Friday, January 28, 2022

Reading :: Information & Organizations

Information and Organizations: The Manager as Anthropologist
By Max Boisot

The link goes to a reseller, since this 1987 book is out of print. I bought my own copy used from Amazon, and to my delight, within its pages was a Visa carbon from 1989 and a few Esso Tiger Tokens. 

Boisot is really interesting to me: an architect turned management theorist who has contributed substantially to theories of knowledge. He has an annual award named after him. Nevertheless, I had not heard of him until I searched the literature for adhocracies in 2013 and found his book on ATLAS. There, he discusses the I-Space, a three-dimensional cube representing the codification, abstraction, and diffusion of information. Fascinating, but also heterodox from my understanding of information in a way that I had a hard time putting my finger on. I made a mental note to go back to this work later and figure out what was going on.

Well, now is later, so I picked up two more of Boisot's books as well as a collection put together in tribute to him after he passed away. This 1987 book is the earliest of the group. It did clear some things up, but it is not as sophisticated or well thought out as his later work, and parts of it I flatly disliked. 

Why? Well, let's recenter on what the book is and what it's trying to do. The book is pitched to a general business audience and according to the back cover, "Max Boisot demonstrates that organization culture is perhaps best understood when examined with some of the tools used by the anthropologist." The book is largely written in a style that may have appealed to a general business audience in 1987 but that introduces real difficulties for an academic audience in 2022. It wedges in a lot of strained whimsy like this one:

Would some knowledge of managerial anthropology help a practical, no-nonsense manager of, say, a packaging firm to manage? This is a tall order, particularly as by 'no-nonsense manager' we usually mean someone who, when he hears 'culture,' reaches for his gun. Can he be persuaded to wait a little while longer before shooting the piano player? (p.13)

The second issue comes just before this passage, as Boisot is explaining that we should apply the basics of cultural anthropology, not to our "exotic" and "scantily clad brethren" of West Africa or the Hopi, but rather to "the British suburban stockholder, the American marketing executive" (p.13). Skipping past the ethnocentrism, there's a point to be made, sure: we can apply cultural anthropology's toolkit to organizations in our own culture. Okay. Which tools? It turns out Boisot doesn't really have any toolkit in mind -- he simply means that we can think about our culture. He even asks: is it possible that in thinking about our information society, we have been doing cultural anthropology all this time without even knowing it? (p.17; he doesn't answer this question, which I think any anthropologist would answer with a resounding no). He summarizes by proffering the concepts in this book, which he says bear "more than a passing resemblance to those used by anthropologists to describe culture" (p.19). He doesn't name these anthropologists—and I suspect that he is referring to the tradition of structural anthropology, although I don't think the conceptual framework he is using, C-space (culture space), closely resembles that work. Maybe a structural anthropologist can weigh in. Later in the book, Boisot describes the famous psychologist Kurt Lewin as a sociologist (p.107)—so I'm not inclined to be charitable about how he's characterizing his antecedents in the social sciences.

Let's pause so I can emphasize that this book is early in his career. I don't think he makes similar mistakes in his later books. And these issues, as maddening as they are, shouldn't obscure the actual contributions of the book, which are in terms of thinking through and representing different types of information in organizations.

That work really starts in the next chapter, when he draws on the idea that all information is coded (Claude Shannon's work is not cited here, but his model of information transmission shows up uncredited on p.38). He claims that we are all "communicating in a code and implicitly assuming that the context in which we interact is sufficiently clear" (p.23). "Only a few of these codes involve speech and writing"—he includes the examples of shaking hands and smiling, or the impropriety of tap-dancing at a board meeting (p.24). When we choose the form in which to communicate, we must choose from a repertoire of codes available to us, and we tend to use the least effortful code (p.25). Codes both structure our experience and communicate it (p.25). And to select is to code (p.28). He argues that we can imagine codes on a scale from direct experience to compression into symbols — that is, from uncodified to codified (pp.29-30). The less codified something is, the more ambiguous and interpretive it is (p.35). 

This brings us to the "simple communication model" on p.38 (the Shannon model) and a brief discussion of "coauthors" Shannon and Weaver (p.40; this is a common misapprehension—Weaver popularized Shannon's model, but they didn't coauthor anything). Following Shannon, Boisot notes three problems:

  • The technical problem: "Is the message received the same as the message sent?" (p.40)
  • The semantic problem: "Is the message received understood?" (p.41)
  • The effectiveness problem: "Does the message lead to the desired behaviour?" (p.41)
So what we're seeing here is a quite structural understanding of communication and information, one that I suppose goes with the structural anthropology from which Boisot seems to have drawn some inspiration. I won't rehearse the issues with such a view at this point; if you'd like a critique, I commend Clifford Geertz' work as a starting place.

Here, Boisot introduces a second graduated scale: diffusion, i.e., how broadly the communication reaches potential recipients (p.45). Are we talking about a private wink between two people, or a broadcast that reaches a whole nation? Critically, he discusses "the diffusion of knowledge" (p.48), and in this early work, he slips between "information" and "knowledge" as if the two are the same. (In later books, he makes a sharp distinction between the two.) He also notes that codification is "the structuring and sharing of information [that] is the collective experience through which culture is built up" and in the private case, "when the experience being structured is to be found in one mind alone," we would only then use the term "coding" (p.49). 

He even provides a formula, D=f(C)

where D is the rate of diffusion per unit of time measured along the horizontal scale and C is the level of codification measured along the vertical scale. f is some function that describes the specific shape of the diffusion curve for a given distribution of the population, communication technology, and so on. We shall not be concerned with elaborating this function... (p.56)

In the tribute collection published after Boisot's death, one author notes that Boisot was not good at math. Although he may have been inspired by Shannon, a mathematician, I don't think that this formula does much work for us. Boisot repeatedly says here and elsewhere that the C-space and later I-space models are conceptual, not empirical, and he does not attempt to make them correspond to empirical measurements. So it's difficult to imagine how one might validate such a formula: how do you meaningfully measure the rate of diffusion or the level of codification? 

