Knowledge, Organization, and Management: Building on the Work of Max Boisot
Edited by John Child and Martin Ihrig
I've reviewed Boisot's books elsewhere on this blog. This edited collection was put together shortly after Boisot's death, containing both Boisot's work and additional work by his collaborators, situated by his framework. Four of the chapters have been published elsewhere: three in prominent journals, one in a book chapter.
The book is in seven sections:
- Setting the stage
- Analyses of the Chinese system
- Organizational complexity
- The strategic management of knowledge
- Knowledge in big science
- Innovations in education
- Concluding reflections
With that introduction, let's get to the chapters.
In "Max Boisot and the Dynamic Evolution of Knowledge" (pp.3-16), Martin Ihrig and John Child say that "while Peter Drucker in the 1960s first drew attention to the fact that increasingly we are living in knowledge societies, Boisot provided a conceptual framework that enables us to appreciate the significance of that trend. His framework offers an understanding of how the ways we choose to express, communicate, and share knowledge are intrinsic to how we relate to one another in organizations and societies" (p.3). The framework started with Boisot's doctoral thesis in 1987, in which he described the codification-diffusion framework known as C-Space (p.3); later he added the third dimension of abstraction to yield I-Space (p.4). The authors describe I-Space this way:
The I-Space is a conceptual framework that facilitates the study of knowledge flows in diverse populations of "agents"—individuals, groups, firms, industries, alliances, governments, and nations. As one of Boisot's most fundamental innovations, it enabled him, and the many other researchers he inspired, to study and advance understanding of the emerging knowledge-based society and the implications of the information revolution. (p.4)
As an architect, the authors say, Boisot thought in terms of spaces (p.7). He applied his insights to other areas in concert with collaborators:
Once Boisot had identified an opportunity for an interesting project, he would try to bring people and organizations together, set up a framework, and then start to research and develop. Typically his collaborators were less well read and less theoretically versed than he was. On the other hand they often brought specialized empirical knowledge and a questioning perspective to the process, which helped to put Boisot's abstractions to the test of validity. (p.10)
In "From Fiefs to Clans and Network Capitalism: Explaining China's Emerging Economic Order" (pp.19-48), Boisot and Child use the C-Space (a four-field with axes of codification and diffusion; p.22) to conceptualize four types of transactional environments: bureaucracies, fiefs, markets, and clans. Boisot defines codification as "the selection and compression of data into stable structures" (p.21, citing Claude Shannon), and argues that "the codification and diffusion of information create a transactional environment that conditions the institutional possibilities to be found in different regions of the C-Space and endows them with some quite specific features" (p.21). For instance, when information is codified but undiffused, you tend to get bureaucracies, in which the diffusion of information is centrally controlled and relationships are impersonal; when information is both codified and diffused, you get markets, in which the diffusion of information is virtually unlimited and relationships are again impersonal (p.22). Obviously, this four-field diagram gives us ideal types and various configurations might exist. Critically, this is a conceptual framework, not an empirical one, so Boisot and Child do not give us guidance on where the boundaries are: at what point can we say that information is codified rather than uncodified? Diffused rather than undiffused? Those of us who conduct case studies can imagine a lot of situations in which the same text behaves differently depending on the frame and the other texts at play—but those issues are unexplored here.
We are, however, told that
abstraction is a prerequisite for the creation of robust codifications and the construction of a rational-legal order. If codification seeks to economize on data processing by assigning the data or experience to categories, abstraction seeks to economize on the number of categories used in the act of codifying. (p.25)
In "Analyses of the Chinese System" (pp.49-58), Child reflects that Boisot developed C-Space into I-Space in order to make sense of China's economic reform and the business systems emerging from it. Boisot based C-Space in part on a 1952 publication by Kroeber & Kluckhorn, in which they conclude that culture "described the ways that people structure and share information" (p.50). The 2x2 of C-Space naturally results in "four transactional or organizational modes" (p.50): one for each resulting quadrant. Later in this chapter, Child notes that Boisot collaborated with Guo Liang Xing, who closely observed "the activities of six enterprise directors, each for a period of six weeks in 1987" (p.56)—I'm not clear on the methodology, but this sounds like an empirical study I'll have to follow up on.
