Predicting the politics of problems

The world today seems filled with conflicts, civil wars, mass demonstrations and terrorism. Around the globe dictatorships impose their rules, countries cut off trade with others, missiles are fired over borders and media outlets are restricted. Geopolitics (the high level politics of countries and how they interact), is a complex and serious business. There are many who would like to be able to predict, to foresee, conflicts before they happen so they can take appropriate steps, and there are a number of computer science researchers around the world working to develop such systems.

Modelling the rules

What needs to be done first is to create a model. A model here refers to a mathematical description of a situation rather than the occupant of a catwalk. Mathematical models are used all over the place, for example in the financial markets to predict stock prices, or in architecture to predict how a stadium terrace will stand up to thousands of fans jumping around. What a mathematical model needs to start is some data, some information to work with, and some rules on how that data goes on to produce a particular effect. Take a very simple financial example. You may know that when Company X produces a new product, its competitor Company Y first loses sales, but then comes up with a new product. So from this rule, and the data 'Company X launches new games console', you can predict that probably Company Y shares will fall for a few months, but they will be working on a new product to counter Company X, and that in a few months shares in Y will recover from the downturn.

Finding the X factors

What's important is to decide what the main factors are that your model has to deal with. In any situation there could be thousands of factors, and you can't work with them all. You need to do something called abstraction: pull out the important bits for the complicated big picture. So when we look at predicting conflicts related to politics the sorts of important data for the model would be in automatically looking at key words in leaders speeches, previous form (has the country had problems before, and if so with who), what are the social conditions like, amount spent on healthcare and education, elections or no elections approaching, and so on. This data collection can be automated by examining news feeds, online news papers, official websites and blogs from around the world.

Linking actors and their actions

From this data software would create actions for actors. Actors are the technical name for parts of computer software that represent individuals or concepts capable of actions. For example a dictator (represented as an actor) would have actions determined by the data coming in, for example were they making peaceful or warlike speeches?

Actors in computer programs don't need to be people, though. Another Actor might be an election: have they been delayed, have they been peaceful in the past, and so on. With this data the appropriate actions can then be attached to each election Actor. The system would then use its rules to determine if the action of all the Actors combined was indicating trouble was brewing, and generate an alert.

Perfect predications?

This sort of software has been successfully developed and has given some useful results, with probabilities attached to the consequences of particular Actor's actions allowing the complexity and uncertainty of the real world to be included. Of course models can't be one hundred percent accurate. They can't model all the factors. They just select the best Actors, and try to find the best data to work with. However as computers become more powerful, more detail can be added, and with the better rules and more accurate data, predications will become better. Will this make the world a safer place? Well that remains an unanswered question. Perhaps we should try and model it.