Tuesday, July 03, 2012

Symmetry as a methodological move, part III

I've been discussing the principle of symmetry in actor-network theory on the blog. In previous posts, I've tried to use commonplace examples to demonstrate how, rather than being a form of pantheism or an antihumanist sociology, symmetry is simply a way of getting at certain relationships that are often overlooked or even buried in other approaches. When we examine these relationships in fairly restricted ways—weight requirements in elevators or falling bodies—symmetry makes a great deal of sense. But when we move to more complex phenomena, symmetry seems like a dicier proposition. 

For instance, let's consider cognition.

Latour famously proposed in Science in Action that we institute a ten-year moratorium on cognitive explanations of science and technology. (Hakkarainen criticized this moratorium in a 2003 Science and Education issue, complaining that Latour mentions this moratorium every so often, always extending it for ten years.) This moratorium came out of Latour's observations of scientific work, in which he noticed that scientists would assemble a large array of nonhumans and humans—instruments, procedures, lab techs, etc.—to perform science, then attribute the results to cognition. Without all of these nonhuman and human intermediaries, Latour argued, cognition could not have provided an adequate explanation.

Let's use an analogy. Suppose that as you walk home, you spot your neighbor on the roof. Do you think to yourself, "I am amazed that my neighbor can jump so high!" Or do you look for a ladder? Analogically speaking, Latour always looks for the ladder. He sees the alternate explanation—based entirely on human ability—as rather implausible. Yes, it's always a possibility, but (Latour might say) let's look for the ladder first.

This does not mean that actor-network theory is anti-cognition, as Hakkarainen and others have interpreted it to be. ANT is no more anti-cognition than a fork is anti-soup. 

No better demonstration of this point exists than Latour's positive review of Edwin Hutchins' 1995 book Cognition in the Wild. Latour declares: 
The results of Hutchins' inquiry are as devastating for psychology as the results of sociology were for epistemology or those of Shirley Strum for primatology. Everything that was crammed inside the mind of individuals is deployed outside and shared collectively with the culture, through the social connections and with the many cognitive artifacts the group has been able to devise. (p.55)
To put it another way: Hutchins, like Latour, looks for the ladder. 

But more than that, Hutchins refuses to separate cognition from mediating artifacts. In Hutchins' account, "cognition" is not confined to what happens inside the skull, but is instead distributed across a given environment. Latour summarizes the central claim of the book this way: 
cognition has nothing to do with the minds nor with individuals but with the propagation of representations through various media, which are coordinated by a very lightly equipped human subject working in a group, inside a culture, with many artifacts and who might have internalized some parts of the process. (p.56, Latour's italics).
To understand what's happening here, let's dive into Hutchins' book itself. Here's what he says about cognition in Chapter 2:
Having taken ship navigation as it is performed by a team on the bridge of a ship as the unit of cognitive analysis, I will attempt to apply the principal metaphor of cognitive science—cognition as computation—to the operation of this system. In so doing I do not make any special commitment to the nature of the computations that are going on inside individuals except to say that whatever happens there is part of a larger computational system. But I do believe that the computation observed in the activity of the larger system can be described in the way cognition has been traditionally described—that is, as computation realized through the creation, transformation, and propagation of representational states. (Hutchins, 1995, p. 49)
So Hutchins approaches the notion of cognition by defining it narrowly, as computation. In committing to this computational account, Hutchins takes a radically symmetrical perspective. Rather than drawing lines between what transpires inside and outside one's head, Hutchins examines computation—and thus propagation of representational states—wherever it happens. He brilliantly demonstrates that much of what we might have assumed was expertise or intelligence is instead attributable to how these representational states propagate. For instance, he argues that by all rights the sailors he studied should not have had the depth of experience needed to perform the navigational feats they performed; but these feats were attributable to the entire system of sailors plus other computational media. 

(As a side note, this book had a deep impact on me as I was planning the study for my doctoral dissertation. For instance, as I was developing the notion of the genre ecology in my own work, I ran across Freedman and Smart's 1997 article "Navigating the current of economic policy," in which they coined the term "genre ecology" and modeled it after Hutchins' concept of "tool ecology" in this book. After reading their article, I picked up Hutchins' book, and its influence was key to my understanding of genres as co-mediating each other.) 

To get at what Hutchins is trying to say, let's leave the ship for a moment and look at what he says late in the book about Searle's Chinese Room thought experiment. Imagine, he says, a room in which John Searle sits. Outside the room, people bring slips of paper on which they have written questions in Chinese. One by one, each person passes her or his slip of paper through a slot in the door; a few moments later, the person receives a reply, also written in Chinese. 

