Keywords

1 I Predict a Riot

Once during what was called ‘the troubles’ I walked up a street in Northern Ireland with international visitors. As we got near the city centre, I noticed that lots of young men were lurking in an unusual way, around the sidewalks. I realised that while none were wearing balaclavas or other disguises, hats were being pulled lower, and scarves higher. There were very few whose faces I could see. But it was more the way they were behaving. There were looks between them. They were all ‘waiting’ and I felt a sort of feverishness in the air. Suddenly I was on high alert. I looked up into the town and realised a march by ‘the other side’ was taking place and the young people could be expected to oppose it. We were right on the interface between divided communities and I suddenly knew: ‘I predict a riot’. It would not be serious, but could injure people. I knew that police—who I could see up the street and were clearly prepared, could fire plastic bullets if things escalated. The chance that my guests would be hurt was slim, but I did not want to take it.

I told my guests, ‘let’s go back’. Some were quite upset with me, when I told them I would explain later. There was trouble as it turned out. No-one got seriously hurt, but it was better not to be around it. Interesting to me, my guests were really surprised at my decisions and the explanation I later gave. They kept asking ‘but how did you know that was going on?’ The told me they had not seen anything untoward and were most surprised. For my part, I wondered how they could not see what was so obvious.

Now I realise I could have answered: I was alerted by my in-built, highly sophisticated, Conflict Early Warning System.

2 Conflict Early Warning Systems: What Are They?

It was not a hunch that triggered my anticipation of conflict; it was my expert local knowledge of conflict put together with observed data. This knowledge came from a life lived in conflict, from having been in similar situations before, coupled with wider political understanding of conflict triggers. This combined to produce an ‘instinct’ that was predictive. You will have had similar instincts in similar situations, and re-examination may show they were based on knowledge and rationality, despite being experienced in the moment as ‘a feeling’.

I called my reaction an in-built Conflict Early Warning System (CEWS), and boldly claimed it to be ‘highly sophisticated’, because it is impossible to parse out all the relevant data sources I unconsciously drew on, and the life-long underlying expertise that informed what looked like an unscientific ‘hunch’.

Nowadays we think of CEWS as technological, data-driven and scientific. CEWS have been a holy grail of conflict analysists. Technology such as we examined in outline in Chap. 2, promises a capacity to deliver a science of early warning drawing on infinite and varied data, at speed, to provide automated systematic predictions of conflict. Can it?

Often times the local person’s highly sophisticated ‘hunch’ will be better, and in practice CEWS often rely on hybrid forms of knowledge across ‘data analytics’ and ‘human expertise’.

3 From EWS to CEWS

Early Warning Systems or EWS provide early indications that something concerning is happening, based on assessing physical events on the ground, and predicting their consequences. They are used widely to monitor weather and natural disasters and climate change: a confluence of natural events can indicate that a major event might be imminent, such as a storm or earthquake. We usually think of EWS as both systematic and automated, and therefore scientific in some way. Yet they mostly depend on human responses to be effective. In the conflict area, they have started to be called ‘Conflict Early Warning Systems’ or CEWS, the term we will now use, and other definitions can be seen in the box.

Early Warning System: A warning system that can be implemented as a chain of information communication systems and comprises sensors, event detection and decision subsystems for early identification of hazards. They work together to forecast and signal disturbances that adversely affect the stability of the physical world, providing time for the response system to prepare for the adverse event and to minimize its impact. (Waidyanatha, 2010)

Conflict Early Warning Systems: Systems to alert people to the risk or onset of militarily violent conflict. Their specific purpose is, ‘to identify and trigger actions to reduce the onset, duration, intensity, and effects of multiple forms of political violence’. (Muggah & Whitlock, 2022, p. 1)

World Bank: A process that (a) alerts decision-makers to the potential outbreak, escalation and resurgence of violent conflict; and (b) promotes an understanding among decision-makers of the nature and impacts of violent conflict. (Defontaine, 2019)

Hague Centre for Strategic Studies: A CEWS warns specifically for the onset, occurrence, escalation or resurgence of various forms of political violence. (Sweijs & Teer, 2022)

4 Who Do CEWS Alert?

CEWS aim to produce a response in a range of different actors (not necessarily all targeted by a single system).

