law, particularly in the section “Non/coherences”, show how ANT can be an effective theory of power. Here they show how the various networks in which the body acts and is enacted offer both possibilities of power (control blood glucose, ward off complications, avoid hypos) and threats of the imposition of power by the unpredictable body (unforeseen variations in blood glucose, complications despite maintenance, hypos despite vigilance).
On the other hand, we can also read Mol & Law's description in the terms Latour supplies (in “Mixing Humans and Nonhumans Together”), of inscribing delegates and having those inscriptions returned to us as prescriptions.
And what I find particularly interesting about that is its similarity to Giddings' structuration. Consider Latour's final paragraph: "what defines our social relations is, for the most part, prescribed back to us by nonhumans". Our world of made things is the mirror that tells us how to be us. We delegate much of our practical consciousness, our knowing-how-to-go-on, to nonhuman actants, so they can tell it back to us.
Want an example of just how D&G are wrong about their rhizome? Well, I'll give you one anyway.
Recall this bit: “a rhizome is not amenable to any structural or generative model”. By “generative model” they mean something like a Context-Free Grammar (CFG), a model popularized by Noam Chomsky in the structural linguistics heyday of the 1960s. (We know because they spend a couple of pages complaining about it.)
Here's what a rhizome is: a highly-connected graph. That just means there are multiple ways to get from one node to another. (D&G suggest that a rhizome doesn't have any nodes, just edges, but that's meaningless; where edges connect you have a node, by definition.) Everything substantial that D&G say about the rhizome fits this representation, as do actual rhizomes.
A highly-connected graph is a particular kind of abstract machine, called a “transition graph” or “non-deterministic finite automaton”. “Finite automaton” may not sound as exciting as “rhizome”, but it smells just as sweet.
And here's the rub. A transition graph is clearly a “structural model”, so clearly the rhizome's pretty amenable to one. And worse, not only can any transition graph be represented by a context-free grammar, but CFGs are formally strictly more powerful than TGs. Chomsky's scorned generative model can do things the rhizome cannot.
This is traditionally demonstrated using something called the “Pumping Lemma”. Really, isn't that at least as cool as “rhizome”? Wouldn't you rather drop the Pumping Lemma casually in conversation? “So, he was all rhizome this and rhizome that, and I was just like ‘Pumping Lemma FTW!’”
OK, so what's the big deal with this rhizome thing? Does it really matter if sometimes Spinuzzi and other ANT- or AT-heads lean on a vapid concept?
Here (finally) is my point: it's symptomatic of a failing that weakens what is otherwise a strong work — a solid, innovative theoretical exegesis and an interesting case study. That's worth pointing out at some length for a few reasons. This is a useful book, describing a useful method, and before trying to reapply it we have to vet it, if we're to be rigorous scholars. More generally, considering the popularity of ANT and AT in this field, it's good to know where people go wrong with them. And finally, I suggested above that one of the strengths of the book is the respect Spinuzzi shows toward ANT and AT, and I'd like to return to that theme, because I think it's important.
The failing I'm referring to is an occasional slip in remembering that a network is a real thing, with strictures that inevitably arise out of the attributes that make it an instance of the category “networks”. In other words, sometimes Spinuzzi seems to forget the network-ness of networks, and produces a claim that is simply incompatible with it. In some cases this is just a silly construction in an aside or minor point; for example, “Here, history develops linearly … bifurcating like a tree” (118). Apparently this is a case where “linearly” means “not, in fact, linearly”!
A more telling example appears in the first chapter:
They simultaneously built the telecommunications network, the political-rhetorical network, and the developing social-cultural-historical network. Or rather, they built a single sociotechnical network, for though there is an abstract difference between these two sorts of networks, there is no practical difference.
This is a stirring formulation. The problem with it is that it cannot possibly be true.
The phone network is a real, physical network. It's constructed of durable material strung through space. Network formalisms apply to it, and altering it has specific economic costs and rewards. Social networks either aren't really networks — which would certainly be a practical difference from real networks — or they are; but in that case they have very different economic cost/benefit matrices than the physical network of wires and such. And that means that even if at some historical moment the physical network and the social network are isomorphic — have “no practical difference” — that situation will not last long.
It's simply vanishingly improbable that there won't be “practical differences” between the physical and social networks, if we require “network” to retain its substantial meaning, its graph-theory meaning.
Ultimately, I'd suggest, this is the danger with network theories. It's not that networks reduce everything to an empty system of relations, or that they are too rigid; it's not that they aren't good theories of development, or that they imply stasis. Spinuzzi ably demonstrates that all these failings can be corrected by mobilizing multiple views of the networks that inhere in the domain of interest.
No, the problem is that all too often we don't really commit to our analytic structures. They just serve as ideas of convenience; rather than being definitive, we let them be merely evocative. They let us support nearly any argument because we don't require anything more of them, and don't let them require anything of us.
Networks, as a theoretical construct, are actants in our theory-making networks; they're tools in our theory-activity networks. To realize their explanatory potential, their meaning-producing effects, we have to inscribe them with their fullness (all that we know about networks, including the formal strictures of graph theory), and we have to let them proscribe their proper operations on us, on the claims we seek to support.
I suggested that part of the strength of Spinuzzi's account of ANT and AT working together arises from his respect for those two bodies of theory. Rather than making one a straw man to argue for the other, or misreading one in terms of the goals of the other, he seeks an honest encounter with the meanings produced by each approach. We need to do the same with the idea of “network” regardless of what theoretical frame we employ it in.
Concepts, like doors, condense labor, extend through time, offer barrier and entry. Used properly they help us maintain an ethical order in the meanings we produce. We owe them, and ourselves, and the other participants in our networks, no less.