How many conflict deaths will there be in Sudan in 2025?
Jeremy’s most recent version of his Multi-AI Oracle is a side competitor with the machine forecasting bots of the VIEWS competition.
Obtaining outside data. Checking internet sources, using keywords: Sudan conflict deaths 2025
Results from around the web: {think}
Jeremy’s note: It’s all over the place with the Sudan question. Just not enough news data for it to reliably forecast, I think. Carolyn’s note: See Jeremy’s and Phil’s previous bot forecasts here —>
Looking at the search results, several sources mention deaths in Sudan’s conflict. Result [1] from Genocide Watch states that in Khartoum alone, over 61,000 people were killed by January 2025, with 26,000 direct war deaths. They also mention that 90% of deaths in Khartoum go unrecorded, suggesting underreporting. Result [4] from ACLED reports over 28,700 fatalities by November 2024, including 7,500 civilians. However, they note that some estimates go up to 150,000. Result [7] from Science.org cites a study estimating 61,000 deaths in Khartoum State from all causes combined by June 2024. Result [8] in Wikipedia’s Sudanese civil war page mentions that as of November 2024, at least 61,000 were killed in Khartoum State, with 8.8 million displaced. Result [10] from CSIS discusses impending famine in 2025, which could lead to starvation deaths. Additionally, [16] mentions a declared famine in Zamzam camp in August 2024.
I notice discrepancies in the numbers. For instance, ACLED’s 28,700 reported fatalities by late 2024 versus LeMonde’s 150,000 estimate. The Yale lab’s data in [7] suggests much higher numbers in Khartoum alone. It’s important to highlight these variations and the factors contributing to underreporting, like blocked humanitarian access and destroyed infrastructure.
Next, the user asked for tables. I’ll compile the data from the sources into a table with period, value, and sources. Since most data points are up to early 2025, projections for 2025 are based on current trends and warnings of famine. For the time series, available data spans from 2023 to early 2025. Calculating changes, means, and standard deviations will require organizing the data chronologically. The uncertainty in numbers due to underreporting should be noted.
Prediction market data isn’t present in the search results, so I can’t include that. The tables will focus on reported deaths, sources, and projected famine-related deaths. Key factors like RSF and SAF blockades, use of starvation as a weapon, and displacement figures are critical to mention for context.
The Sudan conflict has escalated into one of the deadliest wars of the 21st century, with 2025 projections indicating catastrophic humanitarian impacts. Below is a synthesis of key data and trends based on recent reports:
—
### **Reported Civilian Deaths (2023–2025)**
—
### **Key Trends and Projections for 2025**
1. **Humanitarian Catastrophe**
– 30 million Sudanese (65% of the population) require aid, with **5 states under famine alert** [4][10].
– Cholera outbreaks, malnutrition, and blocked aid access could drive starvation deaths to unprecedented levels, particularly in Darfur [7][10].
2. **Ethnic Violence**
– The RSF has systematically targeted non-Arab groups (e.g., 10,000–15,000 Masalit killed in El Geneina in 2023 [1]).
– Starvation, sexual violence, and forced displacement are being used as weapons of war [1][4][7].
3. **Underreporting Challenges**
– Up to **90% of deaths in Khartoum** go unrecorded due to collapsed infrastructure and restricted media access [1][7].
– Satellite imagery and ground reports reveal large-scale grave sites near Zamzam camp and El Fasher [7].
—
### **Time-Series Analysis (Direct Conflict Deaths)**
*Notes*:
– **2025 projections** account for famine-related mortality and intensified fighting in Darfur/Khartoum [10][13].
– Data gaps persist due to SAF/RSF obstruction of aid groups and journalists [1][4][7].
### **Geopolitical Context**
– **SAF Gains**: The Sudanese Armed Forces recently recaptured Khartoum’s presidential palace, but RSF retains control of Darfur [10].
– **International Inaction**: UN/AU peace efforts remain stalled, with the UAE and Egypt accused of fueling the conflict via arms shipments [1][4].
This conflict represents a severe violation of the Genocide Convention, with systemic extermination tactics employed against civilian populations. Immediate international intervention is critical to avert further mass casualties.
OpenAI gives a base rate of probability
The question difficulty is rated 8 (0 to 10)
Sufficient news to provide a good forecast? 1 (0 or 1)
News is relevant, topical and unbiased? 1 (0 or 1)
Question classification: scenario_based_forecast
Expected time (days) until question close: 274
(Carolyn’s note: insufficient data to calculate a numerical value for the base rate.)
The following were considered in order to produce this base rate:
The high base rate is grounded in the current conflict dynamics in Sudan, which have seen significant death tolls reported by various sources. Major factors include ethnic violence, humanitarian crises, and historical patterns observed in similar conflicts. These considerations, combined with reported figures and potential for further escalation, suggest a high probability of significant death tolls continuing.