Humanitarian agencies. These can be alerted to the possibility of humanitarian consequences of conflict such as death or injury, hunger, new health needs or mass movements of people.

The diplomatic community. This community may be alerted to support local crisis interventions, and forms of diplomatic or military or humanitarian intervention, or to plan to evacuate their own nationals. Anticipatory diplomatic responses may be particularly important if the nature of the conflict is such that it could trigger ever wider international conflict, for example if it breaks out along a border, or involves allies of a powerful geopolitical power, such as Russia, China or the United States.

Local communities. Civilians can be alerted to move out of harm’s way, to operationalise their own community responses to violence, assisting those in flight, or even conducting forms of mediation. As my story at the chapter’s start indicates, local actors may in fact have had even ‘earlier warning’ of the conflict than technological CEWS provide.

Military actors. Armed actors can be alerted to produce a range of military responses across quite different types of ‘army’. Responses can be framed as counter-aggression, self-defence, community defence or civilian protection. We may think of some of these responses as peaceful and some as conflict, so again, CEWS may be ‘modular’ and part of both PeaceTech and WarTech.

Are CEWS then PeaceTech or War Tech? They are focused on ‘conflict prevention’, which as noted earlier, can be an important aspect of peace process implementation. However, they are really ‘Conflict Early Warning and Response’ systems, and responses can also be military, focused on security, or war—as Nick’s story illustrated. Arguably, their origins are in SecTech or WarTech, adapted as module to peacebuilding, rather than vice versa.

5 CEWS in Practice

We have already encountered CEWS as PeaceTech in earlier chapters.

Ushahidi. Do you remember Ushahidi in Kenya, in Chap. 4?: the platform that initially began as a response to the violence around the Kenyan elections in 2008? This created a Kenyan CEWS, and now provides a soft-ware platform designed to collect and put together the information necessary to other CEWS. This platform has developed over time from ‘a simple WordPress blog with dots on a map into an entire ecosystem of software and tools built to facilitate the work done by human rights advocates, journalists, election monitors and those responding to disaster and crisis’. It now enables data collection from multiple sources: SMS, Custom embedded web surveys, email, smartphone Apps, Twitter/X, and allows bulk data imports from CSV files. This data can then be presented in a range of ways.

Hala Systems, ‘Sentry’. This system in-essence triangulates sensors in the ground, with crowd-sourced information about attacks, with tracking of flight paths of aircraft from open source material, to provide information on attacks in Syria. It employs technologies that have overlap with Ushahidi, such as smart phone Apps for reporting sightings of planes, but also uses remote-sensing, and AI algorithms to triangulate and verify information. It communicates back warning through an ‘insight portal’ with situation awareness, collaboration tools, and geospatial analytics, not just to ‘inform’ a range of actors, that they may need to take action quickly, but to ‘integrate civilians, local responders and global stakeholders’ (emphasis added).

The Sentinel Project provides a CEWS. It uses data analysis to alert people to a ‘situation of concern’. It has produced a ‘conflict tracking system’ that is publicly available (https://thesentinelproject.org/2015/02/11/the-sentinel-project-launches-conflict-tracking-system/). The focus is on identifying risk of genocide, and conducting ‘operational process monitoring’ regarding a possible evolving genocide. However, the project states an ambition for two more CEWS-type tasks: first a vulnerability assessment that would be able to identify country characteristics and all the actors within a ‘situation of concern’ to identify vulnerability to attack of key communities. Second, in the case of a predicted genocide, a prediction of severe it is likely to be (building upon information from the vulnerability assessment), and what the most likely perpetrator courses-of-action are—all designed to support responses. In practice, this work is still in a form of inception.

