The Formula Is Greater Than The Sword: Data May Hold The Secret To Prevent War

The Formula Is Greater Than The Sword: Data May Hold The Secret To Prevent War
The Formula Is Greater Than The Sword: Data May Hold The Secret To Prevent War Flickr: harnisch

Humans have lived for thousands of years without understanding the rules of physics, safe in the knowledge they wouldn’t float to the moon if they jumped off the ground. For David Mace, a researcher at Caltech who has worked at Facebook, IBM Watson and NASA’s JPL, the laws of war and human nature are the same as the laws of physics. And even though they’re not understood just yet, the answer to how to stop war lies somewhere in the data.

According to Mace, analysts who are experts in two different countries, say ten years studying Saudi Arabia and then ten years studying Iran, are much more accurate than an analyst with expertise in only one country in particular. Computers, naturally, are perfect at combining vast sets of data. Mass atrocity prediction tries to do exactly this, with a top down approach.

“If we can understand those underlying rules, of how people interact with each other, about how ethnic violence is generated, about why people join terrorist groups,” Mace said, “then the hope is that this will allow us in the future engineer solutions around those mathematical equations.”

By analyzing the language used on Twitter, because people use the same pool of words when they're beginning to revolt, it is possible to pinpoint within a seven-day margin of error when a country is about to erupt in violence. However, when Google tried to analyze search queries to track flu trends, they discovered they were overestimating by 10 percent each month.

The problem with data is diversity. There are gaps within the possible data points gleaned from demographic data. As it turns out, there's still not enough data diversity in something like GDELT, which has almost everything, to do this accurately.

This is because the larger the dataset, the easier it is to find a random false positive. According to Mace, those type of approaches are not what's going to be driving the field forward in 40-50 years. Instead, a bottom up approach, like choice modeling, which tries to qualitatively predict human actions, is the answer.

“To really understand these dynamical systems, bottom up approaches I was talking about, you have to have a lot of demographic data on societies, so you have to understand where people came from, what religion they were from, various other things like these on a minute live,” said Mace. “Right now we don't have that. But as people come online, we have the potential to have it.”

While it's easier in hindsight to assign blame based on certain societal factors, information like conditions in bordering countries, rate of employment, neocolonialism, spheres of influences and outside pressure all go into making the catalyst that leads to war.

Join the Discussion