Nevertheless, this work lays the ground for the C-space, which has the two dimensions of codification and diffusion. (Those who are familiar with I-space will wonder where Abstraction went to. In this early work, abstraction is understood as a function or component of codification—see p.109. Only later does he separate the two.) And with these two dimensions, Boisot sets out a number of 2x2 grids. On p.67, he maps the quadrants to personal knowledge, common sense, proprietary knowledge, and public knowledge (discussed further on p.73). On p.78, he puts "minimum entropy" in the codified/undiffused corner and "maximum entropy" in the uncodified/diffused corner. On p.100, he puts types of transactions—fiefs, clans, bureaucracies, and markets—in the four corners. On 116, he maps different org units: the R&D department, the board, the production department, and the sales department. And on p.171. he puts capitalism in the codified/diffused corner and feudalism in the uncodified/undiffused corner. This hammer gets put to work driving in a lot of nails. (Boisot continues this tendency of universally applying his models to the end.)

 Let's back up to see what he's doing with this conceptual framework. In Ch.4, "Knowledge Cycles," he says "it is not knowledge that flows in the C-space but the information on which it feeds" (p.75—yes,  here he is trying to distinguish between the two). He argues that "a number of forces are at work which can be resolved into four vectors"—basically toward each of the corners of the C-space (p.75). These movements "irrigate the C-space with information" (p.77). Yet people forget codes ("who today can read cuneiform script?") and thus "the knowledge that builds up in the different parts of the C-space is subject to the action of forgetting, that gradual process of erosion that in physics goes under the name of entropy (p.77). He argues that entropy is at its maximum in the uncodified/diffused corner, and at its minimum in the codified/undiffused corner (p.78). He also argues that new knowledge has a cyclical flow in the C-space, involving scanning, problem-solving, diffusion, and absorption (p.80). 

Skipping ahead, he divides the codification scale in three: the most uncodified are words and gestures, more codified is writing, and even more codified are numbers (p.109). Similarly, he divides the diffusion scale: least diffuse is "knowledge available only to the firm itself," followed by "Knowledge available to the firm and its competitors" and finally "Knowledge available to the firm, its competitors and its customers" (p.110). He cautions that these can't be translated into a single, universal empirical scale (p.110). And he adds that the "line of least resistance for information flows" is on a diagonal line from uncodified/undiffuse to codified/diffuse. "A more vivid way of restating the proposition would be to say that it is easier to get certain types of data by diffing in the files than by opening up someone's cranium and rummaging around" (p.111). (In his later work, Boisot continues to regard "craniums" or "heads" as substrates that can contain information and knowledge.)

Moving on, he asserts later that organizational growth can be represented as "an internalization of transactions" (p.140). These transactions can be internal and external. Those who have read Boisot's later work and learned about his 3D I-space may be surprised to see a different 3D representation here: the axes are codification, diffusion, and internal/external (p.143). I'm not sure why he dropped this axis unless (a) he really wanted to reclaim that third dimension for Abstraction or (b) he didn't think internal/external was doing a lot of work for him. 

At the end of the book, he tells us that he had three objectives for this book, which he did not state at the beginning:

  1. "to sketch out in a simple and accessible outline a political economy of information consistent with its new role as the key resource in a modern competitive economy" (p.196)
  2. "to show that a political economy of information and a theory of cultural processes were in fact one and the same thing" (p.196)
  3. "to spell out in a tentative way some of the applications that the concepts presented might find in [a manager's] work" (p.196)
He acknowledges that his presentation of C-space included "simplifying assumptions" and did not include "the many qualifications that would weigh down a more academic work" (p.197). 

So how well does Boisot do? 

Understanding his audience, I was willing to cut Boisot a lot of slack in terms of simplifying assumptions and qualifications. But looking at his objectives, I find myself understanding his project much better—and I am thus more confident in how to evaluate it.

Objective 1: a political economy of information. I know very little about political economy. But for a political economy of information, the book does not sharply define "information" or adequately differentiate it from data or knowledge. Boisot does this in his later work, and he does it by extending the C-space rather than starting over. So I'll evaluate this first objective as a partial success.

Objective 2: show that a political economy of information and a theory of cultural processes are the same thing. Boisot does present his political economy of information as a theory of cultural processes. But I find this attempt to be unsatisfactory. His founding assumptions about information seem to be at odds with what we know about cultural knowledge and processes. He draws heavily on Shannon, but Shannon's work was mathematical rather than cultural, and reduces interpretation to decoding. He doesn't draw on other fields that could have really troubled and developed this account, including semiotics, interpretivist anthropology, or sociology. Tellingly, Boisot never gets to the level of examining fine-grained qualitative data in this book (and I don't think he does in his later work either). 

Objective 3: spell out applications for managers. Boisot did well enough that people are still talking about him, so I think his framework is appealing. Although I'm not writing as a manager, I personally am intrigued by the conceptual space of codification and diffusion (and I'll also throw in abstraction) as ways to think through how specific information artifacts are constructed and transported—although I understand "information" quite differently from how Boisot did in 1987.

I've written a lot in this review, and much has been critical. So I want to end by emphasizing that this book's concepts are intriguing enough for me to spend all of this time writing about it. The question isn't whether the book is good or bad—it's what job the book can do for me (and perhaps for you). What is the framework good for? What can't it do? In what ways can it inspire us to think about communication and culture differently? As you've seen, for me this book has been quite generative. And for that reason, if you're interested in exploring similar issues, I recommend it too—after you read Boisot's later stuff.