Exploring “A work in progress”

Let’s start by seeing if I can get in trouble…

I’ve spent some time thinking about David Snowden’s “A work in progress” model for the complex domain. It took me awhile to get my head around it… I’m not convinced that I haven’t actually missed some significant bits of it.  In fact when David (and Steve Holt) start talking about a 3D dimension, I am sure I am missing something.

As a visual person after I pondered the graph for a little while I decided to redraw the graph, so that I could see it “my way”, I think I may have gained a couple insights for myself on the way (your milage may vary).

In this post I’ll do my best to explain what I “see”.

Coherence, Convergence and Coalescence

The “WIP” model refines Cynefin’s domain of complexity by adding 3 dimensions, 4 “danger areas” which are contrasted against a “valid range”.

David Snowden’s “WIP” Model – Copyright © 2007 Cognitive Edge

Here is where my confusion started, the post containing the “A work in progress” model (referred to as the “WIP model” in the remainder of this post) defines Coherence and Convergence but skips Coalescence. I vaguely remembered the terms being defined before and sure enough they are, on the previous post. I’ve reposted David’s definitions here, they are critical to understanding the WIP model.

  • Coherence: the degree to which any need or requirement is structured/defined/understood
  • Coalescence: the level of fragmentation of the requirement and connectivity between fragments.
  • Convergence: the degree to which different interest groups agree on the needs and nature of what is needed

The Cynefin Framework’s dissection of the difference between complexity and complicated was rewarding, similarly this model enriches my understanding by illuminating the subtle difference between terms that at first blush seem so close as to be colloquially synonymous.

Until seeing this model, Coherence has been, for me, the main measure of “valid” structure in narrative. Here the model yields it’s first insight for me, “full” coherence by itself is not only not “enough”, it is outside of the “valid range”. We need to have an additional measure of validity on our pursuit of actionable knowledge, Convergence. Convergence by itself, leads to Pattern Entrainment, where we agree on the way to do things without understanding the structure of the problems we are solving. An additional final validity measure is added by coalescence. Here information becomes more interconnected and pieces start to “fit together” defragementing into a whole.

Valid knowledge in this model emerges as teams bring fragmented “pieces” of knowledge, from disprit viewpoints, together and agree on ways of making that knowledge actionable.

As I examined the model, I realized that “to complicated” hinted that the “Ordered” area in this graph contains both Complicated and Simple areas of the Cynefin model. Which makes sense and is probably obvious to Cynefinites but maybe hidden from the uninitiated. The Cynefin model is often broken down into two domains; Ordered (Cynefin; Complicated and Simple (Labeled ORDERED in the top right here)) and Un-Ordered (Cynefin; Complex and Chaotic, (Both labeled in the mid to lower left here)).

Continuing to carefully review the diagram, in my case by recreating it, left me with some questions.

“complexity and its three boundaries (to chaos, to complicated and to disorder)”

I see two clearly defined boundaries… to chaos (lower left, towards LOW Convergence, Coalescence, and Coherence) and to complicated (upper right, moving FROM High Convergence, Coalescence, and Coherence). I’m unsure of where Disorder lies on this diagram?

Jabe’s Inverted WIP Model


Any intelligent fool can make things bigger and more complex… It takes a touch of genius – and a lot of courage to move in the opposite direction.   

-Albert Einstein

Jabe’s Inverted WIP model

So… here is my attempt to make some sense of David’s WIP model for myself. As Einstein notes above, I ain’t exactly a genius for making it more complex, but I hope to elicit some conversation that will clarify the model further.

The first major difference between David’s WIP Model and mine… I’ve placed Lows (and added ∞) at the extremes of the graph and HIGHs (with a limit of 1 not 0) at the origin. This shifts the perspective of the graph, as constraints on the system increase, a (theoretically) perfectly Coherent, Convergent, Coalesced problem moves towards a singularity. As constraints (as well as knowledge and connections) are removed from the system the possible answer space broadens and expands. This feels more like my experience of problem solving, where Order moves towards a limit but Disorder is nearly infinite. In the past I’ve thought of this movement of information from Chaotic through Complex, toward an Ordered state as a “Cone of Certainty.” The trade off here is that as we move towards Order the decisions we have made constrain our system more and more, forcing us toward potentially suboptimal solutions. Complexity theorist will recognize this as a form of bifurcation, where previous decisions alter the possible solution space.

I’ve added two “danger areas”:

  • CE: Cognitive Ease; Lower Left: As Coalescence, Coherence and Convergence move towards 1, teams risk the chance of believing they understand the problem so completely they don’t need to think about it anymore. This is the realm of oversimplification and myth. Concepts that make their way here can be VERY DIFFICULT to dislodge. With a complete lack of conflict, teams will all agree that they are talking about the same thing, they will claim they all “understand” it and will all agree there is a clear process to solve the problem. Delegation will work well, until the context of the problem changes, leaving teams hurtling towards chaos with little understanding of the “why” of the original solution. This is the domain of “The Bananananananana Principle”. This danger area is in some contention with David’s Pattern Entrainment. I’m somewhat confident that my “danger area” is worth differentiating. 
  • UM: Unmodeled; Upper Right: Moving towards disorder we find ourselves beyond the realm of probability, where teams have no language or models to begin to describe the problems they are attempting to solve. Lacking models to describe the problems, teams maybe either, unable to clearly identify a route towards order OR be completely unaware that parts of their systems are in a state of disorder. This is the domain of being blindsided. This is the domain of Zombie computers, and PEBKAC, where experienced users have difficulty helping inexperienced users, due to a complete lack of a reasonable shared model. 

I’ve labelled the two borders that (I think) are shared by David’s and my inverted model. I’ve indicated with arrows the direction of movement across domains. Again it is important to note that the WIP Model’s ORDERED area contains both of the Complicated and Simple domains of the Cynefin model. In the top right I have added a border “to disorder”, I am pretty unsure… does it belong there? Is the shape correct?

Finally I’ve added a new “valid area”, in yellow just beyond the border of “to chaotic.” This area is valid in a different way than the green area. Probability is valid in this area. We have effective models for describing and examining the performance of a system in this area. Given a large enough amount of input statistical models can be very effective in this area, in order to probe the chaos and “find” interesting patterns. This area is the area of Experimental Mathematics and “The Lean Startup.” Economies change here… many small measurable safe-fail probes become significantly more cost efficient than attempt to bring order to the domain before taking action. Action can be taken quickly with expectation of high rates of failure. Team’s focus on recoverability instead of continuity and stability.

As usual, creating this model and writing this post has given me more clarity around my thoughts… I look forward to hearing yours.