thoughts from the intersection of science, technology and politics

Insurgencies are by their very nature difficult to understand. However each time an attack is launched and every time an IED explodes we start to know a little more about the structure of an insurgency. If we combine together enough of these attacks we start to build up a mosaic picture of the insurgency. Their actions can start to be defined mathematically and we can work backwards from these signatures to understand the fundamental forces that underlie the insurgency. This is exactly what we did in our latest research study “Common Ecology Quantifies Human Insurgency“.

With these models we can for the first time quantitatively understand more about what makes an insurgency successful. From our analysis and modeling we find that there are 14 key characteristics that define a successful insurgent ecosystem; these are listed below with a short name to describe the feature.

1. Many body: There are many more autonomous insurgent groups operating within conflicts than we had previously thought. For example there are 100+ autonomous groups operating in Iraq (as of 2006).
2. Fluidity: The insurgents are loosely grouped together to form fluid networks with short half-lives. This is very different from the rigid hierarchical networks that have been proposed for insurgent groups.
3. Redundancy: If we remove the strongest group from the system another group will rise to replace the previous strongest group
4. Splinter: When a group is broken it does not generally split in half but instead shatters into multiple pieces
5. Redistribute: When a group is broken the components are redistributed amongst the other groups in the system. The redistribution is biased towards the most successful remaining groups.
6. Snowball: The strongest groups grow fastest
7. Tall poppy: The strongest groups are the predominant targets for opposition forces
8. Internal competition: There is direct competition amongst insurgent groups for both resources and media exposure. They are competing with each other in addition to fighting the stronger counterinsurgent forces.
9. Independent co-ordination: Autonomous groups act in a coordinated fashion as a result of the competition that exists between them.
10. Emergent structure: Attacks in both Iraq and Colombia become ‘less random’ and more coordinated over time
11. Evolution: The strategies employed by the groups evolve over time where successful groups/strategies survive and unsuccessful strategies/groups are replaced.
12. High dimensional: Connection occurs over high dimensions (i.e. Internet, cell phone etc) and is not dominated by geographic connections.
13. Non-linear: It is approximately 316* times harder to kill 100 people in an attack than it is to kill 10 people. (*Results for a conflict with alpha=2.5).
14. Independent clones: the fundamental structure and dynamics of insurgent groups is largely independent of religious, political, ideological or geographic differences.

What can we learn from insurgents? Should the US military adopt more of these principles? Can we apply these organizational characteristics to other problems? You can read more about the research over at the TED blog, including the in depth interview I did with them.

{ 5 comments }

I have been working over the last few days with my team to put together the new website mathematicsofwar.com. This website contains a set of resources to help people better understand the details of our research. It is more science/math focused site and includes copies of our latest research papers, background reading, working drafts and details of the modeling and analysis. We’ll be adding to this siteover the coming weeks and months and building it up as a resource for the quant conflict analysis space.

It’s quite a lot of literature to get through and covers a range of different topics including economics, statistics, physics and computer science. But then that’s what the holidays are for right :)

{ 0 comments }

It looks like our research just got picked up on Slashdot. How long before quant analysis of conflict is also on the daily agenda of policy/strategy/military types.

To explain what was driving this common pattern, the researchers created a mathematical model which assumes that insurgent groups form and fragment when they sense danger, and strike in well-timed bursts to maximize their media exposure. Johnson is now working to predict how the insurgency in Afghanistan might respond to the influx of foreign troops recently announced by US President Barack Obama

Hello to any new visitors here via Slashdot.

{ 0 comments }

A good post from John Robb over at Global Guerrillas about the “Ecology of war” Nature paper. Robb argues that the research put forward in the study has applications to what he calls Open Source Warfare. Robb believes that Open Source Warfare has become the dominant form of conflict for the 21st Century and it is defined as

a model of warfare that describes how many small autonomous groups can fight an insurgency despite the lack of a central command hierarchy.

This is an interesting theory, however up until now there has been little in the way of quantitative evidence to explain how insurgent forces could operate in such a decentralized fashion. Robb goes on to say that our model provides quantitative evidence for the mechanisms that would allow Open Source Warfare to successfully operate. These are;

A grouping mechanism. Why groups fragment and form. It assumes a constant insurgent population with a fluctuating number of groups depending on counter-insurgency pressure (more pressure = more groups).

A timing mechanism. A description of how insurgent decision making cycles (assuming lots of groups) impact the timing of attacks. The conclusion is that of cross group communication through the media (stigmergic learning) explains how this mechanism works.

These ground level mechanisms or rules of interaction provide a global structure to an insurgency. This structure is revealed by the nature (size, timing, location) of the attacks. What is interesting about the model is that these structures emerge without a centralized control hierarchy. As a final note Robb has this to say about our research paper

As a scientific study of the area, it does a great job.

Update: As Robb notes, in the Nature paper we use a constant population N for the insurgent forces. However, although it is easier to solve a model with a constant population – the structures and mathematical signatures produced by the model do not change if you allow the insurgent force N to increase/decrease over time. I will follow up tomorrow with a post about non-stationary insurgent populations and other generalizations to the model.

{ 0 comments }

One of the important things to understand about our Nature paper on the ‘Ecology of war’ is that the results in this study go beyond first order statistical analysis and power-law distributions. In this paper we look at the deviations from power-law and we analyse the temporal distribution of attacks. This in itself represents a significant advance in the field of quant analysis of conflict.

But perhaps more important still is that we are able to create a robust model to explain the emergence of these statistical patterns. And this model is applicable across a wide range of different conflicts including modern terrorist activity. With this model we can finally start to understand what it is that makes an insurgent ecosystem successful. Going even further, we can run different strategic scenarios on this model to understand how the insurgency is likely to respond to specific policy changes i.e. increasing troop numbers in Afghanistan. We are simply not able to gain this type of understanding with simple statistical analysis – the modeling is key.

Mikael Huss captures this realization nicely in his latest piece

An interesting article about analyzing and modeling data from wars and conflicts was published in Nature yesterday. On first glance, it looked like one of those “look, we found another power law” papers, but after I had read this interview with one of the authors, I changed my mind – it’s really quite interesting.

Huss also notes that the results are independent of the specific source data used for the analysis i.e. the results are not just a signature of the way mass media reports violence.

Even this data-collection effort is interesting in itself. According to Sean Gourley…the statistical pattern remains similar no matter what data source (governmental/academic/mass media) you are looking at.

{ 0 comments }