I was talking to an associate a little while ago who has links to groups active in the Middle East. The conversation went this way and that and I started thinking as to how you could resolve the situation.
Historically decades of violence tend to lead to power vacuums that then result in an unexpected alteration of power structures. The one that sprang to mind was the Heraclian-Sassanid wars of the early 7th century that left the Middle East highly exposed to the early waves of Islamic conquest. Arguably you could argue that the exhaustion after the Wars of the Roses in England or the wars between the Protestant Union and Catholic League in Germany in the 17th century saw equally significant long term changes as a result. The Tudor’s and Prussia both started on their paths to ascendancy at this point.
So the risk on long term violence in the Middle East is less the threat of Al-Qaeda or IS, or any of the other parties currently involved, as much as the unexpected power structures that arise as a result. Current strategy by the West seems to focus on removing leaders in the belief that they can whack faster than the moles can burrow.
This is broadly similar to the approaches that the Crusaders and the Seljuks applied to rooting out the Nizaris in the 12th and 13th centuries. Caliphs and commanders continued to be assassinated. The threat was only removed when the Mongols, under Kitbuqa, carried out a policy of extermination, genocide really, in the Elburz mountains.
That’s not really on the cards today despite the British and Russian armies having similar policies in Afghanistan in the 1880’s and 1970’s.
Whack-a-mole may work but it seems that we have the possibility to start using big data and analytics to get more predictive confidence in our analysis of warfare in Syria.
One argument in the social sciences is that predictions is difficult as you only have one data set. No other country is like Israel, no city is like New York, no time is like Birmingham in the 1970’s
Another argument is that we have little insight into the organisational politics and behaviour of the 2000+ groups involved.
That may be true but I am sure that there are many hundreds of analysts attempting it for the security services of many countries.
This approach is different.
We start modelling the historical data. We have a huge number of cases where fragmented armed groups have struggled for control of a country, driven by ideology or not. Many people, Toynbee among them, have tried to build theories of history, to find patterns from historical human behaviour. That has failed for many reasons, most importantly because historical analysis is bluntly subjective, massively simplifies and is prone to bias. The models that you can build based on the primary and secondary unstructured sources that make up most of the historical record are weak.
However this data can be structured.
How can we structure historical data to allow us to build predictive models?
At this point we need to step back a bit and think about what it is that we would want to structure and why? All the groups in the Middle East (and any militant groups) are looking for power. We can look at people like Greene, Machiavelli, Sun Tzu who are concerned with the nexus of war and power. They give some insight but to a large extent (Sun Tzu excepted) they are more concerned with power being part of the agency of states. and wars being, on the whole, bi-partite.
On the other hand if we look at the academic literature we have Lukes, Foucault, Giddens and others – but they are almost entirely focused on power as a function of the nation state and it’s internal and external manifestations (marx after all was focused on the analysis of the power struggle between capitalists and workers).
Finally we have the studies of power within organisations – Handy, Hofstede et al – but modern business organisations are more constrained in the use and abuse of power that Al-Nusra or the PKK. Studies are also highly focused on achieving shareholder value rather than the diverse goals of warring groups.
If you consider a countryside in arms with multiple warring groups, not quite a state of nature, but where status and success is based on being a good war leader, in a tribal sense, then we have very little modern political or organisational analysis that helps us to understand this – in part because of the humanist consensus that we have moved on since the wars of the Renaissance and Reformation which was the last time Europeans faced multitudes of non-state actors fighting.
That humanist movement though is key, especially Gutenberg and the proliferation of printing presses that he released.

By the early 16th century hundreds of books a year were being published.
Unlike today the majority of them were non-fiction and there were also large numbers of pamphlets and broadsheets reporting the news.
The result is that the depth of historical knowledge that we have about the people involved in events grows hugely from a few centuries before. The cast of historical characters grows hugely even in places such as the Balkans and the Germanies where continual warfare was taking place. Importantly there is also far more information about who they were and how they died. A good example is the Munster uprising in 1533 where a number of non-entities, merchants and religious fanatics seized the city and held out for 2 years. We have exceptionally detailed information about those individuals who would never have featured in Geoffrey on Monmouth’s writings 4 centuries before.