Child adds that Boisot's reliance on his root paradigm did pose some limitations. One was that
Max generally treated information as a cultural phenomenon, regarding the way it was articulated and shaped reflecting cultural norms. He called his original framework the "culture" space. When he depicted institutional or organizational arrangements in terms of different configurations of informational dimensions, he was implicitly regarding them as cultural constructions. This tended to overlook another factor that influences the shaping of institutions and organizations, namely power. (p.57)
Child goes on to allege that Boisot recognized this issue, but "put the issue to one side because it would unduly detract from the elegance of his framework. Yet three of the quadrants of the C-Space are structures of asymmetric power" (p.57; the exception is markets). Child notes that whereas Boisot overlooked power, the Chinese Communist Party did not, and the CCP's consolidation of power explains why China is not moving toward a market configuration (p.58)!
In "Extreme outcomes, connectivity, and power laws: Toward an econophysics of organization" (pp.61-92), Boisot and Bill McKelvey argue that organizational science is built to be nomothetic (i.e., to expect predictability via regularities), but this approach has let us down. After some discussion that I will skip over for lack of interest, they map three ontological regimes (ordered, complex, chaotic) onto the axes of Variety of Stimuli and Variety of Responses, forming "Ashby Space" (p.76). These are associated with strategies: routinizing, adaptive, and "headless chicken," respectively (p.78).
In "The creation and shaping of knowledge" (pp.109-128), Boisot argues that knowledge management is old in science, new in management (p.110). (Side note: Here and elsewhere, Boisot tends to portray science as ahead of the rest of the culture. This might be why he was so interested in collaborating with CERN.) He asks: why have we been so slow to knowledge management? And he argues that it's because
- knowledge isn't observable or measurable
- information and communication technologies have "led to the 'dematerialization' of economic activity -- the substitution of data and information for physical resources" in many areas
- "one cannot manage a knowledge resource as if it were a physical resource" (p.111)
He outlines three problems:
- What is being managed?
- Is knowledge a social phenomenon?
- How does knowledge relate to power? (p.112)
Although he says he doesn't have the space to deal with these problems in detail, he proffers the I-space as a conceptual framework to help address them (p.113).
What is knowledge, and how does it differ from data and information? Boisot asserts that "data can be viewed as a discernible difference between different energy states" and draws on Bateson to define information as the data that make a difference to someone, i.e., data that "will modify an agent's expectations and dispositions to act in particular ways," i.e., its "knowledge base" (p.113). For that agent to be knowledgeable, "its internal dispositions to act can be modified upon receipt of data that has some information value"—and here, inexplicably, he cites Latour and Woolgar (1979) (p.113).
Just a side note: it seems jolting to me that Boisot would ground his theory of information in Claude Shannon's work, then cite Bateson and Latour, whose understanding of information seems to be radically different. But such is the danger of an eclectic mind.
Thus, Boisot argues, knowledge doesn't flow; data does (p.114). Thus when he discusses "knowledge sharing," he "will actually be referring to some degree of resonance being achieved between the knowledge states of two or more agents following some sharing of data among them" (p.114). Knowledge is not dispositional and thus it doesn't have solid contours: perhaps two agents' understandings will loosely "resonate," but they will not be identical (p.114). "People are concerned with saving time and resources required to articulate and transmit knowledge. They are thus more likely to share knowledge that is clear and umambiguous than knowledge of a more tacit and elusive nature" —he gives the example of sharing stock market figures by fax as opposed to describing a Pollack painting over the phone (p.114). And "the articulation of knowledge, in effect, calls for two kinds of cognitive efforts: abstraction and codification" (p.114):
- "Abtraction either invokes or creates the minimum number of cognitive categories through which an agent makes sense of events": the fewer the number of categories, the more abstract its "apprehension of events" is (p.114).
- "Codification, by contrast, refines the categories that the agent invokes or creates so that it can use them efficiently and in discriminating ways. The fewer data an agent has to process to distinguish between categories, the more codified the categories that it has to draw upon" (p.115).
