Approximately 2 billion people live in conflict-affected areas. Yet the field of behavioral science has developed most of its theories, run most of its experiments, and built most of its interventions in stable contexts. When behavioral scientists do work in conflict settings, we tend to follow a familiar pattern: take what worked elsewhere, “localize” it, and evaluate it. But this isn’t enough.
This is an urgent problem. Conflict is reshaping lives globally at an unprecedented scale. In 2024, 123 million people were forcibly displaced—the highest number on record. Civilian casualties in Gaza, Lebanon, and Ukraine have reached levels unseen in over a decade, including for children.
I’ve spent nearly a decade applying behavioral science in conflict settings across 15 countries and 4 continents, and here is the hard truth: Our field is still missing critical understanding about how behavioral science works in active conflict—and there’s more to learn than we might realize.
At the Neuropaz conference this past February, I was asked to speak about the hard truths those working at the intersection of behavioral science and peacebuilding need to confront.
What follows are five of those hard truths, and what we can do about them.
Hard truth #1: We don’t have much evidence about what works from conflict settings
The vast majority of theories, experiments, and interventions we use in applied behavioral science were developed in stable contexts. Political science has studied conflict dynamics; development economics has studied postconflict recovery; and social scientists have run studies in conflict settings evaluating specific programs. But we lack systematic evidence about how conflict reshapes decision-making, how to co-design interventions grounded in what communities actually need, and how to measure and scale what works in these contexts.
Why don’t we have good evidence from conflict settings?
First, doing research amid conflict is difficult—logistically, ethically, and politically. Access and security is limited, populations are displaced, and the basic infrastructure needed for research can be absent (though we’re beginning to figure out ways to navigate these challenges).
Second, our field tends to bring existing frameworks, measures, and theories, all developed elsewhere, and then adapt them at the margins. What we need more of is systematic understanding of how conflict shapes psychology and what communities actually need from the outset, before deciding what interventions to test. Too often, qualitative research with communities happens after an intervention has been designed and evaluated, to explain what worked or didn’t. By then, we may have missed the chance to get the fundamentals of the intervention right, for the people who need it most.
Third, there is also a deeper problem: Research in this space can be optimized for measurability over impact. Research questions are often chosen based on what we can easily evaluate, not necessarily where intervention would matter most. As a field, we haven’t invested as much in developing the foundational knowledge for conflict settings as we have for stable contexts.
Conflict isn’t just harder development or more extreme poverty. It fundamentally shapes how people think, decide, and cooperate. We’ve made some progress moving beyond WEIRD research in recent years, but we cannot assume that evidence from stable contexts—even in the Global South—readily transfers to conflict settings.
Hard truth #2: Conflict creates different psychologies, and we often misdiagnose behaviors we don’t understand as “biases”
In northeast Syria, years of conflict and climate change have devastated farmland and dismantled the seed system. Starting in 2021, our teams at the Airbel Impact Lab—the International Rescue Committee (IRC) research and innovation unit—partnered with IRC’s Syria Country Team and local stakeholders to address this crisis. Working with farmers, they developed an innovative seed multiplication program where farmers would grow climate-resilient wheat varieties and donate a portion of their harvest back to a collective pool, spreading high-quality seed across their community.
The model’s success depended on farmers being willing to set aside part of their harvest for collective benefit—trading today’s security for tomorrow’s gains. For this to work, we needed to understand how people in active conflict think about the future and make trade-offs between immediate and delayed rewards. So we ran a standard time discounting survey to understand how people in this context think about the future.
We asked farmers: Would you take 500,000 Syrian pounds today, or 1 million in one month?
Forty percent chose the immediate amount, a daily discount rate of 10 percent, compared to a global average of 0.3 percent. The behavioral economics literature would call this hyperbolic discounting, or present bias—a behavior to correct.
But when we talked to farmers, we found something different. As one farmer explained: “I cannot be sure I will be alive in one month to collect a larger payment.”
This wasn’t bias; it was a rational adaptation to chronic uncertainty. When the future is genuinely unstable, when conflict is ongoing, when you might be displaced next week, prioritizing the present is survival, not error.
This is why we need to be more thoughtful with the term “bias” in crisis settings. When behavior looks “irrational,” we need to get more curious about people and context. Ask: What’s the logic of this behavior in this environment? In our program, we didn’t try to “correct” farmers’ time preferences. We adapted our approach to provide benefits sooner rather than later.
