We are finally getting better at predicting organized conflict

People have been trying to predict conflict for hundreds, if not thousands, of years. But it’s hard, largely because scientists can’t agree on its nature or how it arises. The critical factor could be something as apparently innocuous as a booming population or a bad year for crops. Other times a spark ignites a powder keg, as with the assassination of Archduke Franz Ferdinand of Austria in the run-up to World War I.

Political scientists and mathematicians have come up with a slew of different methods for forecasting the next outbreak of violence—but no single model properly captures how conflict behaves. A study published in 2011 by the Peace Research Institute Oslo used a single model to run global conflict forecasts from 2010 to 2050. It estimated a less than .05% chance of violence in Syria. Humanitarian organizations, which could have been better prepared had the predictions been more accurate, were caught flat-footed by the outbreak of Syria’s civil war in March 2011. It has since displaced some 13 million people.

Bundling individual models to maximize their strengths and weed out weakness has resulted in big improvements. The first public ensemble model, the Early Warning Project, launched in 2013 to forecast new instances of mass killing. Run by researchers at the US Holocaust Museum and Dartmouth College, it claims 80% accuracy in its predictions.

Improvements in data gathering, translation, and machine learning have further advanced the field. A newer model called ViEWS, built by researchers at Uppsala University, provides a huge boost in granularity. Focusing on conflict in Africa, it offers monthly predictive readouts on multiple regions within a given state. Its threshold for violence is a single death.

Some researchers say there are private—and in some cases, classified—predictive models that are likely far better than anything public. Worries that making predictions public could undermine diplomacy or change the outcome of world events are not unfounded. But that is precisely the point. Public models are good enough to help direct aid to where it is needed and alert those most vulnerable to seek safety. Properly used, they could change things for the better, and save lives in the process.

The path from on-the-ground event to prediction requires complex analytical machinery, so we drew up a diagram of an idealized model. It follows these basic steps:

In the world of conflict prediction, there is a truism: the best predictor of violence is a history of violence. One illustration is the Early Warning Project’s 2019 predictions for the sites of new mass killings, defined as the death of over 1,000 civilians in a year due to the deliberate action of armed groups (2020 figures weren’t available at press time): the Democratic Republic of Congo, Afghanistan, India, and Myanmar rank among the 30 highest-risk countries.

The global rankings also highlight some of the model’s shortcomings. Venezuela ranks low, despite a widely held belief that extrajudicial killings by security forces have been rife. So does the US, despite a rising threat of white supremacist gun violence.

Myanmar — There have been talks of repatriating Rohingya refugees and giving them government protection, but ethnic violence against the Muslim minority has sadly continued apace.

India — In February 2019, a suicide bomber from a Pakistan-based militant group blew up Indian paramilitary trucks. Since then, new instances of violence have kept springing up, centered on the disputed region of Kashmir.

Venezuela — A United Nations report in July 2019 suggested that the government had carried out more than 9,000 extrajudicial killings in the previous 18 months. The model didn’t code them as systematic political killings, resulting in a lower risk rating.

China — China’s population size, limited freedom, and history of mass violence contribute to its risk of new mass killings. Tensions seem to be rising, as protests in Hong Kong have drawn accusations of police brutality.

United States — In September, the US Department of Homeland Security recognized white supremacist terror as a national security threat. Several mass shootings targeting minority groups indicate a trend that the model doesn’t capture.

In April 2018, Abiy Ahmed was sworn in as prime minister of Ethiopia, promising to end years of ethnic unrest and antigovernment protests. Much of the international community thought Ahmed, who is of mixed Oromo and Amhara ethnicity, might usher in an era of unity and reform. He was awarded the Nobel Peace Prize in early October, mainly for his work on a peace deal with neighboring Eritrea after a 20-year dispute.  The selection was controversial, however, as Ethiopian citizens have not experienced a dramatic new peace. By last December, ethnic violence had forced almost 3 million people from their homes.

Throughout the violence, which is still ongoing, Uppsala University’s ViEWS model has been making predictions on what will happen in Ethiopia.

The results show how far the field of conflict prediction has come: ViEWS can forecast three different types of conflict risk—state-based, one-sided, and non-state—in a geographical grid with cells just 55 kilometers on a side and take into account even a single death attributable to organized violence. That kind of resolution, impossible just a few years ago, promises to make predictions far more useful to the United Nations and humanitarian organizations that are trying to help turn the tide back toward peace.

To better illustrate how this works, we’ve identified five key moments between June 2018 and July 2019 when conflict in Ethiopia escalated. We then compared them with ViEWS’s output to see whether the violence was properly predicted ahead of time and how, when the model missed its guess, events changed its ensuing predictions.

August 2018 — The Ethiopian government moved to abolish the Liyu police force in the state of Somali—home to a large population of ethnic Somalis—and tried to oust its president. This set off a series of violent clashes

The model had identified this particular region as having a very high risk of conflict—a 1 in 6 chance. That proved right. 

September 2018 — Soldiers from the Oromo ethnic group, which make up a third of Ethiopia’s population, returned to Addis Ababa after a conflict in Eritrea. Sectarian tensions boiled over, killing 35 people—most from other minority groups.

The model did not do well at predicting this. The capital, Addis Ababa, is generally at a lower risk of conflict thanks to relatively good infrastructure, policing, and economic growth.

November 2018 — Prime Minister Ahmed carried out a series of ethnically charged arrests across the country. Violence erupted between Oromo and Somali ethnic groups, resulting in 22 deaths in the eastern village of Tuli Guled.

The arrests were distributed across the country, and the country-level prediction for violence shot up to 67%. The model didn’t precisely predict where the violence would occur, which accounts for the low score in this region.

December 2018 — A wave of violence hit restive Moyale in December after regional governments handed over security to the federal government. The violence culminated in a shooting inside a local hotel that claimed at least a dozen lives.

Refugees often flee from Ethiopia to Kenya through Moyale, and the Oromo and Somali ethnic groups have a history of clashing here. ViEWS correctly labeled the area as having a relatively high risk.

June 2019 — After an attempted coup against the Amhara state government by military leaders, the federal government ordered a five-day internet blackout. Police arrested over 250 people suspected of conspiracy in the coup, many belonging to the Amhara ethno-nationalist group.

Coups are, by their nature, hard to predict. The model did not foresee this one, but the probability of conflict rose dramatically after the coup.