Modeling Chaos

Snowden's Cynefin framework says that modeling is primarily of value in the Complicated/Knowable domain, and that the Unordered domains (Complex & Chaotic) are not effectively modeled.

While I think the basic point is right, a superficial reading of Cynefin might overlook the fact that most decision contexts have a mixture of simple/known, complicated/knowable, complex, and chaotic elements. And, Snowden has a lot to say about tactics for moving unordered aspects of a context into an ordered domain to enable better decision-making.

This was prompted by an article in Network World about a new capability that IBM is touting for managing a crisis (which usually has a large chaotic component). The basic approach is stochastic programming.

My initial reaction was negative...based on perspectives like Cynefin, but also in reaction to statements like "The model allows all unforseen challenges to be solved, mostly within an hour..." I suspect even the most ardent proponent of stochastic programming would find this statement a bit extreme.

I suppose Moore's Law makes stochastic programming increasingly attractive...and, as with all IT-centric tools, it may, in a relatively short time, become relevant to a wide range of needs. However, as with all "better mouse-traps", there are significant switching/transition costs/investments associated with changing existing processes, policies, and (most of all) the world inside the heads of those affected by the change.