"Abstraction and codification are mutually reinforcing" and "the agent that is able to economize on its data processing resources through successive acts of codification and abstraction will be able to transact with other agents more economically and hence more extensively than will the agent that cannot" (p.115). Boisot adds that
A problem arises, however, when much of the knowledge that is of potential value to other agents is of a more tacit nature and hence not readily available to trading. Much of an organization's technological know-how, for example, may be of this kind. It is the fruit of a slow accumulation of idiosyncratic experience, and it resides in the heads or the behaviors of employees, working singly or in groups. (p.115)
Here and elsewhere, Boisot often refers to knowledge as residing in heads—which, honestly, makes his earlier reference to Latour even more baffling. On the next page, he says: "From an intellectual capital perspective, knowledge management is about the capture, storage, and retrieval of knowledge located either in the heads of employees, in the heads of outside collaborators, or in documents" (p.116). He concedes that
by their very nature, abstraction and codification are highly selective processes. Only a small part of a tacit knowledge base can ever be subject to articulation and structuring if genuine data processing economies are to be achieved. Thus, much tacit knowledge inevitably stays with its possessors whatever efforts at codifying and abstracting it have been subjected to—and much of this tacit knowledge will be valuable. (p.117)
Next, he gets to the question of "social learning," which "occurs when changes in the stocks of knowledge held by one or more agents in a given population trigger coordinate changes in the stocks of knowledge that are held by other agents in the population" (p.118). He briefly cites Piaget and Weick (p.122), then describes the social learning cycle as following an S-curve in I-space (p.123). The six steps are:
- scanning
- problem solving
- abstraction
- diffusion
- absorption
- impacting (p.124).
In their commentary on this piece, "The strategic management of knowledge" (pp.129-139), Martin Ihrig and Ian MacMillan note that after 2006, Boisot "focused on two areas: mapping critical knowledge assets, cultural and organizational structures, and associated learning paths; and simulating strategic knowledge management processes, in particular knowledge flows derived from knowledge-based agent interactions" (p.130). Mapping in I-Space "allows us to represent an agent's knowledge as a portfolio of knowledge assets, as a network of nodes and their links to other nodes" (p.136). Mapping these nodes in a network allows us "to consider its dynamic behavior" (p.136).
Skipping ahead, in "Knowledge in big science" (pp.155-166), Agusti Canals asserts something that I found key to understanding Boisot's theorizing: Boisot claimed that he had an "inability" to deal with mathematics (pp.155-156), and this inability kept him from a career in the natural sciences. So—and this is my commentary—Boisot repeatedly frames science as a vanguard for the rest of us to follow, and he prefers to model interactions in terms that resemble those of physics.
In "The three phases of Max Boisot's theorizing" (pp.205-211), John-Christopher Spender characterizes I-Space as suggesting "a complex economics of information, an information-based approach to political economy" (p.206). Spender notes that Boisot characterized information in energy terms (there's the physics influence again) and because I-Space was "self-contained" in energy terms,
the flow around the Social Learning Cycle ... 'worked' because the unit of information flowing could not be at all places in the cycle at the same time—it is trading-off the contrasting energy natures and values of the different types of information itemized in the Keio paper. There was a corresponding change in entropy as information moved around the cycle because in the real world, as opposed to the abstractions of neoclassical or "Newtonian economics," generating, transforming, codifying and deploying information is entropy-raising work. (p.206)
My commentary: This passage was tremendously illuminating to me because it emphasizes how Boisot's guiding metaphor of Newtonian physics (economics?) captured his theorizing and resulted in what seem to me to be very odd claims. Boisot uses I-Space to track transformations in information: how is "a unit of information" abstracted, diffused, etc.? But consider a study that Boisot briefly cited earlier, Latour and Woolgar's Laboratory Life. In that study, Latour describes cascades of rerepresentations, in which (for instance) the result of cutting off a rat's toes is represented in a column of numbers, then in a graph, then in other representations leading up to published papers and finally assertions that can be made in textbooks. Yet such representations are not just transformations in a single stream, they are yielded by combining previous representations, and they are themselves combined and compared to yield insights. They also don't disappear—scientists at the Pasteur Lab and elsewhere keep these representations so that they can unwind their arguments at any point, producing evidence at each point in the chain. In actual studies conducted by an ethnographer of science, we don't see a serial set of transformations but a tangled web, and information must continue to exist in different parts of the cycle at the same time. It is not the unit of information (whatever it might be—Boisot is vague about what constitutes information in the SLC) but the relations among different representations that makes science, and arguably other endeavors using complex information, work.
Let's leave it there. As always, I find Boisot's work to be fascinating but heterodox, and I look forward to continuing to think through its wrinkles. If the question of information interests you as well, definitely take a look.
(2003).
Posts
