From prediction to adaptation

I've been attending the NECSI summer school on complex systems. One of the key concepts I've gained from it is the reason of why traditonal methods (such as calculus) are not useful when dealing with complexity. The problem lies in the "averaging assumption", which means that the state of each component is independent of others. This is not true in many systems, so the central limit theorem does not hold.

Considering also deterministic chaos, one realizes the limits of prediction in complex systems. It is not that we cannot know them. The thing is that new information is generated by the interactions, so we cannot predict (compress) the future of the system without "running" it.

As human societies are becoming increasingly complex, it seems to me that there is a shift between the usefulness of prediction and adaptation. Now, both are useful, but when we can predict less and less what will be the state of the world economy next year, or even next week, we need to be able to adapt more than to predict (anticipate).
Another component is also relevant: robustness. If a system does not "break down", i.e. maintains its functionality, it will have greater opportunities to adapt to a changing environment.

A combination of adaptation, anticipation, and robustness is necessary to cope with complexity (see Sections 2.6 and 3.5.3 of my thesis). All of them are important. The only thing I'm saying now is that the relevance of adaptation compared to anticipation increases with the complexity and predictability of the system.

Only a generation ago, it was common to go out of college and get a job, the same until retirement. You would live and die in the same city that you were born. Prediction was sufficient up to a few years, so one could make a detailed plan for the future. Not that it always worked, but now things are different. In a globalized society, very few people remain in the same job for more than a few years (or even living in the same city, or with the same partner). Predictability has decreased to the order of a few months (see the post "Overwhelmed by Choice?"). Lacking predictability, we are forced to increase our adaptability. This is a shock to our value system, because it is difficult to "settle down". Our values limit our adaptability. Will this bound the complexity increase of our society, or our values will change?


RBH said…
My firm evolves artificial agents to predict the behavior of a complex adaptive human socials system and control processes in it, and have found that the 'balance' between adaptation and persistence to be a significant problem. Briefly, while we want our artificial agents to be able to adapt to changing dynamics, we don't want them to merely react to last week's noise. The latter is a recipe for whipsawing. We address that problem with a population of varied agents rather than one or a few optimizing agents. We're still working on it, though. :)
Carlos said…

Indeed, diversity is very useful for exploring and coping with complexity.

Best regards,

Popular posts from this blog

The Past, Present and Future of Cybernetics and Systems Research

flocking memes...

PhD thesis: "final" version online