Navigating Complexity

I was invited to be part of IBM's Alamden Institute Navigating Complexity workshop. I have connected with IBM a number of times at Almaden over the years, and the spectacular site in the hills above San Jose really does expand your mind. It's much more interesting than the Almaden site in Second Life! I do applaud IBM and the leader of the session, Myron Flickner, for pulling this meeting together.

I was surprised at the number of people there that I had met in the past. Stuart Kauffman, of Santa Fe Institute and Bios Group, Doug Engelbart, inventor of the mouse. William Rouse - organizational modeling expert. Brian Arthur - synthetic economist. Also made lots of new conenctions.

Surprising too that even the definitions of complex systems are still in debate. There is also the subspecialty: Complex Adaptive Systems, which assumes the models are adapting. Not unusual, though the time scales can be very different. William Rouse presented a paper developed for presentation to the NSF which tries to button down definitions and research directions.

The presentations could be divided roughly into two groups. One about the architecture and function of the brain, and the other about business-related topics. Certainly the brain is a complex system. Complex systems also exist in business. Supply chains, organizations, software, communications and innovation flows.

The work on brain systems is intellectually very interesting and I found it fascinating, but it really should be handled in a separate meeting. I could not have brought my business clients into that meeting. Separately, with the right kind of specialists, they could be more useful.

Almost everyone in the group was either in academia and research. The exceptions were me and Ronald Johnson from Boeing. I challenged the group that we should look much more closely about how we can translate complexity systems to make them more useful for business applications. This should include improved architectures and methodologies that include experimentation frameworks.

One question that was asked was what kind of problems have truly been solved using this technology. There are some examples, but far fewer that there should be. Several examples where shown, notably some work at CMU on how to analyze organizations, that looks very promising. I will provide more details in a forthcoming post.

What ever model your build needs to be trustable. Why else would management use it to support decisions? Complexity models may not be verifyable in the normal sense, but for their results to be trusted, there needs to be a method to overlay an experimental design that can gain that trust.