This book is a thoughtful look at how organizations acquire knowledge. I'll describe the main thrusts of their argument, and consider how it relates to software development.
Their thesis is that Japanese and Western societies and companies have developed different understandings of what it means to learn and to innovate, but the way forward is through a synthesis of both styles.
Knowledge is taken as the basis for what an organization does, but it's important to know that creating knowledge can be as important as processing knowledge.
A key idea is that some knowledge is tacit (i.e., internalized) and other is explicit. Western philosophy is described as having struggled to understand whether knowledge is based on what we experience (empiricism) or inherent truths (rationalism), and to have focused more on explicit knowledge. Japanese thought has tended to treat tacit knowledge as more important (but not really addressed epistemology as a philosophical tradition).
The quote Peter Drucker (Post-Capitalist Society, p. 24), saying: a skill "could not be explained in words, whether spoken or written. It could only be demonstrated." And, "the only way to learn a techne [skill, in Greek] was through apprenticeship and experience."
Knowledge creation is the process of making tacit knowledge explicit. The example of the Honda "Tall Boy" (a new car style that moves away from the sedan look) demonstrates how this can happen. First, metaphors and analogies are used to help provide a framework for thinking about things that can't be easily described. "As such, metaphor is highly effective in fostering direct commitment to the creative process in the early stages of knowledge creation." ("Contradictions inherent in a metaphor are then harmonized by analogy.") Second, knowledge moves from an individual to an organization; teams provide a shared context and multiple points of view. Finally, ambiguity and redundancy are used: ambiguity lets many things fit within a framework; redundancy lets a group explore many aspects of a direction before deciding where to go.
The authors consider the learning theories of Bateson, and Argyris and Schon: single-loop learning and double-loop learning. "From our viewpoint, the creation of knowledge certainly involves interaction between these two kinds of learning, which forms a kind of dynamic spiral." (p. 44) Nonaka and Takeuchi are not as enamored of Senge's system learning; they think it doesn't escape the Cartesian viewpoint.
There are two key dimensions in a theory of organizational knowledge creation: an epistemological dimension--whether the knowledge is tacit or explicit-- and an ontological dimension--whether it is known at the individual, group, organization, or inter-organization level. Knowledge is about beliefs and commitment, about action, and about meaning. Western philosophy has the formulation "knowledge is justified, true belief" and has focused on what truth means. The authors view the "justified belief" part as the critical part.
Knowledge conversion (between explicit and tacit) is a crucial part of the social job of sharing knowledge. There are four key modes:
The result of all this is (or can be) a knowledge spiral. It is sustained by using dialog to move from socialization to externalization; by linking explicit knowledge to move from externalization to combination; learning by doing to move from combination to internalization; and field building to move from internalization to socialization. Notice how it moves back and forth between explicit and tacit, and how it can increase its level (individual to group and beyond).
All this doesn't happen accidentally; there are conditions at the organization level that can promote it:
With these conditions in place, the organization can move from tacit knowledge by "sharing tacit knowledge, creating concepts, justifying concepts, building an archetype, and cross-leveling knowledge." (p. 84) This moves to explicit knowledge in the market, which can feed back information and help the cycle continue.
Management and Organizational Structure
Top-down (the traditional hierarchical model) and bottom-up management approaches both have limitations. Top-down presumes the answers from from above, but bottom-up can be too independent. Both models ignore the ability of middle management to reconcile the problems. The middle-up-down approach recognizes that middle managers often create knowledge (the front-line is too busy with today, top management is out of touch). This puts them in a dynamic position, and belies the trend to eliminate middle management.
Middle-up-down acknowledges that top management has dreams, but middle management is the group that makes it happen with a mid-range theory. Middle managers act as catalysts.
A bureaucratic structure is too locked in to repeating its past; task forces or projects are important but need overall structure. The solution is a hypertext organization: combining a business system layer, a project-team layer, and a knowledge base layer. Team members can shift layers, but belong to only one at a time (unlike a matrix approach). Knowledge is combined across layers. Projects tend to be controlled more directly by top management, letting the team focus on short-term needs and speeding up communication.
"These new organizations: (1) tend to be flatter than their hierarchical predecessors; (2) assume a constant dynamic rather than a static structure; (3) support the empowerment of people in building intimacy vis-a-vis customers; (4) emphasize the importance of competencies - unique technologies and skills; and (5) recognize intellect and knowledge as one of the most leverageable assets of a company." (p. 162)
The authors describe three approaches to organizing teams:
This is the traditional waterfall approach. It has a long lead time, and slow learning, and won't be discussed further.
Rugby style involves creating a cross-functional team that works together directly. It handles changes in requirements well, and has a short lead time. Long and continuous interaction among project team members clarifies the grand concept (business strategy), mid-range concept, and the product concept.
Benefits of the rugby style:
Drawbacks of the rugby style:
American Football Style
To balance the benefits and drawbacks, the authors propose "American football" as a new metaphor, suggesting a way to get short lead time and high performance levels.
These are the key ideas:
The authors propose these guidelines for an organization involved in knowledge creation: (page 227)
Finally, their hope for the future is hypertransformation, that is, multiple transformations across multiple dimensions.
Discussion, Relative to Software Development Processes
Nonaka and Takeuchi have focused on an important problem, knowledge creation. It's important in physical manufacturing, and clearly central to software.
Agile teams bring people together to work on problems, in an intensive way. (Collocation, pairing, frequent inspection, and feedback help sustain this intensity.) In working together, the team socializes its values and practices. This lets the team work together efficiently, but makes it harder for outsiders to understand what's going on.
The middle-up-down and hypertext organization approaches propose a mechanism by which middle managers can have an important impact on their teams and the organization as a whole. Too many middle managers spend too much time acting as aggregators of status reports rather than driving and sustaining the creation of organizational knowledge.
The comparison of Rugby (i.e., Scrum) style and American football is especially interesting in the light of efforts to "scale" agile processes. When a job is too big for one team, a small team sets the initial tone and direction, and then rugby-style teams operate in parallel. (Note that this has three parts: big job, cohesive guiding team, and parallel rugby teams.)
Scrum addresses large-project situations with the Scrum of Scrums approach. Grizzly (being developed by Ron Crocker) has practices around team coordination and architecture development (and XP-style practices within teams). XP, to the extent it says anything about scaling, says "conquer and divide": solve the essence in a small team before splitting into parallel teams. All these approaches seem compatible with the football approach.
Other approaches to large teams often seem to miss the essence of the rugby or Scrum approach: they don't value the intense socialization, and may ignore the importance of a cohesive guiding team. In this, they may unwittingly give up opportunities for team learning and performance improvement.
I highly recommend this book: The Knowledge-Creating Company, by Ikujiro Nonaka and Hirotaka Takeuchi.
Copyright 1994-2010, William C. Wake - William.Wake@acm.org