Can Artificial Intelligence predict armed conflicts?

It is not easy to guess when a new conflict will erupt. Sometimes the causes can be as varied as a birth boom, a bad harvest or, if we remember our history lessons, the murder of an archduke …

The problem
Despite the difficulty of the effort, there have been interesting initiatives that try to determine which countries are most at risk of deliberate actions by armed groups that cause civilian casualties.

They are usually based on the analysis of different risk factors such as: a previous history of similar events, restrictions on freedom of movement, coexistence of different ethnic groups and type of political regime. The models created convert these risk factors into input variables, such as:

  • Population density. – GDP growth. – Travel time to the nearest city. – The proportion of uncultivated land. – The years since independence. – The type of government.

From these input variables, the models generate a risk or score value.

Some models
For example, in 2011 the Peace Research Institute in Oslo created a model to predict conflicts between 2010 and 2050. However, it was not able to predict the civil war in Syria.

Therefore, it is usual to combine different models, maximizing their strengths and minimizing their limitations. A scheme of this type of model assembly would be similar to that shown in Figure 1.

In 2013, researchers at the US Holocaust Museum and the University of Dartmouth published their first “ensemble” model: The Early Warning Project.

The objective is to generate an early warning system, capable of detecting in which countries the risk of outbreaks of armed violence is high.

One of the notable, though tragic, successes of the project was the ability to predict a massacre in Myanmar, shortly before the massive persecution of the Rohingya began in 2016. Myanmar then ranked first in the report.

Other models, such as ViEWS, created by researchers at Uppsala University, offer unthinkable resolution levels a few years ago. The ViEWS model is focused on conflicts in Africa, and offers monthly predictive readings on multiple regions within a given state. You can predict the risk of three different types of conflict – from the state, unilateral and non-state – in a geographical grid with cells only 55 kilometers away and take into account from the first death attributable to organized violence. In Figure 3 you can see an example of prediction taken from the last monthly report prepared by the VieWs project team.

conclusion
Lately, unfortunately, when we see the terms and “War” and “Artificial Intelligence” together, the use of the latter seems to be focused exclusively on the improvement of weapons capabilities.

However, we see that it can also be used as a peace tool. A tool capable of generating early warnings that allow the UN and other humanitarian organizations to direct aid where necessary. Also, alert the most vulnerable in advance so they can find a safe place.

Artificial intelligence, like any technology, is not something “good” or “bad.” So are the objectives with which we use people.

It is possible to read the original post in Think Big Companies. To keep up with LUCA,

Spread the love

Leave a Reply

Your email address will not be published. Required fields are marked *