So the theory is that we can map historical multi sided conflicts through the records of individuals being engaged in them. Mentions in contemporary manuscripts and books is an indication of the importance of and success of figures. In many cases we know the causes of death. John of Leiden was executed. Martin Luther died of a stroke. John of Leiden’s movement died with him. Luther’s didn’t.
By mapping these networks we can then get a much deeper understanding of the historical forces in play. This though only tells us about individual conflicts and doesn’t help us generate useful rules predicting future conflicts.
Before I move onto prediction I’ll dive into how this can be done. Google and many other institutions (such as the British Library) are busy digitizing their book and manuscript collection. OCR is sufficiently advanced that significant information can be extracted on a programmatic basis. So we can identify actors by location and time period. With some semantic analysis we can identify whose side (if any) they were on, characteristics about them (the proud, the cruel, the fat) and their eventual fate. And we can do their for hundreds of thousands of books.
There are certainly some pretty huge data quality issues – language, naming and spelling conventions for one – duplication and information gaps for another, but we can build very large scale data sets that capture the vast majority of those actors who contemporaries thought were important.
Once the data set is built, or partially populated, perhaps seeding with something like Wikipedia or Encyclopedia Britannica, (which would allow an easy low cost Minimum Viable Product) then you produce the network map. Wikipedia already has links which enable tagging but at a deeper level relationships would need to be assigned on a probabilistic basis – using strong/weak ties as popularised by Granovetter.
The model would also be time sensitive so that the relationship and engagement between individuals could also be related to the exercise of power on the ground. This would require additional modelling. how do you quantify power? Should a historical version of Jane’s be developed, or a economic or a territory based model? How accurate would it be? To start with territorial control would probably be sufficient as in many cases that was the measure of success at the time.
So that is a model of a particular conflict (or group of conflicts) in a particular time period. The next step is that you repeat the process for multiple conflicts. Then you start feeding them into an AI
There are a bunch of different ways of doing this. One way is to emulate the AlphaGo team which had a three stage process. First they trained the AI on the data using a deep neural network. Then they used reinforcement learning and had different versions f the AI play off against each other. Finally they used all these results to feed into another neural network that was focused on predicting consequences.
If you look at this way a go board is quite similar to a network of relationships. Certainly there are more positions to play with and more states than the 3 state go board (black, white, empty) but in essence what we have is a game, fact recognised by everyone from the ancient Shahs of Persia with their fascination for Chess (whether they invented it or not), to Sun Tzu and countless European strategists.
I am under no illusions that this would be easy, but once you strip away all the parts you have
- A lot of data
- A game / conflict where rules are consistent over time
- The computing power to understand it
A lot of the predictive power of any programme would really depend on the model. The more complex the model the scarcer the data (or more expensive to collate it), the simpler the model the more nuance that is lost. It’s a classic case of Non sunt multiplicanda entia sine necessitate or use the simplest model that gives you useful results. We don’t know at what point that is.
I remember a long term ago using a VAX mainframe to try and model the impact of the British Army using a 120mm or 105mm barrel on some equipment. The problem was that the model was so complex that we never got the simulation to give us useful results. It’s the same here. We have centuries of unstructured data. How little can we get away with structuring, and to what level and would we get any useful results?
If it worked it would then be possible to start considering, to finally bring drone strikes in, who to kill and when to have the most powerful long term impact on the contact. Cut off one head and two more appear. Continual warfare hydrates the enemy, teaching them how to counter your moves.
In contrast knowing how changes will impact the enemy gives you mastery of the situation and I think deep knowledge of how organisations change in war conditions will be of great value
“Be extremely subtle even to the point of formlessness. Be extremely mysterious even to the point of soundlessness. Thereby you can be the director of the opponent’s fate.” Sun Tzu
Sorry it’s been a bit of a ramble today. Next time there will be a bit more structure :).

