How big data could anticipate violence and promote peace

Big data – the use of advanced computing techniques, like machine learning, to analyse massive amounts of data – is continually demonstrating its benefit across scientific fields and the business world, as it produces fresh insights.

Most are familiar with such advances related to medicine; they span disease diagnosis to tracking and reducing the spread of COVID-19. Big data boosts to driverless vehicles and industry automation are also often discussed.

However, this technology is only just beginning to have an impact on our ability to better forecast violence and conflict, and hence inform potential preventive steps.

Recent trends show deteriorations in peace worldwide since 2008, threatening to reverse encouraging developments made in decades prior. Big data has the potential to help buck the trend and reduce the prevalence of violence: sexual assault, election-related and political violence, and climate-aggravated conflict included.

Here are three major areas where big data can be further applied to advance efforts along these lines.

1. Understanding online extremism to stop offline violence

In recent years, there has been a substantial increase in the spread of extremist narratives and content across social media platforms, at times being linked to offline acts of violence such as the 2019 New Zealand terror attack.

Most attempts to counter these threats have focused on banning content and users, yet research has demonstrated that this approach is largely ineffective and at times counterproductive.

The group Moonshot CVE utilises big data techniques to understand how extremism online might lead to violence offline. This leads them to develop techniques that more effectively respond to extremism both online and offline, helping to prevent individuals from moving down a path toward violent extremism.

Their work has been used against ISIS as well as across the Unites States to counter the growing threat of far right extremism.

Similarly, IBM has developed tools that allow for hundreds of millions of tweets to be monitored in order to analyse sentiment, potentially pre-empting criminal activity.

And yet, such developments do not come without drawbacks. There are growing concerns that misusing big-data techniques could exasperate social tensions, with some criticizing the development of racially biased algorithms, for example. The use of such algorithms could potentially interact with other sources of bias and should thus be treated with caution, particularly in the context of crime prevention and policing,

Safecity is an initiative that uses big data to create maps that identify assault hotspots based on user updated instances of assault. This proves useful in assisting people to avoid areas where their chances of being assaulted might be greater. Similarly, HarassMap utilises georeferenced and crowdsourced data to highlight where assaults have taken place, increasing public awareness and thereby reducing spaces where such assaults can take place.

2. Foreseeing political unrest

Recurring instances of violence in many countries are a major concern for peace and hamper global efforts towards sustainable development, trapping populations in a self-perpetuating cycle of impoverishment and further destabilisation.

This is a serious challenge for a number of African countries in particular. Better predictions of such violence would be extremely valuable and allow for more effective prevention efforts.

Researchers behind the project ViEWS have begun to integrate machine learning into the creation of high-resolution predictions for political violence across Africa, up to 36 months in advance. With promising initial results, this early warning system is an important beginning in attempts to increase our knowledge of where and when such conflict will occur, and therefore interrupt cycles of violence.

Following the deadly violence of the 2007 Kenyan election, a group of technologists founded the group Ushahidi (meaning 'testimony' in Swahili). Ushahidi has used big-data tools such as natural language processing to analyse crowd-sourced data to monitor election violence in real-time, helping to combat its spread.

They have since extended their work to include more than 160 countries and project types such as the Syrian civil war, Occupy Wall Street and the 2011 Christ Church earthquake.

The violence that followed the 2020 US presidential election highlights how widespread such challenges are and hence the need to address them at a global scale. Big data can and should play its part in addressing these issues.

3. How climate change corresponds to conflict risk

Experts have repeatedly warned about the impact of climate change on the risk of violent conflict globally.

Climatic stress and its effects on human lives and livelihoods has become a new frontier in conflict research; specifically, accurately predicting when and where violence is likely to occur in connection with this.

In a communication with the director of ViEWS, Professor Håvard Hegre argued that achieving this goal would depend on using diverse open-source data on climate, as well as a number of intervening socioeconomic and political variables.

Modelling the impact of climate-induced food production shocks on conflict risk, for example, is important in this effort. The World Food Program has developed HungerMap Live, which integrates data on food, nutrition, weather and economics to forecast food insecurity.

Such predictions could be combined with information on political sentiment and polarisation from databases like GDELT, a vast project monitoring all sources of news (print and online), as well as social media, in over 100 languages in almost every country around the world.

This massive and constantly updated stream of data has proven useful in monitoring the time-sensitive aspects of the escalation of social unrest in the past, demonstrating its potential for identifying emerging security threats in the wake of climate change.

The World Resources Institute's Water, Peace and Security initiative has developed a water conflict-related early warning tool. They employ machine learning to analyse diverse data including economic, socio-political and environmental indicators, in order to forecast violent hotspots in some of the more water-stressed regions across Africa, the Middle East and South-East Asia for up to 12 months in advance.

So far, the model is correctly forecasting an overall occurrence of 86% of future conflicts, cautioning, however, a difficulty in forecasting with high spatial precision.

A cautionary tale from COVID-19

Big data applications have certainly begun to demonstrate their potential for predicting violence and how to help prevent it. However, progress made in foreseeing crises and conflict must not be at the expense of individual liberties.

Current concerns about data-driven responses to the COVID-19 pandemic tell a cautionary tale. Likewise, possible biases against minorities need to be averted, as well as the risk that authoritarian governments may leverage big data for further persecution.

However, there is strong potential to improve foresight and prevention capacities. With assistance from government and funding bodies, as well as input from experts and field practitioners, opportunities abound for advancing the field of computational conflict analysis and helping to improve conditions that are conducive to peace.