Strata. A ‘sister’ project at University of Edinburgh is an ‘Earth Stress Monitor’ developed with the UN Environmental Programme (UNEP). This monitors not just climate impacts, but also security vulnerability. It is not strictly an EWS or CEWS, but does use algorithms to show ‘hot spots’ where multiple risks coincide and people reside, and so points to places to pay close attention to. In marrying conflict and climate risks, the tool points to the need to understand multiple forms of conflict and climate risk together.

Beyond these examples, there are multiple diverse CEWS: this is a fast-proliferating field, often driven by research projects and non-governmental organizations. It is also a key focus of governments, and regional and international organizations often for their own intelligence and operational purposes. The following list provides links to other examples, although is not exhaustive (for good and recent overviews see Hegre, 2022; Sweijs & Teer, 2022).

Atrocity Forecasting Project (AFP), Australian National University,: aims to enhance forecasting of mass atrocities, performing in essence a ‘Research and Development’ function for CEWS (Butcher et al., 2020).

Conflict Forecast (CF), an FCDO-funded project that forecasts ‘outbreaks of political violence and escalations into internal armed conflict’ (Mueller & Rauh, 2018, 2022a, 2022b).

The Comprehensive Planning and Performance Assessment System (CPAS), a UN system for supporting adaptive management in UN Peacekeeping missions. An example of a long-term foresight tool as it supports long-term assessments as to how missions are going and how to improve them, rather than immediate ‘early warning’.

Preview, German Government, includes a ‘conflict forecasting component, see description here (Bressan, 2021).

The Violence Early-Warning System (ViEWS), provided by the Uppsala Conflict Data Programme: generates monthly predictions of the number of fatalities in impending state-based conflict 1–36 months ahead, and probabilistic assessments of conflict (Hegre et al., 2019).

UN Situational Awareness and Geospatial (SAGE), United Nations: a computer based reporting system that supports an event database including enabling UN peacebuilding and peacekeeping operations and staff to categories incidents in real time (see Druet, 2021). It is powered using the Ushahidi software, but is able to ingest a very broad amount of data including data that is structured, unstructured, quantitative, qualitative and visual.

Volatility and Risk Predictability Index (VRI), ACLED: tracks conflict surges, as part of the ACLED conflict data suite of datasets. (ACLED)

Vigil Monitor Ltd, a company that works with the XCEPT research programme to use satellite imagery to understand the drivers, dynamics and direction of conflicts, discussed more next chapter.

6 Variation in CEWS

It is interesting to note that the examples all have different approaches to:

  • the purposes and drivers of the CEWS

  • the decision-makers they seek to reach

  • the time horizons over which they seek to influence change

  • the methodologies and technologies for gathering, processing and presenting EW information

  • the geographic scope the CEWS covers, from specific countries or areas within them, regions, or ‘the world’

Let us examine further.

6.1 How Early Is the Early? Variation in Time Horizons

CEWS operate to warn on different time scales, and these link to the different ‘prevention’ tasks they aim to serve.

Nowcasting: warning about the emerging and ongoing. Ushahidi in its original incarnation, offers an example of a ‘what’s happening right now’, system. It was designed to warn people of where they should go and not go in the moment. Hala works somewhat that way too with a focus on enabling civilians to protect themselves. But of course a wider range of more structural responses to violence may also be enabled—in the case of Ushahidi’s initial work in Kenya, community responses to violence, or mothers making sure teenagers and kids did not go and join in. International Crisis Group has a Crisis Watch, conflict tracker, that it talks of as a ‘nowcasting’ tool because it tells people what is unfolding and may require a response.

Nowcasting may be so ‘now’ that it does not support timely intervention. Other purposes are also claimed for some of these now-focused CEWS, in what can seem like a form of hedging. Hala notes that its system is not only about warning but is also about ‘accountability’ after-the-fact, because it documents attack details that can potentially provide a basis for any international assessment of human rights violations and war crimes in the future. Other initiatives such as Sentinel, for example, and Vigil Monitoring that we will look at next chapter, similarly claim to really be about post-hoc accountability, although they offer a CEWS dimension.

