From Robb:

Given a situation where decision making is falling behind the requirements of the environmental reality, we can expect inevitable catastrophic failure at some point in the future. When this occurs, one of the following new approaches will emerge:

C) Decentralized decision making via a market mechanism or open source framework. This approach is similar to process “B” detailed above, except that a much wider degree of diversity of outlook/orientation within the contributing components is allowed/desired. The end result is a decision making process where multiple groups make contributions (new optimizations and models). As these contributions are tested against the environment, we will find that most of these contributions will fail. Those few that work are then widely copied/replicated within components. The biggest problem (opportunity?) with this approach is that its direction is emergent and it is not directed by a human being (the commander).

Robb’s thought of building a market platform leads to an interesting, important and far-reaching question: if one were to attempt to engineer a market platform for decentralized decision making, how would one design the incentive structure? Put another way, what heuristic would one use to determine rewards?

Waste due to the cost of failure becomes a concern when one starts to look at evolutionary approaches where most “contributions” fail. One way to lower the cost of failure is to test new approaches in a simulated environment. Such an approach would have its own challenges, however.

For example, in a discussion in the comments section over at Zenpundit, Moon and I discussed the applicability of genetic optimization algorithms to the problem, as well as potential challenges. Concerns include:

Getting buy-in. This can be difficult due to the probabilistic nature of the process and the potential to just reach a “good” solution as opposed to an optimal solution.

Time. Evolutionary processes require repeated iterations. If each iteration takes too long, the entire process can become impractical.

Stability. In a dynamic fitness landscape, genomes can converge to solutions that are no longer relevant for later environments. In competitive environments, thinking adversaries will be working to make this happen. A competition develops between the rate at which the adversary can change the environment and the rate at which we can adapt to it (i.e. the classic OODA-loop competitive dynamic).

This final point is an irreducible characteristic of warfare, and it distinguishes it from private sector competitive dynamics (as well as many optimization and decision analysis techniques from systems engineering and operations research). Some of my previous posts discuss this further [1] [2] [3].

[1] The No Free Lunch Theorem and Resilience
[2] What about this big thing?
[3] The Limits of Universality (Includes an example of inappropriately applying an industrial optimization model to warfare.)

By way of information aesthetics, found this mash-up visualizing Zimbabwe Election Watch data.

While this sort of heavy-weight functionality is probably beyond most of the cell-phone based browsers many Zimbabweans use to access the internet, it is fascinating for its potential use in rallying and focusing international attention. Like Gapminder, this sort of visualization can provide access the existing data being gathered by NGOs in a vastly more relevant and useful manner.