Two interesting articles (well more than that actually, but two in particular) in the current edition of the New Scientist (no. 2802 5th March 2011). One is a great example of fractals which shows that elliptical galaxies can grow in the same way of ice crystals build up around a microscopic dust grain as it falls to earth. This provides an alternative to the previous theory that they happen when two smaller spiral-shaped galaxies collide. Similar patterns between systems which are radically different in scale.
Then we get a slightly more mixed article entitled How to predict when a dictatorship is ready to fall. This has some wise words from Martin Scheffer who argues (as many of us have over the years) that its a mistake for analysts to focus on a causal trigger rather than the underlying system To quote the quote We cannot predict the spark, but we can say when a forest has accumulated dangerous levels of kindling. In the same article we have a claim from Yaneer Bar-Yam that he has successfully modeled 90% of recent ethic conflicts in Kenya, central Asia and former Yugoslavia. Great for getting a headline, but I would be more impressed if it had been used in advance. Well it may have been but then it is surprising to see it go unreported.
Now I don’t challenge that the claim of 90% is the case, but I have seen many a similar report in many a conference and they are all retrospective models. It’s easy in retrospect to see the things in a model that provide coherence and it can be useful. However even if the model has predictive capabilities it will have the issue of false positives and also the question of how to convince someone to action. How (to use the possibility opened up in the article) to get a repressive regime to realise it needs to change, before change is catastrophically thrust upon them. I should also admit in my one encounter with Bar-Yam that I thought he placed an over reliance on models and failed to take account of the unique aspects of human complex systems namely identity, intention and intelligence. I said so at the time but I don’t think listening was on the agenda, but we all commit that sin from time to time.
Interestingly the article qualifies the value of models with the following critical comment though good social data may be hard to find. Now this relates to my post of yesterday. Agent based models, however sophisticated, are limited by the fact that humans do not have discrete agency (changing and shifting identities), they do not operate from simple rules but make decisions based on complex patterns of both the past, present and possible futures. Some of those decisions are intentional, which is not the same thing as predictable. Think of altruism, archetypal role models etc. etc. Some exhibit intelligence and design some are reactive and they can move between those modes very quickly.
So all of the models are a very useful method, but they are not the whole picture. Firstly you need the data. SenseMaker™ was in part designed to do this, and to do it in real time from multiple perspectives. By gathering the micro-narratives of day to day existence, by allowing those micro-narratives to be self-interpreted at the point of origin we start to approach a situation where the system creates its own modeling data. We need more synthesis and experimentation here, but the realisation that we need to shift from anticipation based on prediction to anticipation based on awareness is growing and I no long feel I am banging the drum in isolation as I did on those first DARPA programs over a decade ago.
Galaxies and snow flakes operate from basic laws, humans while they are physically limited are not limited in their capcity for myth, metaphor and magic, it is both our damnation and potentially our salvation