Engineering Peace

Global development has a war problem. The world overall is seeing more conflict every year, and the majority of the damage is in developing countries. And while war in industrialized societies can even support economic growth, at least in the short term, developing countries can take decades to recover the growth that violence wipes away.

Yet there is a belief in the development sector that war — and its catastrophic effects on weak economies with fragile infrastructures — is an inevitable feature of human society. That even if we had more effective tools, the challenge of war is far bigger than the resources that the development community can muster to meet it. In that sense, war is treated much the same as a volcanic eruption or a cyclone: We can build resilience once it happens, but we can’t prevent it from happening. 

Pessimism reigns following resurgences of conflict in areas that had seen huge investments in “stabilization,” like Afghanistan. As Yvonne Helle, former special representative to the United Nations Development Programme’s Programme of Assistance to the Palestinian People, told me days after the October 7 war in Israel and Gaza broke out, “I feel so useless and naive in believing that development was possible.”

It’s true that many traditional development approaches may not be great investments during wartime, when interrupted supply chains, scattered communities, and scrambled local power relationships all make such work far more expensive. But it is too fatalistic to treat war as an inevitability.

A new wave of empirical research is starting to expand our tool kit for preventing cycles of violence, at least when it comes to non-state actors. Our next step is to figure out what really works — and how much it costs.

Peacebuilding is notoriously difficult to evaluate. The chaos of war and displacement makes it difficult to survey communities. Conflict takes on different forms in different places, and extensive testing is needed to make sure promising approaches hold up in multiple environments. Destructive violence, though frequent, is rarely predictable enough to observe how it responds to a controlled intervention. It’s hard to recruit potential violent actors, often on the fringes of society, into studies, and even harder to observe how they behave.

These challenges shouldn’t obscure how far we’ve come in understanding the causes of violence. Since the mid-2000s, a revolution has been underway in conflict studies, driven by the marriage of novel large datasets on incident-level violence and observational methods developed in more traditional fields of development economics. This union was spearheaded by academic collectives like the Empirical Studies of Conflict project¹ and the Irregular Warfare Initiative, which use incident-level data to test core assumptions behind efforts to stabilize regions suffering from irregular warfare. In Iraq, for instance, it turns out that young men’s job circumstances don’t have much of an effect on their decision to join insurgencies, overturning much conventional wisdom and casting a pall over job creation schemes that USAID and other donors had heavily invested in. 

This first wave of studies of conflict based on surveys and other observational data paved the way for randomized field experiments. These experiments borrowed techniques from the randomized controlled trial movement in public health, and were often hosted by the same institutions, including groups like the Abdul Latif Jameel Poverty Action Lab’sCrime and Violence Initiative and Innovations for Poverty Action’s Peace & Recovery Program. This work has helped validate existing approaches, but measuring the impact of any intervention on real-world violence remains a challenge.

One well-known paper illustrates just how tough this line of research can be. Working in the slums of Monrovia, Liberia, University of Chicago political scientist Chris Blattman and his co-authors found that eight weeks of cognitive behavioral therapy followed by a $200 cash transfer significantly reduced the likelihood that young men (previously engaged in illicit activity) display a range of violent and criminal behaviors at least a year after the study began. To show this, they had to find brave local enumerators willing to spend several hours each day over four days in casual conversation with men who were at high risk for violence, probing them with questions about their behavior. Closely and rigorously studying high-risk populations like this is critically important — new organizations have even sprung up in other parts of the world to test whether these findings hold up — but also critically difficult. This type of research is, therefore, vanishingly rare.

Instead, most experimental research focuses on intermediate outcomes, like extremist attitudes or hostility toward opposing groups, because they are easier to identify and measure. While this work has limits, it can still be very useful for identifying the seeds of future interventions. 

For instance, research has shown that making complex identities salient can show people that they have something in common with members of groups they might think of as their enemies, that communicating through respected leaders can reduce support for extremist actors. It suggests that teaching and practicing the skill of seeing another’s experience as if it were one’s own can improve social cohesion and that mass media (for instance, via a radio soap opera) can increase positive interactions with members of opposing groups.

Some of the greatest contributions have illuminated what doesn’t work. Most people might intuit that simply bringing people from opposing groups together might lead to greater understanding and lower the motivation to fight in future. But dozens of experimental studies illustrate that mere intergroup contact does not show reliable impact and may even make things worse. The reason is simple: Spending time with members of rival groups, whatever the benefits of getting to know them in the long term, makes salient the fact that there is a conflict in the short term. Without carefully constructed conditions in which the contact occurs, such interactions can cause excess stress and even harden the resolve to fight, in turn preventing the kind of bonding needed for long-term inhibitory familiarity to develop.

The fuzzy picture of peacebuilding effectiveness may now be coming into greater focus. A 2021 meta-analysis by the impact evaluation organization International Initiative for Impact Evaluation identified 37 high-quality empirical evaluations of peacebuilding programs — only three of which existed 10 years prior to publication — but they all measure these intermediate outcomes, like trust or tolerance of other ethnic groups. There are strong theoretical and empirical reasons to believe that building social cohesion is an effective way to reduce political violence, but much more work is necessary before we can be confident that these peacebuilding initiatives are really a cost-effective way to reduce war itself.

The gap between experimental results and ultimate outcomes is one reason funders are wary of the peacebuilding space. In 2022, when investigating “civil conflict” as a possible cause to recommend for philanthropic donations, charitable foundation Coefficient Giving declared itself to be “quite uncertain how much of a difference micro-scale interventions like cognitive behavioral therapy will make on macro-scale events like war.” (Asterisk receives funding from Coefficient Giving.) 