Does Searle know Chinese? Perhaps. Let's suppose that in one scenario, Searle knows how to read and write Chinese and is composing answers on the fly. In a second scenario, Searle doesn't know Chinese, but has several baskets of Chinese characters plus a rulebook that tells him how to combine them to respond to certain strings of characters. (The second scenario is very similar to that of Fred Saberhagen's first Berserker story, "Without a Thought.") In the second scenario, Hutchins argues, Searle doesn't know Chinese, but (for our purposes) the room does. Or to put it another way, the "cognitive properties of the sociocultural system" are different from "the cognitive properties of a person who is manipulating the elements of that system" (p.362). Cognition—in the narrowly defined sense that Hutchins uses the term—is stretched across the human agent and the environment, resulting in accomplishments that could not have resulted from the human agent alone. Searle has a ladder.

And with that, we can see why Latour does something that critics such as Hakkarainen haven't caught. He lifts the moratorium on cognitive explanations:
When I published Science in Action in 1987, I proposed a "moratorium" on cognitive explanations, which had been so freely and cheaply entertained by epistemologists. ... I did not know that I would be able to safely lift the ban, less than 10 years later, since in the meantime, cognitive explanations would have been dissolved past recognition [by Hutchins]." (p.62)


Mark Mills said...

>He brilliantly demonstrates that
>much of what we might have assumed
>was expertise or intelligence is
>instead attributable to how these
>representational states propagate.

What is a 'representational state'?

Can a scientific instrument measure or alter it?

Clay Spinuzzi said...

Hi, Mark. Yes. Hutchins uses a materialist, externalist approach to cognition. So, instead of speculating about internal mental states, he observes evidence of how information propagates in the environment. From the study, that evidence includes observable representational states such as written marks, positioning of tools, and speech.

For instance, in one chapter, he examines how the ship takes bearings, following a chain of representations from sighting specific landmarks (that is, identifying readings from a tool pointed at each landmark), to speaking those readings through an intercom, to writing those readings on scrap paper, to setting the hoey (a protractor-like tool) with those readings, to placing it on the map, and so forth. Each link in the chain is observable, verifiable, and measurable.

For that and many other reasons, it's a remarkable book that I can't recommend highly enough.

Mark Mills said...

Hello, Clay.

>instead of speculating about
>internal mental states,
>he observes evidence of how
>information propagates in
>the environment.

Good enough. Talking about how ' "cognition" is not confined to what happens inside the skull' certainly has advantages.

It isn't clear to me where you are going with this, though. I'm fine with the symmetry discussion, and I'm fine with the distributed cognition argument, but I'm puzzled by the combination. Assuming some would say 'cognition is in the head', then using 'head count' symmetry one could conclude 'more heads is always better', which certainly fails. Searle's riddles and Hutchins instrumentalism might suggest an explanation for that failure, but I suspect few would find the argument removes doubt. For example, some would agree that heads are never symmetrical, but there is no need to posit cognition outside the head. They might argue, atoms of 'cognition' are symmetrical, but some heads work at faster speeds (faster cpu cycles). It is an explanation that seems to conveniently rationalize the problem.

I'm left wondering if the subject is Hutchins, symmetry or a topic that has escaped me entirely.

Clay Spinuzzi said...

Right, I think that perhaps the underlying piece is the notion of mediation, although that notion is understood a bit differently in Latour, Hutchins, and (beneath it all) Vygotsky.

Some versions of hardcore cognitive psychology conceive of cognition as occurring strictly within the skull, and understand cognitive interaction with the outside world in terms of modeling parts of the world in working memory (and I apologize for the gross oversimplification here). In the Vygotskian tradition, however, interaction with the world is understood as mediated by physical and semiotic tools; the tools change the problem space, making feats of cognition possible that would not be possible otherwise—in other words, spreading cognition across the problem space.

An example of the dialogue between the two views shows up in Mind, Culture, and Activity 5.3, in which Ericsson and Simon are frankly puzzled by Smagorinsky's explanation of mediation. Why can't we just understand mediation as represented in working memory? they ask.

The answer according to Vygotsky is that mediation actually changes the task (a point made separately by Hutchins and—if memory serves—Norman).

Clay Spinuzzi said...

For an easy example, imagine having to multiply two 8-digit numbers. Doing this in our heads is very difficult; I couldn't do it. Doing it on paper is eminently possible—if we know an algorithm and can perform it on paper. The algorithm that I was taught in school involves placing the numbers in precise columns, then working right to left and knowing how to "carry" excess digits. Alternately, one could use alternate approaches using representational states, such as chisanbop or blocks or a calculator, each of which might involve a different understanding of the problem. (Not incidentally, one such representation system became the basis of writing.)