Medium Term Adaptive management—warning about the risks of conflict onset. Other systems, aim to more link to wider preventative strategies, some that try to anticipate and produce responses aimed at further down the road, operating perhaps more at regional and organizational level. UN’s CPAS that focuses focusing on adaptive management in peacekeeping missions has more this type of function, with SAGE providing more immediate incident response-support. Interestingly, although both are UN peacekeeping innovations the systems do not operate simultaneously in all missions and seem to operate as two different platforms.

Futures horizon-scanning—warning about long-term risks and conflict trajectories. Longer-term approaches to CEWS aim to anticipate conflict dynamics down the road. This type of system shifts the methods from immediate assessments of how conflict is unfolding, or even medium-term adaptive management, to help ‘scenario plan’ as to how to address some of underlying structural drivers of conflict that may need strategies for change. This type of approach is often embedded in governmental and international systems such as SAGE, Preview and CPAS.

Do-It-All (DIA) Systems. This is my term and reflects that increasingly CEWS are being brought together with all possible tools to address instability, to try to provide all of these CEWS functions, and also support quick access to good practice guides to support responses. For example the US is developing an Instability Monitoring and Analysis Platform (IMAP) to enable policy-making in line with evidence. It aims to make open access data and technology widely available through US government departments, ‘to inform U.S. strategies, policies, and programs on conflict prevention and stabilization’. While DIA Systems sound attractive in terms of efficiencies, they run counter to the ‘do one good thing’ approach, and risk making so much data and analysis available that the right information at the right time remains difficult to access.

6.2 Who Are the ‘Decision-Makers’?

Different CEWS are used by different organizations and therefore differ with regard who they reach as ‘decision-makers’. For example, Ushahidi and Hala aim to reach ordinary citizens. However, Hala’s system also requires consideration of how information should be verified and documented for any future domestic or international criminal law process, meaning that it also responds to professional needs, say of lawyers. Some systems are specifically organizational, such as the UN’s SAGE and CPAS for peacekeeping operation support, or the US’s IMAP.

Muggah and Whitlock have drawn a distinction between ‘first mile CEWS’ that are people-centred and bottom up, and ‘last mile’ that focus on threats and top down (2022, p. 8). They use this spatial reference to indicate proximity to people affected (first mile), and proximity to international responders (last mile). These are two different sets of decision-makers, whose time needs are different, and whose response capacities will be different.

7 Digital Innovation and CEWS

Questions of purpose, timing and decision-makers to be supported, tie into the types of digital innovation that is employed.

7.1 Innovation in Data and Data Analytics

It is difficult to generalise about the data used in CEWS as they vary as examples have illustrated. Contemporary CEWS that aim to operate at scale, tend to use predictive data analytics. While statistics or ‘descriptive analytics’ tell us what a situation is, predictive analytics try to use a range of techniques of analysing data to predict what might happen next.

The data in CEWS can be quantitative data about conflict, or qualitative about structural conditions such as poverty, that we think generate conflict. It could be expert data, gleaned from reports, or from expert surveys, or crowd-sourced data, such as reporting by people, as in the Ushahidi example. Or remote-sensed data such as Hala uses. Or social media data. Amongst these data are a number of ‘old-fashioned analogue’ types of methodologies—such as asking experts—whether local or academic or political—their opinion. For example, International Crisis Group’s Crisis Watch works entirely on the basis of expert analytical assessment but visualises it on an interactive map. However, this is the exception. The multiplicity of possible data for CEWS, creates the possibility of millions of data points—in other words ‘big data’.

7.2 Innovation in Data Gathering Tools

Digital transformation has also re-shaped the ‘data gathering tools’, as has already become apparent. SMS surveys, or tools such as KoboBoxTool enable large scale survey and analysis of information, all with a level of online-offline capacity and security. Webscraping of news stories, or Twitter/X feed, all now feed into CEWS. A range of innovative tools operate within missions to enable people to log incidents in ways that automate their integration into the analysis and visualization of the CEWS. Druet’s detailed description of SAGE provides an example of just how diverse the types of data drawn on can be (Druet, 2021).