Researchers trying to close this gap have two problems. First, as we’ve seen, work like Chris Blattman’s, which actually measures whether individual-level interventions reduce violent behavior, is rare. Second, even these studies are too limited to address the question we’re really interested in: How much does reducing individual violence affect the chances that a large-scale conflict will break out?

At Peace Per Dollar, the new initiative at UC Berkeley’s Center for Effective Global Action, we are trying to answer all these questions. Our goal is to show that approaches that have demonstrated consistent effects at the individual level — like school-based empathy education programs — can also deliver real value in preventing larger-scale avoidable violence, and that they can do so cost-effectively. In fact, given the enormous destruction wrought by war, we believe that such tools could even end up being as effective as, for example, bed nets for preventing malaria deaths. If the development community knew the return on their investment, they could invest more confidently in preventing destructive violence.

Our first task is to understand the scale at which peacebuilding interventions would need to be deployed in order to do their job. Would absolutely every denizen of a conflict-prone region be required to undergo, say, a highly intensive exposure to outgroup narratives in order to meaningfully avert the likelihood of a future return to violence? Surely not. But, then, how many would and to what degree of intensity?

An analogy may be found in global health with vaccine efficacy. Vaccine trials demonstrate the ability to preserve individual health in the face of a pathogen. But to assess whether a given serum can prevent an epidemic, you need the bridging concept of herd immunity. In peacebuilding, we are still figuring out if the serum works; we haven’t yet gotten around to thinking about what population-level “immunity” might look like, let alone funding the kind of large-scale interventional trials required to actually observe it in the field. 

While we wait for such trials, however, it’s not impossible to imagine how war prevention might work. Of course, we won’t be able to prevent all conflict. Many large interstate wars, including Russia’s war on Ukraine and the U.S. and Israel’s war on Iran, are widely viewed as the product of the determination of a small number of leaders with extraordinary access to military power. 

But in the developing world, war is much more often the product of a readily available recruitment pipeline. Cut off the supply of willing fighters, and there is no war to be fought. We often imagine such fighters to be ideologues, but this is usually wrong. When individuals participate in political violence, it’s almost always as a result of following a pattern of behavior that begins long before they are even aware of it. Each choice they make along the way — for instance, deciding to start hanging out with a particular group of friends — pushes them unwittingly into a situation where they may be required to act either as accessories to or as perpetrators of violence. Getting out of this situation, perhaps by abandoning a social circle and actively seeking a new one, requires more and more effort as they are pulled deeper in.

This is not to say that people who eventually join violent groups don’t care about the cause. They do, often to the point of holding irrationally “sacred” values. But this is equally true of people who join nonviolent protest movements instead of violent ones. The difference between the two can be as small as a momentary lapse in judgment during a critical period of life. Accepting an invitation by a stranger. Scrolling through an innocuous-seeming news article. Responding without restraint to a perceived insult in a public forum.

All this implies that potential fighters can be deterred with less effort than it would take to change an ideologue’s mind. But how many participants are required for a wave of violence to be considered a war, and how much effort would it take to deter them?

An upper bound could be found in the case of the Rwandan genocide, one of history’s worst atrocities. According to one estimate, up to 14% of Hutus participated. Now imagine a radio program that increases the likelihood of an individual expressing dissent by around 25%, meaning that four people would have to be exposed in order to produce at least one individual who refuses to follow orders. According to this simple calculation, sitting 56% (4 x 14%) of Rwanda’s Hutus down to listen regularly to such a program might have produced at least as many refusers as willing killers.The cost of reaching so many people in this way might be enormous. But there might also be ways of bringing it down. If you believe in some version of herd immunity, you wouldn’t need to flip all potential perpetrators into refuseniks, just enough to create a visible norm of dissent. Technological progress can help drive down delivery costs further by locating the highest-risk people at their highest-risk moment (say, right after they receive orders but haven’t yet decided to follow them). AI-powered social listening tools can detect toxic discourse that may predict future violence. At the individual level, a better temporal understanding of the stages of radicalization can lead to more precise and thus less costly deployment.

Development practitioners need an engineering mindset to instrumentalize the new tools that researchers are fine-tuning in real time. This would allow them to work in areas where war is likely with the confidence that they can actually do something about it. If they had more clarity on the scale required for war prevention, funders would also be better able to assess whether the costs of deployment would be worth the likelihood of successfully preventing catastrophe. (They’d first have to demand greater cost transparency from their grantees: Only one of the 73 interventional studies we reviewed at Peace Per Dollar contained basic intervention cost information.)

Greater alignment toward war prevention might therefore help build an investment case to bring in funding from philanthropists and social investors who previously would not have considered it. Like a natural disaster, war is effectively priced into the cost of every other development intervention. It’s also very expensive to clean up: A 2018 UN analysis estimated that spending one dollar on violence prevention could remove the need to spend up to $16 later on, even when accounting for a reasonably low likelihood of success. 

Enormous opportunity now exists to align incentives toward war prevention in the development community. What would a focused research-to-practice pipeline look like? We are talking about an effort to create a “weapon of mass peace,” arguably on the scale of the Manhattan Project. In place of centrifuges, we would need large-scale cluster randomized trials where outbreaks might be possible over a reasonable time horizon and the patience to see them through. But you can see a future in which we have the tools to avert violent conflict if you squint. Engineering peace is hard — but not impossible.