Tuesday, 22 December 2009

Ecology of war: statistical patterns of insurgency

I haven't posted on this blog for a while, but this paper inspired me to. These guys have managed to make a pretty good stab at understanding terrorist motives by simply analysing the casualty data from various wars. Here's the blurb:

Many seemingly random or chaotic human activities have been found to exhibit universal statistical patterns. Among these is human conflict: the size distribution of casualties aggregated over entire wars follows an approximate power-law distribution. But do the events within individual wars share any common patterns? Neil Johnson and colleagues show that they do. Using detailed data sets from a wide range of conflicts, including Afghanistan, Iraq and Colombia, they show that insurgent wars share common patterns with each other, and also with global terrorism. They explain the size and timing of violent events in terms of ecological interactions between human groups. Their model is consistent with recent hypotheses concerning insurgency and establishes a quantitative connection between insurgent warfare, terrorism and ecology. Similarities to financial market models point to a link between violent and non-violent human behaviour.

1. This immediately springs to mind despite being quite blurry. In the Dynamical Systems and Chaos course at uni, one of the problem sheets was to work out the phase diagrams for competing systems, like whether malaria would die out under certain conditions, how rumours are spread amongst physicists, and who wins the hypothetic war etc, which are described by sets of PDEs. But the problem of why these are never true representation of reality is the inadequacy of the many constants used to depict reality. So if we can fine tune these constants, in theory we could increase the accuracy which is what demonstrated here. I may be completely wrong. Frankly don't remember much from the undergrad days.

2. "As far as the laws of mathematics refer to reality, they are not certain; and as far as they are certain, they do not refer to reality."

Don't overestimate the utility of these modelling exercises.

3. Fortunately we can understand the deviations from reality with error analysis.

Besides, what do you suggest as the alternative?

4. Use modelling but understand the limitations.

5. Ed, isn't that what Marc has said? Or what science is about?

6. More on limitations here.