7.3 Innovation in Statistical Techniques

Also rapidly evolving is development of new analytical techniques for CEWS purposes. These are often difficult to follow and assess. Well established statistical methodologies of prediction continue to be used, the area has increasing methodological experimentation.

The Global Urban Analytics for Resilient Defence (GUARD) project of the Turing Institute, uses ‘spatial interaction’ theory, and links such as, number of people killed at major road junctions and onset of conflict, for predicting conflict. Guo, Gleditsch and Wilson, the project’s creators, have brought deaths in conflict data, such as that mentioned earlier of ACLED and UCDP, all geocoded in space, to overlay it with other quantifiable information relating to political dynamics, and set it on a map, to attempt to predict where conflict will occur (2018).

7.4 Innovation in Technology of Communication of Risk

Other digital innovations have focused on connecting the CEWS information to decision-makers. For big data and systems, these often involve dynamic interactive dashboards or interfaces, such as the platforms of SAGE, Sentinel and CPAS. Some will be built using software such as Microsoft PowerBi, others using conflict-tailored software such as Ushahidi. Knowledge foundries, discussed below, are another innovation.

8 What Does It Take for a CEWS to Work?

On 14 April 2012, when the Titanic hit an iceberg, Thomas Andrews from Ballymena Northern Ireland and who had designed the Belfast-built ship was on board. When he heard the extent of the damage, he knew categorically that the ship would sink. He also likely knew there was insufficient lifeboat capacity to save all those on board. This was data. A signal was sent immediately to nearby ships. One, the SS California was only five miles away and saw distress flares. However, it did not respond. Two inquiries—a British one and a US one—found that had it responded, all those lost could have been saved. Another ship, the Carpathia, was 58 miles away and arrived three hours later—after the Titanic had sunk. It saved the lives of over 700 people but over 1500 people had already perished in the sea.

Those in charge of the SS California either misinterpreted Titanic’s signal as not signalling imminent distress, or decided not to risk travelling through icebergs. Earlier the SS California had stopped amidst the ice field and sent a warning to other ships nearby—another EWS. However, miscommunications on the Titanic, meant that this warning did not reach the bridge, and the ship powered on to its fate—another EWS gap. Andrews, who in fact is my distant family relative, went down with the ship. The SS California was in the end sunk by a German Submarine in the First World War.

Advance information—or early warning—of icebergs existed, and was communicated at two points, once by the SS California, which if heeded could have prevented the ship’s course into an iceberg, and another by the Titanic to the Carpathia, that could have resulted in a more successful rescue operation. But the seriousness of both communications were not fully appreciated, not fully communicated to the people with decision-making power, or not taken seriously by them when it was, and so did not trigger the necessary decisions for action.

The Titanic example points to ‘EWS Gaps’. In the conflict field, similar CEWS gaps of communication and action exist, but are even more likely because conflict risks are very difficult to quantify. Also, forms of response, such as use of force, are themselves high risk and often unlikely to be used until conflict is actually unfolding, if at all. Research has also shown that ‘who’ communicates the risk in terms of how senior they are, and how much they are valued in the organization, is as important to how seriously a warning will be taken, as the reliability of the information—statisticians may carry little weight with army generals (De Meyer et al., 2019, p. 273).

For a CEWS to work as a peacebuilding measure, it would need to do the following:

  • Provide a very early indicator that violence is escalating, or is about to escalate, that is reliable and trusted

  • Get that information to people that are in a position to influence whether violence develops or not

  • Connect to those people’s mechanism of decision-making for intervening to stop the violence

  • Produce responses that are effective in preventing conflict

If any part of the chain breaks, the CEWS is unlikely to be preventative. Certain things need to be in place for each stage to be successful.

Reliable Analysis of what predicts conflict. Knowledge of what causes conflict is limited. There is consensus about factors that are linked, but no real academic consensus on how and when they combine to ‘cause conflict’. Conflict events such as killings or bombs might indicate that more conflict is on its way, but may not. Similarly, ‘disagreeing about the sovereign status of a region’ might cause conflict, or ‘poverty’ might cause conflict, but neither of these things predict conflict, because there are many instances where disputed status of a region, or poverty, do not cause conflict. Any predictive analytics capable of providing a CEWS, will find it difficult to decide what factors to explore correlations between, and what predictive conclusion to draw from them.

