Reading :: Collisions and Collaboration

Collisions and Collaboration: The Organization of Learning in the ATLAS Experiment at the LHC
Edited by Max Boisot, Markus Nordberg, Said Yami, and Bertrand Nicquevert

I picked this book up primarily because it was one of the few recent books out there to use the term "adhocracy"—a term that I'm currently researching for a project. But it turns out that Boisot is quite well known for his writings on information and strategy (though not so well known in the circles where I typically read and publish, alas). After reading this book, I can understand why: his I-Space framework is useful for thinking through the relationships between organizational characteristics and the circulation of information.

Let's talk about I-Space first, then get back to the book project.

I-Space is depicted as a three-dimensional space, a cube with three axes:

• Codification: from very uncodified to very codified information. "Codification is indexed by the amount of data-processing required to distinguish between categories and to assign events to these" (p.33).
• Abstraction: from very concrete to very abstract information. "Abstraction is indexed by the number of categories required to perform a given categorical assignment." (p.33)
• Diffusion: from very concentrated to very diffuse. Here, we're talking about the degree to which information can be diffused over time. "The greater the degree of codification and abstraction achieved for a given message, the larger the population of agents that can be reached by diffusion  per unit of time." (p.34)

Within this three-dimensional space, Boisot et al. say, we can map a social learning cycle. This cycle follows six phases:

1. Scanning
2. Codification
3. Abstraction
4. Diffusion
5. Absorption
6. Impacting (p.39)

Mapped in I-Space, the social learning cycle typically looks like an S-curve. But it looks different for each organization, since each organization's I-Space tends to be different—that is, different organizations have different characteristics of codification, abstraction, and diffusion. Indeed, "to the extent that individual agents can each belong to several groups, each locatable in its own I-Space, they will participate in several SLCs that interact to form eddies and currents" (p.38).  

In fact, if we map parts of the I-Space to cultures and institutional structures, we find that different structures have different "homes" in the I-Space:

Fiefs. These thrive in situations with concrete, undiffused, uncodified information. Characteristics:

• "Information diffusion limited by lack of codification to face-to-face relationship." 
• "Relationships personal and hierarchical"
• "Submission to superordinate goals"
• "Hierarchical coordination"
• "Necessity to share values and beliefs"

Bureaucracies. These thrive in situations with abstract, undiffused, codified information. Characteristics:

• "Information diffusion limited and under central control"
• "Relationships impersonal and hierarchical"
• "Submission to superordinate goals"
• "Hierarchical coordination"
• "No necessity to share values and beliefs"

Markets. These thrive in situations with abstract, diffused, codified information. Characteristics:

• "Information widely diffused, no control"
• "Relationships impersonal and competitive"
• "No superordinate goals—each one for himself"
• "Horizontal coordination through self-regulation"
• "No necessity to share values and beliefs"

Clans. These thrive in situations with concrete, diffused, uncodified information. Characteristics:

• "Information is diffused but still limited by lack of codification to face-to-face relationship." 
• "Relationships personal but non-hierarchical"
• "Goals are shared through a process of negotiation"
• "Horizontal coordination through negotiation"
• "Necessity to share values and beliefs" (p.49)

These different institutional types are similar to those of Ronfeldt's TIMN, Cameron and Quinn's Competing Values Framework, and Mintzberg's categories of structures, but they're defined via the three axes in the I-Space. Essentially, Boisot et al. are arguing that certain institutional structures thrive because they best adapt to specific conditions in a given I-Space. 

Yet they also posit that modern information and communication technologies (ICTs) are shifting the curve, making it easier to diffuse messages and thus to reach more people at a lower level of codification and abstraction (pp.50-51). This shift is a big deal. Up to this point, to be widely diffused, knowledge had to be highly codified and abstracted—something that naturally favored bureaucracies and markets, which are set up for those conditions. But due to ICTs, the shift in the curve might favor clans as well as markets, to the comparative detriment of bureaucracies (p.51). "Over time, this is likely to favor clan-like network cultures that have a tendency to closure ... rather than the more open market processes" (p.51).

So there's the I-Space framework. In this collection, Boisot (who is listed as an author on every chapter) and his coauthors apply this framework to the ATLAS collaboration, a complex multunational scientific organization that uses ATLAS, a high-energy physics detector at CERN; it forms part of the Large Hadrion Collider (LHC) (p.8). This organization coordinates without central managerial authority (p.55). 

In Chapter 3, the authors examine the ATLAS collaboration as an adhocracy, drawing from Mintzberg to identify the structure and then locating it "to the right of the region in the I-Space labeled 'clans'" (pp.74-75)—that is, further diffused than clans are. "As a geographically dispersed, loosely coupled adhocracy, the ATLAS Collaboration extends beyond the bounds of clan cultures as conventionally understood. Yet ... it remains driven in large part by clan values and practices" (p.74). Later, in Chapter 4, the authors add: 

As a loosely coupled adhocracy, ATLAS operates to the right of clans along the diffusion scale of the I-Space, in the region where, on account of the lack of structure of the knowledge being exchanged and the large number of interacting players, things could quickly become chaotic. The collaboration remains culturally cohesive, however, held together by shared commitments, norms, and a common focus on an infrastructure of boundary objects that over time deliver a clan totem: the detector itself. (p.114).

Boisot et al. credit modern ICTs for the success of ATLAS' adhocracy: "It is mainly the bandwidth effect [in which more people can be reached per unit of time] that makes giant adhocracies like the ATLAS Collaboration possible, and that is likely to give birth to a new scientific culture" (p.257). For instance, such adhocracies can involve "a re-personalization of science, a move away from the impersonality of anonymous peer reviewing and journal publications as the only way to get on and towards the building-up and exploitation of personal networks" (p.263). Via ICTs, "the culture of clans can be extended to more loosely coupled and geographically scattered adhocracies" (p.264).

Let's inject a bit of caution here. Many of the authors are involved in ATLAS or other aspects of CERN; this collection is not (exclusively) an outsider's account and shouldn't be considered a set of studies per se. In tone, the contributions are often a bit congratulatory.

But on the other hand, the collection does a great job of explaining and illustrating I-Space. If you're interested in how organizations work, or how I-Space works, this book is a good solid introduction with some intriguing analysis.