Early warning and fundamentals
With all things international in free fall, I would like to take aim at a sacred cow. Specifically, I think it is time — as a community — to stop pretending we accurately predict conflict. Our focus should be on monitoring, diagnosing, warning, and responding, as we can do all of these well.
Prediction is a big beast in the conflict analysis industry. I distinguish it from early warning and “anticipation” in its time horizons, with “warning” reflecting short-term, evidence-supported patterns (as we do in our Conflict Alert System (CAST) and monthly reports). Prediction’s scope is to uncover conflict patterns between one, five, and even 30 years into the future, and often for the onset of “new wars” rather than conflict evolution. Its popularity surges with uncertainty, and with hopes that we can automate and AI it into reality.
My concerns about it are that:
a) No one has ever predicted anything surprising about actual conflict zones or rates. The novelty is the model, and despite a pretty healthy respect for quantitative methods, constant references to new models or methods as the keys to unlock poor out-of-sample prediction results is a horse that has been over-flogged. No model — especially those using automated or media-tone data — can improve upon a casual observer’s solid guess that Pakistan and India will fight soon, where the timing will be dependent on elements internal to both countries. More useful to planners and responders would be to assess “Where is India likely to attack when it does? How? And what level of counter-attack will be deemed sufficient?”
b) Prediction models have overtaken knowing anything about conflict. I am not against models; I am fervently against novelty as a substitute for knowledge. Predicted results can’t stand alone because they are a poor manifestation of reality. Accurate timing is what most prediction models are trying to capture — and that really is a fool’s errand. For example, for close to 30 years, Zimbabwe’s former President Robert Mugabe was reliably predicted to be constantly “dying” or “overthrown” — and this was in a very consistent political environment which was eventually overturned by a poisoned ice cream cone given to the current president (and the fact that Mugabe would eventually die). Relatedly…
c) Their purpose. A cynic might say that prediction has been popular because it is a good way to deflect. If the focus is on “predicting out” conflict patterns five or 30 years from now, then no one will ask why they don’t know what is going to happen next week. For a leader of an international organization or a state’s foreign ministry, current concerns are how to prioritize resources immediately, how to domestically adapt to the changing world order, and how to anticipate the immediate reaction of enemy actors to important assets in the next six months to a year. Everything else is gambling.
To test my doubts, I will often ask conflict predictors two questions: first, “What causes conflict?” and the second, “Is the US going to attack Iran?” Answering either requires an analyst to assess their fundamentals. But models don’t have fundamentals. Debating how states or armed groups enact an agenda, sustain funding, or shape their larger trajectory can be clarifying and productive — these are the components of patterns we can model and apply. My point is that you need to know something about conflict and countries before trying to model these upsets.
The sole basis of effective and useful warning comes from monitoring conflict reliably, effectively, and flexibly, and explaining patterns you are seeing. There are no crystal balls. You can be innovative in how you monitor and in how you discover patterns (whose characteristics can be modelled for consistency, deviations, and longevity, etc.), but warning is down to whether someone’s ears and eyes are receptive to a well-evidenced message you have crafted. It is a give and take.
So, let’s discuss the fundamentals: These are the elements of a conflict you think are driving violence patterns. Every conflict analyst has their own fundamentals, and mine are largely about the political and conflict environment. They involve interrogating how armed groups operate — their conflict procedure. They include:
- Which direct and indirect security risks are present? (e.g., has someone important just been fired? Is an election coming?)
- How will/could they manifest? (e.g., links between elites and armed groups; are mutinies likely?)
- Who and what controls the direction and cycle of risks? (e.g., have specific political parties been associated with militias in key regions? Who is making decisions?)
- What is their agenda, and can they implement it? (e.g., how are groups forming and what is their scope and scale?)
- What obstacles and incentives exist once violence has begun? (e.g., where does the state have security capacity? Which local elites may turn? Where are other armed groups operating?)
- What will be affected (relationships, elections, people, business, etc.) intentionally and unintentionally? And, of course, how does each of these affect a group’s (or combination of groups’) conflict procedure?