Conflict doesn’t just amplify existing behaviors. It fundamentally reshapes how people think about the future, trust, risk, and cooperation. The psychological adaptations are different in kind, not just degree.
Hard truth #3: We can’t just adapt existing interventions. We must co-design from the beginning
In the past decade, there has been significant focus on “localizing” behavioral interventions. The aim of localization is to ensure that programs are culturally appropriate, contextually relevant, and led by local actors rather than imposed by international organizations. The push for localization is a long-overdue shift away from decades of top-down development.
But in practice, localization often becomes: take what worked elsewhere, translate the materials into the local language, add some cultural elements, and evaluate it. This falls short of what is needed.
Take the case of social-emotional learning (SEL). SEL interventions have demonstrated significant impact with vulnerable children globally—improving academic achievement, attendance, and mental health. Based on this evidence, international organizations began adapting evidence-based SEL programs developed in stable contexts for use in conflict, crisis, and displacement settings. The goal was to test whether these adapted interventions could improve both learning and well-being outcomes among conflict-affected children.
IRC ran a series of randomized controlled trials (RCTs) evaluating SEL programs that had been adapted and implemented in northeast Nigeria—a place which has experienced ongoing conflict with Boko Haram for over 16 years. The RCTs were rigorously designed to test program impact on children’s social-emotional and academic outcomes.
However, the studies showed limited or null effects on the social-emotional outcomes they were designed to address. The programs had been translated, cultural elements added, frameworks adjusted, but they were still built around assumptions from elsewhere.
Teachers weren’t using them, as they found them disconnected from what mattered locally—both the skills and values children needed to succeed in their communities and the classroom realities teachers faced daily.
Sometimes the problem isn’t that researchers don’t adapt interventions or even that communities won’t adopt behavioral science research. It’s that interventions designed in different places, with different people, under stable conditions are less likely to work in conflict settings.
It’s not that we can’t draw from the broader behavioral science evidence base, but we have to acknowledge that adaptation can only go so far. Developing effective interventions requires co-designing with the community from the start.
In Nigeria, we shifted to this approach. Our behavioral science team at Airbel joined the project to identify barriers and co-design interventions that could overcome them.
We didn’t ask ourselves, “How do we get teachers to use these SEL activities?”
We asked teachers, parents, and community leaders: “What do children in your community need to grow up to become successful adults—beyond academics? What values matter here? What skills? And how can behavioral science help design supports that work for what people actually need?”
Community members identified skills and values that mattered most to them. Some overlapped with standard SEL frameworks, like self-regulation and empathy. Others were different, like respect, obedience, and discipline.
As one teacher told us: “Discipline is the most important thing because you need to be disciplined to achieve a lot in life.” This wasn’t in our original SEL framework. But it was central to what communities saw as essential for children’s success.
We then co-designed new activities with teachers and community members around the skills they had identified—discipline, respect, tolerance. We kept the active ingredients—the psychological mechanisms that make SEL effective, such as conflict resolution and social problem solving, and focused on short, targeted activities that build specific skills. But the specific activities were co-designed from the ground up with teachers and community members around the skills they had identified.
Same psychological mechanisms, different expressions, and co-designed from the start.
In a pilot study after this, teachers were implementing SEL activities for an average of 18 minutes per day—the intended dose—compared to the previous program, where most teachers reported they hadn’t used the SEL activities since their initial training.
This is the difference between the light touch localization and genuine co-design. Localization takes what works elsewhere and adapts it. Co-design starts with what matters here and uses behavioral science to support it.
Hard truth #4: We still need to learn how to sustain and scale what works
In northeast Syria, years of conflict have also created deep tensions between internally displaced people and host communities—those who stayed in place during the war. Our behavioral science team at Airbel, again in partnership with IRC Syria, is currently working on a project to increase peacebuilding behaviors between these two groups, including internally displaced people and hosts serving together on local committees, and jointly deciding how to resolve intergroup conflict and allocate resources within the community.
In an early workshop testing behavioral interventions to build these behaviors, an internally displaced person said: “I used to think the hosts are racist, but after the exercise I realize that we are the same.”
That moment of rehumanization was profound. And it happened in a single workshop with a few dozen people.
In many cases, exciting behavioral science interventions that address conflict risk remaining one-offs. Even when we find interventions that create real change, we lack clear, evidence-based pathways for scaling them. Contact interventions, one of the most well-documented interventions in conflict settings, show short-term prejudice reduction. But how do we move from workshops with dozens of people to population-level change? How do we design for scale from the beginning, in a way that is both deep (contextually relevant) and wide (reaching most of the population)?
In our recent seed-security Syria work, we’re designing and testing this from the beginning. What behaviors need to spread for social cohesion to improve at a population level? How do we design and measure interventions that can reach entire communities across a region, not just individuals? Who are the most effective spreaders of behaviors? What motivates them? Under what conditions do people invite others into programs?
Hard truth #5: We’re measuring the wrong things
Most peacebuilding research measures individual-level attitudes, like prejudice reduction, trust, or empathy. But we need to measure actual behaviors, not just attitudes. Do people cooperate across group lines? Do they engage in violence? Self-reported attitudes are easier to measure, but behavior is what matters.
And still there’s a deeper mismatch: We’re intervening at the individual level and measuring at the individual level, but often the things we’re trying to prevent—for example, intergroup violence—happen at the community level. We can’t assume that changing individual attitudes or behaviors translates to preventing community-level violence without actually measuring whether it does.
* * *
A different approach: Five principles for research in conflict settings
If our field were to approach behavioral science in conflict settings differently, here’s what it might look like.
First, start with communities, not problems. Conventional approaches often pick a problem first, then ask, “What’s measurable?” This means optimizing for what we can count, not what counts. Instead, ask people directly: What challenges do they face in their day-to-day lives? What are their goals? What does a better, healthier, safer life look like to them? This scopes where intervention would matter most.
Second, co-design with communities and with local researchers as true co-scientists. Local teams bring lived experience, cultural expertise, and community trust that outside researchers cannot. This should shape intervention design from the start, before deciding what to test. This includes specific wording, delivery mechanisms, and measurement approaches. We need to design for the realities people actually live, not the ones we assume.
Third, identify the active ingredient, then ask how it works here. Don’t just translate an intervention. Ask: What’s the psychological mechanism that worked elsewhere? Then ask: How does that mechanism work in this specific context? For example, in our initial testing of “conversational receptiveness,” an intervention that helps people listen across disagreement, in Syria, we don’t plan to just translate scripts. We’re asking: How do people actually express active listening across group lines in this context?
Fourth, test assumptions at a small scale before full rollout. This is good practice in any setting, but it becomes even more critical in conflict settings, where the context is likely to be novel and rapidly changing. Start by prototyping—turning solution ideas into simple, tangible mockups (drawings, paper concepts, role-plays) and testing them directly with users to learn what resonates and co-design what needs to change. Then run small behavioral tests or experiments to check: Does this actually work as we think? Does it help or hurt? This rapid, iterative testing with users allows teams to surface contextual misalignments early, before investing in full-scale implementation.
In our seed-security work in Syria, we found that farmers donated 38 percent more to others when they perceived the “others” as similar to them, confirming our hypothesis that emphasizing common ground can increase cooperation.
But small-scale testing also caught assumptions that would have derailed the program. We hypothesized that women would need to bring a family member to enroll, based on what we’d been told about cultural norms. Testing revealed this was wrong in some areas and right in others for completely different reasons: In areas with recent attacks, women needed someone for physical safety traveling to sites, not for cultural reasons. Without testing, we would have applied a one-size-fits-all rule that failed everywhere.
Finally, design for scale from the beginning, but define what scale means for your context. Scale doesn’t always mean exact replication or global reach. Given that conflict creates different psychologies and contexts vary enormously, an intervention might only spread effectively within a specific area or region—and might need to adapt as it spreads. The question isn’t, “How do we scale this everywhere?” It’s: “What’s the right scope for this intervention given the context? And what conditions allow it to spread within that scope? Ask from the start: What behaviors need to take hold for this to work at the target population level in this setting? Who can help them spread, and what social infrastructure needs to be in place?
As global foreign aid faces unprecedented cuts and humanitarian needs reach all-time highs, we cannot afford to keep testing interventions that aren’t grounded in the realities of people’s lives. We cannot afford to treat 2 billion people as an edge case.
This work would not have been possible without the IRC Syria and Nigeria Country Programs who co-designed, led and implemented these projects, the Climate Innovation Team at Airbel, Vaidehi Uberoi, and Tahirat Omolara Eniola.