So mediation changes the task, making it simpler but also affecting it in other ways. Vygotsky argues for a dialectical understanding of cognition, in which the medium and the mediated (us) mutually affect each other, developing in tandem. Arguably, adopting a medium also involves adopting some of the sociocultural assumptions that went into its development. That is, as we expand our cognitive capabilities by adopting more and more mediators (tools, algorithms, texts), we ourselves tend to soak up the assumptions and values that go into learning how to use these mediators. (I go down this path myself in my first book.) That is, not just in cognition but in assumptions and values, human activity is no longer strictly human (i.e., residing in individual cognition, volition, etc.) but spread across the range of mediators that the human being uses in her or his activity.

And this leads to the question of agency, which is often the sticking point in discussions of Latour. That's because many who are willing to go as far as Vygotsky (understanding activity as mediated) are reluctant to go the next step, as Latour does (understanding agency as also mediated, as an emergent property of the system—that is, as symmetrical). (Again, I apologize for the gross oversimplification). That is, people understand themselves as making choices and initiating actions; they read Latour as claiming that those choices and actions are not theirs, but their tools'.

Latour doesn't help these people much because his style of argumentation is florid and metaphorical. My aim in this blog series is to break down the notion of symmetry into small chunks with everyday examples, and to examine it as a methodological stance that allows us to examine interactions from a certain angle rather than as a foundational, overarching explanation of human activity. That is, I want people to get comfortable with trying on symmetry to see how it fits, like trying on a pair of shoes, rather than feeling that they must adopt symmetry as a worldview, like converting to a new religion!

I hope this explanation doesn't wander too much—I'm dashing it off before replying to a ton of email. Please do write back if it seems unclear, unsatisfactory, or completely wrong-headed :)

Mark Mills said...


>in which the medium and the mediated (us)
>mutually affect each other, developing in tandem.

There is an interesting tension here between the notions 'mediation' and 'symmetry'. This reminds me of questions that keep me up at night: Is 'science' a search for a preexisting truth or a debate over heuristics? In the context of the symmetry essays, is the 2000 pound weight restriction a truth or a pointer proven over time to cause useful reactions?

>understanding agency as also mediated, as an emergent property
>of the system-that is, as symmetrical

It would probably be helpful for you to elaborate here. As I understand it, you are saying 'agency' is an emergent property or consequence of some sort of balanced mediation. I would be more comfortable if 'agency' was broadened to 'human agency'. This ties into another uncomfortable phrase used above, 'human activity is no longer strictly human'. It isn't clear to me that 'human activity' was ever 'strictly human'. Wouldn't it be just as logical to say 'human activity' and 'human agency' are emergent from mediated 'non-human agency', and never has been, nor will be 'strictly human'?

>My aim in this blog series is to break down the
>notion of symmetry into small chunks with everyday examples,

Ah! more of that tension between the 'always' and the 'emergent'. :-)


Clay Spinuzzi said...

Right, you've put your finger on one of the disagreements between Latoureans and Vygotskians: agency. I wrote about this disagreement in my book Network, and Kaptelinin and Nardi have another (smart, Vygotskian) take in Acting with Technology.

The Vygotskian take on mediation is that the human being acts, mediated by physical and psychological tools. In the final analysis, though, we're talking about human activity, activity that is directed and controlled by human beings. Although activity theory, which descends from Vygotsky, is often characterized as distributed cognition, at the heart of its understanding of cognition is the human being acting in collective human activity. That is, we're talking about human agency—and activity theory, despite its distribution of cognition, is ultimately an asymmetrical theory. This perspective is very amenable to humanists, which I think is part of why people in my field (professional communication) have been eager to embrace it.

The Latourean take on mediation is much more distributed, with agency being an emergent property of the system. That is, it is genuinely symmetrical, with any element of the system (human or nonhuman) potentially being an actor. Furthermore, "system" becomes harder to define, since any actor is also a network. For instance, we could examine a hammer as part of an actor-network, but we could also see the hammer's components as separate and important in different actor-networks—a hammer is suddenly differentiated into components, for instance, if we carry it past a strong magnetic field (which acts on just part of the hammer) or throw it into the fire (in which case part of the hammer will be burned up). That's a long way of saying that Latour's actor-network theory is ontologically focused.

Hutchins belongs to neither tradition, though he is conversant with both. But—and I thought this was the most interesting thing—he made the methodological choice to focus on propagation of representational states because that focus and perspective allowed him to get at the question he wanted to ask: how does the ship know what individual sailors don't?

And that brings us to the question you asked. Is science a search for truth or heuristics? For Hutchins—and, I think, for Latour—symmetry is a methodological move in that it allows us to set aside long-held assumptions and ask: what if we view it this way instead? And what we find, I think, is that it allows us to look at phenomena from very different angles and investigate very different components of a phenomenon. A scale and a spectrometer give you very different data, but that doesn't mean one is more truthful than the other. Similarly, these theoretical "instruments" let us get at different phenomena that we normally might not consider. Like a scale and a spectograph, these are not exclusive, nor are they irreconcilable.