These questions are an explicit rejection of what is not fundamental to understanding future actions: I don’t think “why” is important because the given logic is either much too large (e.g., “ideology”) or much too small (i.e., we have strip-mined individual and collective grievance logic for too long). For me, the “tails” matter — the small movements, assassinations, or defections — to understand how low-frequency warnings derail high-frequency occurrences. You need people from these places to explain how small moves lead to big changes. Why bring this up now? Because given the level of disruption to the international system in the very recent past, we have an opportunity to reset our priorities. The international system is unmoored, conflicts are shifting, and no amount of pretending or modelling otherwise changes this. Models cannot deal with choices in a political environment that is both increasingly decentralized and totalitarian, coupled with simultaneous and unaligned conflicts. Things are changing, and we need to make sure we adapt. Holding onto old certainties will not help us understand what is coming.
Webinars
Two years of war in Sudan: How the SAF is gaining the upper hand
On 17 April 2025, ACLED hosted a webinar to discuss the findings of its report marking two years of war in Sudan and the important shifts that led to the Sudanese Armed Forces’ advances in the war and the impact this new phase of the war will have on the dynamics of the conflict. It was moderated by me and featured speakers ACLED’s Nohad Eltayeb and Ali Mahmoud Ali. You can view the webinar recording here.
Rwanda in Mozambique: Limits to civilian protection
On 24 April 2025, ACLED hosted a webinar to launch its report examining the impacts of the Rwanda Defence Force’s shift in posture toward civilians in Cabo Delgado. In the webinar, which was moderated by ACLED Senior Analyst for Africa Ladd Serwat, we heard from the report’s author, ACLED Southeast Africa Senior Analyst Peter Bofin, and Zitmar News journalist Mariam Umarji. You can view the webinar recording here.
Select ACLED in the media!
- ACLED’s Luca Nevola’s insights on the Houthis were featured in The New York Times and MSNBC. Ameneh Mehvar was quoted in Arab News. Nasser Khdour was quoted by Reuters on Turkish demonstrations.
- Kieran Doyle commented on declining rates of US disorder in Slate and Huffington Post.
- Al Jazeera featured ACLED’s data in its article “Animated maps show two years of war in Sudan” and quoted our own Muaz al-Abdullah on Syria.
- Our data was also featured in The Wall Street Journal’s article “The Brutal War Complicating Myanmar’s Quake Response.”
Notes and notions
Check this out: Opinion | To Understand Global Migration, You Have to See It First. These migration estimates were drawn from location data of 3 billion Facebook users and have extraordinary detail.
Scorn as you wish, but this discussion with Secretary Rubio was interesting.
Even the most dedicated jobsmith (or, DOGE) would be put to shame by the administrative structure and policies of jihadi groups. Here is the great Elizabeth Kendall mentioning a public document about an al-Qaeda leader in Yemen, who (as she notes) is in trouble, but the reaction of the organization is so interesting: “#AQAP does not often choose to wash its dirty laundry in public,” and this “implies concern to keep local opinion on side.” AQAP – just like us.
A lot of China material is going around these days, including the very well-produced podcast “To Catch a Thief” as well as a detailed discussion about the perils of underestimating the US-China codependent relationship. A final piece about “The Eight Tribes of Trump and China” considers the way that the China debate has divided those in and around the new administration. Well worth your time.
I have very little “Irish news,” only to note that because football seems to be constantly on in my house, we caught (Irish-ish person) Declan Rice (of Arsenal) scoring two goals (2!) against the brilliant Real Madrid (who were playing Mbappé!), and the look on his face told the world this was the best night of his life! I could take or leave football, to be honest, but sport does allow you to see some real high points in the lives of complete strangers, and that is lovely.
While recently looking at an old global map made in the Mercator projection, I thought perhaps this whole fuss in the United States about Greenland could be because President Trump may have only seen maps made in this projection, and therefore assumes that Greenland is huge (which it isn’t, of course).
