orgtheory.net

cognition as networks

With the increasing interest in the cognitive structures that underlie organizational and market activity, I think it is important to take a step back and think more carefully about the constructs and how they are used.  A quick glance at the literature reveals a long inventory of cognitive structures used in research – categories, frames, schemas, logics, scripts, recipes, etc …  Currently within organizational theory there is much emphasis on categories and how they constrain and enable market behavior.  While there certainly is good reason why categories should matter, we should ask ourselves if categories are the right unit of analysis.

Let me propose that a more tractable way of thinking about cognition is to treat categories as embedded within a broader network of other categories through a series of relationships – essentially what is called a schema.  Empirically, this means capturing the nouns/phrases as categories and the verbs as relations that connect them.  See Kathleen Carley and colleagues efforts to draw such cognitive maps.  In my own work, with Christopher Bingham, we have applied this technique to analyze how the insurance industry conceptualized the early business computer.  But, what do we gain by looking at the network as opposed to the individual categories?

One reason why considering the conceptual network is important is that cognitive mechanisms, such as analogies, leverage the relational structure and not the category structure.  Gentner and colleagues have characterized analogies as mapping a relational structure between something familiar and the new concept.  Something new is familiar because of the shared relational structure as opposed to sharing the same category. In fact, in our work with the computer, we observe two distinct analogies – one comparing the computer to existing office machinery and the other, to the human brain.   Just focusing on categories would miss this powerful mechanism to expand and develop new categories.

Another reason has to do with meaning.  With respect to meaning, focusing on categories places an emphasis on identifying the salient characteristics that define the category as well as boundary maintenance.  However, our research has shown that the meaning of relations and categories can change not because of internal changes to the category itself but because of changes in its associations.  For example, early on the relation “verify” was associated with clerical workers and office machinery and took on the meaning of checking up on the office machinery to make sure it was doing what it was supposed to.  Over time, what “verify” was associated with changed.  Programs replaced clerks such that computers would self-verify and other kinds of users and information were added.  “verify” shifted from confirmation through comparison between humans and computers to the computer acting as the ultimate proof.

Third, by thinking of cognitive networks encourages us to think more deeply about higher-order structures that influence cognition (see Klaus Weber).  For instance, Martin Ruef has argued the new categories are likely to emerge in less densely populated areas of the network.  Another intriguing approach would be to incorporate network measures in our analysis of cognition.  For example, how do clique structures respond to new information as opposed to more diverse cognitive structures?  At what point are relational structures developed in relation to categories?

I am curious, what do you think of the advantages and disadvantages of thinking about cognition as networks?

Written by Steve Kahl

July 14, 2011 at 10:49 pm

5 Responses

Subscribe to comments with RSS.

  1. I love it.

    http://brokensymmetry.typepad.com/broken_symmetry/2009/05/an-organic-economic-theory-an-economy-as-a-network-of-networks.html

    There is a need to be precise, and I’m not so familiar with the “category,” “schema,” &c. jargon. But there is no doubt to my mind that cognitive structure reflects a dynamic process of mimicry and feedback.

    Andy Clark is the clearest I have read on cognitive structure as it relates to the external world:

    Click to access burrowspaper.pdf

    But I haven’t seen Andy or others really focus on the details of topology and dynamics of social networks in cognitive frameworks.

    Like

    Michael F. Martin

    July 14, 2011 at 11:02 pm

  2. Here are some interesting source materials on the dynamics of networks, coincidentally published today:
    http://markorodriguez.com/2011/07/14/graphs-brains-and-gremlin/

    Category structures are cognitive models produced naturally by the network structures of the brain – degenerate networks in a sense. Network structures on the other hand are new and cognitively intractable. They are definitely the path to higher-order understanding, but I think we’re still missing the conceptual tools to think beyond data graphs and statistics.

    Like

    Rick Thomas

    July 15, 2011 at 4:53 am

  3. […] cognition as networks […]

    Like

  4. In a way, your description of categories and networks reminds me of the ANT description given by Latour at EGOS last week. At least visually, he also imagines categories and networks co-evolving. The main difference, I think (and I confess not to understand ANT very well), is that ANT doesn’t have a theory of cognition underlying it.

    There is other work out there that shows that categories/culture may influence the relational structure, rather than the reverse always being true (e.g., Vaisey and Lizardo). In Srivastava’s and Banaji’s recent ASR piece, they explicitly argue that the pathway between culture and relational structure is cognitive. That is, how people categorize themselves shapes how they view themselves, which in turn affects who they form collaborative relationships with.

    Like

    brayden king

    July 15, 2011 at 3:07 pm

  5. Great question, Steve.

    I agree with Michael that there is a need to be precise in conversations like this.

    In particular, I find it very helpful to distinguish between “cognition” as a noun versus as a verb.

    To explain, cognition is often used as a noun to refer either to the relatively durable contents of socially shared classification systems, or to individually held copies of this knowledge. At the same time, cognition is also used to refer to more action-oriented aspects of knowing about something. This means the same word is used to describe both knowledge and the processes of knowing it, and this can be confusing. (The OED definition of cognition is, “The action or faculty of knowing; knowledge, consciousness; acquaintance with a subject.”)

    This distinction helps me understand how “cognition” both reproduces extant patterns of social interaction and generates new ones. That is, cognition-as-knowledge (the noun) tends to focus the process of knowing things toward either what we already know or toward things suggested by what we already know, but cognition-as-knowing at least theoretically permits us to jump those rails and sometimes know and do new things—even collectively.

    (As an aside to Brayden’s post, I was also struck by Bruno LaTour’s talk last week at how compatible much of his perspective is with what I would recognize as relational sociology. I’m not sure Brayden’s right that cognition isn’t really part of ANT, but I’m not ANT expert either. I’d be willing to suggest, however, that the “collecting” function LaTour describes is cognition-as-knowing that draws on cognition-as-knowledge. Whether that counts as a theory of cognition or not, well, probably not.)

    All that is to say, one of the disadvantages of thinking about “cognition as networks” is the potential for confusion over what we mean by cognition (noun or verb), but in my view, being clear about that clears the way for productive inquiry into how the structure and dynamics of network models of cognition (both types) not only support the durability of culture or social structure, but also how they support change.

    Like

    mtkennedy

    July 18, 2011 at 6:37 pm


Comments are closed.

%d bloggers like this: