Creating Paradigm Shifts in Conflict Early Warning Systems


Approximately 160 million people died in wars during the 20th century.  Nearly a million more have died in the 21st century, including during current conflicts in Iraq, Sudan, the Democratic Republic of the Congo, Afghanistan, Burma, Syria, and more.  Claims that we live in the safest time in human history are true for some, but tragically false for others.  The war in Syria alone has claimed over 100,000 lives and shows no sign of stopping.  Across the globe civilians are threatened by oppressive regimes, violent opposition groups, and battles over natural resources.  The realization of the late 20th and early 21st centuries that perhaps the international community, especially governments and advocates in the West, could prevent these types of conflicts outside the realm of high-level diplomacy was transformational.  However, the last several years with failure to prevent after failure to prevent proved that it’s time to question the foundations on which our understanding of conflict prevention, especially atrocity prevention, lie.  

Conventionally, technology-based atrocity prevention takes the form of early warning systems (EWS).  These systems are hierarchical, academic, require significant technical expertise, utilize a vertical information feedback loop, and are typically designed and located in the West.  A system’s success is often based on the accuracy of its prediction capacity rather than how many lives it saves.  Patrick Meier, a leading scholar on early warning systems, identified one of the major flaws with the convention when he said, “Reports don’t protect people, nor do graphs. People protect themselves and others.”  This means that even the most cutting edge early warning systems that have the best of intentions are, at best, minimally effective.

(LLAMA – Locally-Led Advance Mobile Aid,  Source:

Systems that warn communities about impending natural disasters, however, are relatively advanced and often successful.  A local flooding early warning system (LFEWS), the most common type of natural disaster system, consists of four key elements: 1) Risk Knowledge 2) Monitoring and Warning 3) Dissemination and Communication 4) Response Capability.  


In other words, a community assesses their risk of flooding, monitors river and water table levels, warns community members when levels are high, and then responds appropriately.  This type of EWS is most famously used in the Philippines, but also exists in sub-Saharan Africa and central South America.  The key to the system is simplicity.  The technology is minimal, usually consisting of river gauges to monitor impending floods and a SMS texting network, radios, or bullhorns to alert community members.  This allows for immediate response to danger.  Moreover, the emotional and physical urgency involved in these systems is tremendous.  No one has more motivation to quickly flee from danger than those about to be harmed or killed.   

These are lessons that the atrocity prevention community should humbly study and consider incorporating into their EWS.  The goal should be to make systems work for people in harm’s way which means making them horizontal, localized, and accessible.  In this case, fancier isn’t better and, as Meier notes, reports don’t save lives.  

A local conflict early warning system (LCEWS) would look similar to a LFEWS – citizen-based, ultra-responsive, and highly localized.  It would be guided by an overarching, comprehensive concept of human security which would intentionally provide protection from armed violence and, perhaps unintentionally, empower communities.  There is, however, a place for advocates based in the West.  Experts can act as a support system for communities that are operating LCEWS in technical and strategic capacities.  Additionally, predictive models, hate speech databases, crisis mapping, and other tools can be used to reduce the threat of violence over the medium- and long-term, similar to the way natural disaster systems differentiate between tornado warnings and climate change warnings.


In the coming weeks I will explore what the core components of a LCEWS should be and how they can be applied.  My goal is to envision a system that puts the power into the hands of people most affected by atrocities and more effectively protects those unjustly ravaged by violence.  



  1. Michael Donahue

    I find it ironic that this post is generated from Toronto, Canada, given the recent events around natural disaster, yet I do find this to be an interesting beginning to a valuable conversation that is yet only a brief piece of the efforts. Kenya’s violence in 2007, though reported by media, govnt and NGO’s as a surprise, was not for many on the ground, yet still occurred, and visa verse, in 2011 when the same media, govnt. and NGO folk reported the impending and likely violence, yet none of us have heard why they failed in that warning. Maybe has something to do with money, as well as conventional reporting.

  2. muhammad feyyaz

    interesting indeed. i am of the opinion that perhaps early warning may not be possible in certain situations which are changed by external dynamics. for example, US invasion of Afghanistan during 2001 was never thought to have brought with it or in its immediate wake the kind of conflicted conditions that we know witness in tribal areas of Pakistan. one can argue that its implications could be studied before hand, yes. but just go back in 2001 with 9/11 in mind, and imagine the hysteria pervading global mindsets against terrorists, coupled with willingness of nation states to ally with US. who would have bothered what this invasion or military action will do, and today we have lost over 40000 people in Pakistan alone. northern Ireland is another example.
    i will propose that early warning notion is relevant to intrastate conflicts only, and secondly it is agreed that local communities can do much better for themselves than outsiders, but then concept is not applicable universally, it is germane to certain societies that have the potential of generating violence, characterized by religious / sectarian cleavages, ethnicity, clanship / tribalism, faith based polarisation etc. consequently, by using such a framework, we can target where this concept is more effective, and can consequently customize it according to local dynamics. i am sure this has been done in many ways already. thank for providing opportunity for this discussion.

  3. Sean Langberg

    Thank you, Michael and Muhammad, for your responses. I am glad the post generated some interesting thoughts.
    Michael – I agree that the 2007 Kenya elections were a turning point for the atrocity prevention community. Despite the fairly obvious indicators that you mentioned, it caught many people off guard. In my opinion, this is because of attention placed elsewhere and a focus on unconventional conventional drivers at the time rather than conventional ones such as elections. With regards to 2011, in my experience, there were several warnings about violence in Kenya prior to the elections and it is good that those predictions were wrong. There were follow-up pieces about why it didn’t happen. One of the inherent challenges with atrocity prevention is that we are predicting events that we don’t want to happen. If we are successful, it’s hard to prove a victory because we are warning about violence that doesn’t happen.
    Muhammad – You are correct in saying that EWS primarily deals with intrastate conflict. This is partially because creating warning models in a domestic context is easier than an international one and because most mass atrocities of the late 20th and early 21st century occur during intrastate conflicts i.e Sudan, Burma, Bosnia, etc.
    Thank you both for providing feedback. Feel free to continue the conversation.

  4. Greg Stanton

    Sean Landberg’s comment on how to best develop early warning systems is excellent. The longer Genocide Watch has worked on prevention, the more we have concluded that local early warning is best:
    1. Local knowledge is nearly always better than knowledge reported at a distance.
    2. Such early warning allows for early intervention, often without resort to violence, by local people to prevent violence. It can also make use of local law enforcement authorities, churches and mosques, schools, and citizens groups.
    3. The so called “international community” is a myth. The UN is seldom able to act to prevent violence because of the paralysis of the UN Security Council, and because of the lack of action by the General Assembly, which no longer uses its “Uniting For Peace” powers. Regional organizations are much more effective when military force is needed, but even they can be stymied by regional politics.
    4. The 2013 Kenya case is an example of the success of early warning. When Kenyans saw what it took to pull Kenya back from the brink of genocide in 2007, all sides were determined not to repeat that scenario. A massive civil society and governmental effort was made to counter hate speech, and to deploy police to voting sites and potentially volatile areas. This preventive effort was organized locally by Kenyans with some international assistance. Rather than claiming that Kenya in 2013 was a “false positive”, it should be seen as successful preventive action. The Sentinel Project is to be commended on its participation and assistance to that effort.

  5. Karen Volker

    Thanks, Sean, for starting this conversation. We also think about the problem of stopping violence requires a systematic approach. But we would argue that the system should not only contain the warning but also the response. Our approach to interrupting and stopping violence is based on the WHO’s method for interrupting and stopping the spread of epidemics. We’ve adapted the methodology to apply to violence, treating violence as a contagious disease, a learned behavior. Both elements are critical – knowing when something is going to happen and then doing something to stop it. I’d like to work with you to continue thinking through this problem with the added layer of creating a global network of violence interrupters who can work locally to interrupt violence before it happens. We can also think about how we can use technology to interrupt violence. Feel free to contact me offline to continue the conversation. Karen Volker (

  6. muhammad feyyaz

    thank you indeed for your reply. i am trying to examine a possibility that mere onset of an intrastate armed conflict in a nuclearised region such as in India and Pakistan, can or should itself be conceived as a warning for a possible major interstate war under nuclear overhang. early warning literature from this dimension has yet not received any attention in the existing peace and conflict literature. any idea you like to share on this aspect of warning paradigm?

  7. Priyan Seneviratne

    It was interesting to go through this article. I was part of a team who introduced a citizen based early warning and response system in Sri Lanka during the ceasefire between 2002- 2008 period in the eastern province of the country. By the time it was introduced it was the first time where an EWS system was incorporated with an organic ER mechanism. Actually the system was really a comprehensive model that has gone unnoticed since . It had GIS mapping capabilities of conflict hotspots, a relational database to monitor micro level conflictive indicators which was lacking in the FAST system that was made operational by SWISSPEACE sometime ago, an sms alert system which connected the grassroots to the track 02 and 01 levels instantly, and an organic community platform with organic members with lot of social capital for engaging in rapid response. Sad to say all of that hard work went down the drain . However it has been fortunately picked up by some other organizations operating in different parts of the world.
    Majority of the failure of an EWS i believe lies with the problem of lack of or all too late response which i believe Barrs is also attempting to address through the proposition of the LLAMA model. In practice the issue was raised when developing CEWARN but was never addressed at the micro level. From experience all i can say is that for a successful EWS and ER system to be realised there needs to be a certain degree of political will between the protagonists to the conflict to prevent violence. It is more successful at preventing violence at micro level than at macro level because the at the macro level.
    At a moment if violence has already erupted between communities then the system should be inclusive enough to have all elements in the communities (including spoilers) who could be mobilized for its containment. Social capital of different actors at all three levels are vital in this regard and they are the ones who need to be mobilized in the long term.
    At the community level when violence erupts the most important next step would be to identify where it might spread to and contain it at the location it erupts. This requires prevention of spreading of rumours and mobilization of local community associations. ( if you already have not can refer to Varshney and Horowitz in this regard). In the Sri Lankan case can say religious leaders, farmers associations etc have been vital in this regard.



  1. A Blueprint for Local, Nonviolent Responses to Mass Atrocities (Part II) | The Widening Lens - [...] the information and the means to move to lower-risk areas when physical escape is the best option.  LCEWS are another important…
  2. the four dilemmas of a mass atrocity | securing rights - [...] processes are based on the systematic gathering of mass-atrocity evidence. As Sean Langberg observes, these two categories feed into…

Leave a Reply to muhammad feyyaz Cancel reply

Your email address will not be published. Required fields are marked *