Its been a long time coming but I'm getting to the next stage of the Knowledge Sharing Across Silos series where I will start to move from analysis to the problem to solutions over a series of posts. However I found when I started writing the first solution post that there was still a story to be told, namely the origin of the Cynefin Framework (or at least its early stages). It all started as a means to understand how informal networks and supporting technologies allow grater connectivity and more rapid association of unexpected ideas and capabilities than formal systems.
A long time ago now I read Boisot's Knowledge Assets for the first time and thanks to the agency of Yasmin Merali met the author and started to work with him; I'll be teaching with him in Hong Kong next week as it happens. The I-Space model shown is a cube based on three axes: abstraction, codification and diffusion. The social learning cycle (red on the picture) shows how as knowledge is increasingly moved from concrete experiential Zen type knowledge to codified highly abstract (expert language etc) it is increasingly easy for it to defuse independently of the knowledge holder. Once internalised it moves back to the concrete. Now that is very brief, you really need to read the book.
Initially in a workshop at Warwick University and then in a series of articles I started to take some of the ideas in the I-Space, added much, modified much and ended up with the Cynefin model. That was the first time I have taught with Max, there have been many times since. This was at the height of the knowledge management movement, then dominated by the SECI model and a focus on codification. My first move was to modify the I-Space to create a different perspective on what would become one axis of Cynefin. I took the abstraction dimension, but looked at that in relationship to the cost of codification.
The result is shown here, and more fully elaborated in this article. Its fairly simple; at the highest level of abstraction (you have a conversation with yourself) the cost of codification is very low as you have a 100% shared context (hopefully). On the other hand if you want to share your knowledge with everyone then the cost of codification will approach infinity. Basically the less the shared context the higher the cost in money, time and effort of creating a knowledge artifact of artifacts which will successful allow knowledge to diffuse without the direct mediation of the knowledge holder. The net result is that is a very narrow range in which it makes any sense to codify at all, which I call the zone of effective diffusion (I used to call it the zone of acceptable abstraction but that is not as accurate and its too wordy to boot).
Now any effective diffusion is dependent on shared context, but it also varies on the degree to which the knowledge context is dependent on cultural aspects. Highly explicit cultures find it easier to codify what they know – think of engineering where there is a body of knowledge, an established education and training programme and a body of published material. On the other hand much knowledge is informal, is deeply dependent for understanding on common shared experiences and deep trust built over time.
Given these two elements I created a model with a vertical dimension based on the balance between low and restricted levels of abstraction and the horizontal one flexing between teaching and learning cultures. At various times I also used explicit and tacit, rule based and ideation based and other language. It isn't the Cynefin model, I was only just starting to study complexity and was proceeding with care. Far too may people read up on something quickly and use the language without real real understanding. You see a lot of that with complexity theory and neuro-science, and there are now some cases with Cynefin itself. Either way its not somewhere I wanted (or want ) to go. You can see the evolutionary path that led from this model to Cynefin but Cynefin only really arrives when complex adaptive systems theory is used for the first time. The model allowed me to look at four types of community or knowledge sharing context. as follows:
Low abstraction – Teaching
Here we are dealing with material that has to be known and understood across the whole organisation without human intervention. Its very basic stuff like expenses and the like. This domain is never an issue for cross silo sharing. Well that is not strictly true, its where all the problems are as far too many organisations attempt to reduce all their sharing activity into the highly structured forms of this domain and that is where things go baldy wrong.
Restricted abstraction – Teaching
The domain of the expert. Context is provided by professional education and formalised training. The mechanisms for communication are well established – papers, report logs and the like. Membership of the community is by dint of training and/or function and is formal rather than informal. Within the bounds of established (or possibly establishment) thinking and language transfer is pretty effective. However in an inter-disciplinary environment, or in a field where novel ideas that challenge the establishment are emerging this is not a good space.
Restricted abstraction – Learning
Its important to understand that the abstraction level here is much more orientated to common experience than it is to specialist language. To communicate in these environments you have to live the life, share the experience, intuitively understand the values. This is the domain of the shadow or informal networks on which all organisations depend. It's also the area of serendipitous encounter. To take a classic case, two employees of the same organisation attending a conference have a conversation about an embarrassing side effect of a drug, result Viagra and we can add the glue that didn't work which gave rise to PostIt™ notes and many others. Innovation, linking across silos is far easier in an informal network than in a formal system.
Low abstraction – Learning
Here we are dealing with novel and the unexpected. The abstraction level is open due to that novelty. No specialist language has yet evolved and there is little or no experience in any part of the communication. We have to develop practice, allow language to emerge through our interaction with reality. It may be uncharted but we can still navigate it if we understand some of the principles of how to allow new ideas and concepts to interaction and co-evolve with real world problems. I didn't know it at the time, but this was one of the spaces where social computing would provide much utility but also at least in part fail on its promise. The space is too unstructured, two anarchistic for meaning to emerge. I hadn't fully understood the utility of partial constraints, and that is a subject I am still exploring.
It was early days but the ideas were forming. I am OK with the abovemodel, although I think I misused symbolic and I wouldn't fall back to it. That version went on to be a part of a book chapter before complexity theory crept in and the early forms of what is now Cynefin started to emerge. More of that in a future post