Even where good analysis exists, it may not be easily to combine possible relevant conflict triggers into a predictive methodology. Often political analysis of what causes conflict will have to be translated into some sort of algorithm that works with quantifiable data. A wrong calculation will lead to inaccurate prediction. Plus, one might be sceptical (I am, could you tell?!), about explanations for conflict that do not factor in the unpredictable factor of: human agency. Without going into the many deep and complex debates, it seems clear, to me at least, that some sort of complicated combination of structural conditions and the agency of armed actors, determines whether conflict results. It feels like a leap of faith, that probabilistic data based on quantifiable drivers of conflict, would produce good prediction.

Appropriate and accurate data. Knowing what to measure, and how to compute the measurement, can still leave a difficulty of what measurements to use. Even if we know what factors can be used to reliably predict violence, we need capacity to monitor and collect data on those factors. This is not simple.

Deaths in conflict, for example, would seem critical to conflict prediction in an immediate sense, and in principle deaths in conflict seem quite a certain thing to measure. However, counting such deaths is a massive, complex and even political undertaking. It requires both defining what is ‘in conflict’ (does it include the person who took a heart attack when they heard about their relative’s death or not?). It also requires a methodology for actually counting the deaths in question. For example, both ACLED and UCDP measure ‘deaths in conflict’, but they define conflict in different ways and use very different methods to ‘count’. UCDP defines conflict in terms of warring groups that have a publicly stated ‘incompatibility’ (or dispute) and uses two verified news sources to count deaths in desk-based review. ACLED counts different types of violent death—through protests, or shootings, using a network of in-country experts. This means, for example, that for UCDP cartel drug violence in Mexico is not conflict, and for ACLED it is. Both methodologies are thorough and valid, but the methodologies are different and will produce different results (see further Raleigh et al., 2023). Both may be built into CEWS, but unless transparently so, we may not know exactly how they are used to provide any predictive analytics, and therefore may not easily be able to assess how the hidden biases of either might affect the prediction.

If we move from ‘deaths in conflict’ to a range of other conflict-relevant data, such as the level of corruption in a country, the levels of relative poverty, or something like ‘the belief your situation can improve’,—all possible measures of conflict likelihood, then whether reliable data exists, or could be collected, becomes even more challenging. Moreover, apart from deaths in conflict data, most other relevant data is ‘slow data’—produced annually and tracking slow-moving processes of change. Whether data is ‘fast’ or ‘slow’ affects the type of conflict prediction it is useful for.

Good communication of risk to decision-makers that have capacity to respond. In the field of conflict, a more common gap arises in the lack of political will or capacity to respond. In Srebrenica during the conflict in Bosnia Herzegovina, what has since been called a ‘slow genocide’ occurred over months, and eventually atrocities were committed in sight of UN Peacekeepers. Yet nothing was done to intervene, due to a range of factors that constrained the political will of peacekeepers on the ground and their political bosses back ‘at home’. There was no will for engaging Peacekeepers in fighting genocidal military chiefs, or to use aerial bombing, or other forms of intervention (for an account of the unfolding of the ‘slow genocide’ see the chilling record in the International Court of Justice Application of the Convention on the Prevention and Punishment of the Crime of Genocide (Bosnia and Herzegovina v. Serbia and Montenegro), 26 February 2007).

If any one of these issues gaps arises, the CEWS will be rendered ineffective, or worse—counter-productive.

9 New Generation CEWS: Hocus-pocus Tech?

There are signs emerging of a new generation of CEWS that promise to overcome the warning-response gap by connecting data to more rapid and on occasion automated decision-making.

These are systems that claim to apply deep machine learning and Artificial Intelligence (AI) ‘to generate increasingly parsimonious assessments that are, crucially, then matched to a range of possible real-time decision options elucidating their inherent risks and payoffs’ (Muggah & Whitlock 2022, p. 4). In other words, to link automated predictive analytics with automated decision-making responses.

A range of ‘knowledge foundry’ systems now claim to be able to absorb and crunch a massive range of structured and unstructured data. For example, Palantir who are listed in NYU CIC’s ‘PeaceTech’ ecosystem mapping, are one such company, and already have provided services to the US/NATO in Afghanistan as well to multiple police forces (we return to them next chapter).

When I say ‘purport’ to do this, a key issue is that often exactly how this works is not entirely clear, including: what data is being fed in; whether there are fire walls between data given for different reasons and projects; how the algorithms are used to make determinations; how and when they trigger an automated response; and when human decision-making interacts with automated response. Customising these systems to particular contexts is also very expensive work—so much so, that Palantir have often not made a profit. Founded in 2003, its first profit-making quarter was reported in February of 2023 (Capoot, 2023). The phrase ‘connect to real time decision-making’, also rather glosses over what those decisions might be.

Further innovation appears on the cards. Palantir, for example, aim to ‘take AI to the edge’. By that they mean—off the cloud and into machines such as drones or remote-sensors. Remember the cloud/edge distinction? The push for faster analytics, means trying to embed Artificial Intelligence at the level of the machines themselves so that analysis will link even faster to automated response. This, for example, could mean having the algorithms calculated in drones, planes, submarines and satellites who communicate directly with each other to process information and do calculations and make decisions faster than cloud computing would allow.

In the words of Palantir this innovation involves a ‘“cascading chaotic process that has to be standardised”—this sounds a little scary. Taking AI to ‘the edge’ also involves stripping it down so that it can work in low-code environments with things like battery energy sources that are not limitless, so not everything is possible at the edge, that is possible in ‘the cloud’, posing the question: what might go missing and will be able to evaluate how that affects the process? (see further Palentir Edge AI).

The future direction of CEWS, for better or worse, seems to involve, innovation in:

  • Types of data collection tools and diverse types of data that can be incorporated

  • The data science of prediction

  • The ways of communicating that data back to end-users in ways that enable and support decision-making

  • Creating connected automated data collection and decision-making, in ways that are deeply decentralised and happen between pairs of machines, using ‘stripped back’ Artificial Intelligence

10 Predicting Peace—Peace Early Warning Systems?

In conclusion, what about ‘predicting peace’. It is striking how bad social scientists, peacebuilders and militaries are at this. Interestingly, they are often more doom and gloom than is warranted. There are no ‘Peace Early Warning Systems’, nor much talk of them, although Crisis Watch puts peace talks on its crisis map, and the United Nations Development Program (UNDP) ‘Social Cohesion and Reconciliation Index’ (2015), both arguably moving in this direction. Does it matter that we do not have PEWS, or are they just a nice idea but not necessary or useful? I think it matters.

If you are a civic actor in Yemen, it might be important to know what type of ‘peace’ might emerge, if Saudi Arabia negotiates secretly with Houthis, regarding the fate of your country and life? It might be important to a UN Resident Coordinator’s job, to know if the country’s landscape is going to change and he or she will suddenly be managing a peace process rather than a country office. It might be important to react quickly with forms of confidence-building measure when a ceasefire is called, or even help one on its way. But more than all that, it might be important to ‘see’ more clearly the projects of civicness in play, to understand the possibilities, constituencies and conditions that are conducive to change. Why do we map conflict events, but neglect mapping dialogue processes, or mediation efforts even after-the-fact when risk of damaging them is over, or even where peace processes are ‘active’, or might ‘break out’?

Questions

  1. 1.

    Have you had an instinct in a situation of what would happen, that you can now examine and see was rooted in knowledge and rationality?

  2. 2.

    How much do you ‘believe’ CEWS can be effective?

  3. 3.

    Which use of CEWS, nowcasting, medium or long term, seems the most useful?

  4. 4.

    Why do we map conflict and create multiple CEWS, and not map peace dialogue, or have even one PEWS?