Ideally, the news feed would contain the following sorts of information for a better forecast:
Access to real-time, comprehensive ground reports and a more detailed understanding of both RSF and SAF strategies would improve forecast quality. Data on humanitarian aid efficiency and local negotiation efforts would also be valuable.
Some potential divergent considerations that might effect the base rate:
A successful international intervention or peace agreement could drastically reduce the number of fatalities. Additionally, if humanitarian access improves, the situation might stabilize, leading to a decreased death toll.
The following chain of events are necessary for the question to resolve positively:
– Continued conflict between the Sudanese Armed Forces (SAF) and Rapid Support Forces (RSF), resulting in increased casualties. Likelihood: Highly likely. – Famine conditions leading to significant increases in indirect conflict-related deaths. Likelihood: Likely. – International diplomatic efforts remain ineffective, allowing conflict to persist without resolution. Likelihood: Highly likely. – Humanitarian access remains restricted, preventing accurate recording of casualties and aid delivery. Likelihood: Highly likely.
Querying Claude (AI predicts: [“Less than 1000”: 0.01, “Between 1000 and 3000”: 0.02, “Between 3000 and 5000”: 0.05, “Between 5000 and 8000”: 0.12, “Between 8000 and 12000”: 0.3, “More than 12000”: 0.5] – confidence: 6)
Querying Mistral (AI predicts: [“Less than 1000”: 0.01, “Between 1000 and 3000”: 0.05, “Between 3000 and 5000”: 0.15, “Between 5000 and 8000”: 0.25, “Between 8000 and 12000”: 0.3, “More than 12000”: 0.24] – confidence: 6)
Querying OpenAI (AI predicts: [“Less than 1000”: 0.01, “Between 1000 and 3000”: 0.05, “Between 3000 and 5000”: 0.1, “Between 5000 and 8000”: 0.2, “Between 8000 and 12000”: 0.3, “More than 12000”: 0.34] – confidence: 6)
Question Type: Multiple Choice
Confidence: 6
# LLM responses: 3
Model value:
* Less than 1000: 0.01
* Between 1000 and 3000: 0.05
* Between 3000 and 5000: 0.1
* Between 5000 and 8000: 0.2
* Between 8000 and 12000: 0.3
* More than 12000: 0.34
The reasoning behind the prediction of high death tolls in the ongoing conflict in Sudan primarily centers around the intense and persistent clashes between the Sudanese Armed Forces (SAF) and Rapid Support Forces (RSF), as well as the dire humanitarian crises, which include declared famine conditions and restricted humanitarian access. These factors, coupled with historical patterns of escalating conflicts and systematic targeting of civilians, especially in regions like Darfur, contribute to the projection of a high death toll. The predictions are supported by recent data and trends indicating significant fatalities, and monthly death rates exceeding 2,000. However, these projections could change if successful international interventions broker peace or improve humanitarian conditions, reducing indirect deaths. Furthermore, concerns about the reliability of reported figures, due to restricted access and possible underreporting, add uncertainty to the projections. Regional shifts, such as withdrawal of support by regional powers or internal exhaustion leading to de-escalation, could also potentially lower death counts.
Past forecasts on Sudan’s conflict fatalities —>
Guide to Multi-AI Oracle’s Reports (some of these only apply to older versions)
Median: We collect the output values from the LLMs in an array. We then take the median value. This is between 0 and 1 (i.e. 0.15 = 15%).
Base rate: Currently we take OpenAI’s value as the base rate. We are working on a better value using a different prompt that takes into account process. The base rate helps to determine if the median makes sense.
SD: The standard deviation between the median and the base rate.
Confidence: We query each of the LLMs on how confident they are of their predictions (between 0 and 10) and take a median. Because the LLMs tend to be overconfident, we take anything lower than 6 as being low confidence. This factors into the overall model.
Conf Mode: Based on the confidence value. >=9 is high confidence. Below 6 is low confidence (this is also triggered by an exceptionally high SD).
Mellers: This refers to Barbara Mellers, specifically a paper she wrote that includes a formula for moving values towards an extreme (i.e. 0 or 1).
Reverse Mellers: This uses the formula from above, but with a sub-1 coefficient to move the values closer to 50%.
Theory of Mind: We ask the LLMs what they think other LLMs would predict. We hope that this makes them consider the questions more deeply.
Beta Distribution: Currently unused, but possibly of interest. This is based on the median, the base rate and the SD.
Close Type: We noticed that the appropriate base case for some questions is closer to the extremes, while others are closer to 50%. When we have a low confidence value, this helps us to determine whether to extremize or de-extremize the value. ‘A’ implies closer to zero. ‘B’ implies closer to 50%. ‘C’ implies closer to 100%.
# LLM Responses: The count of LLMs that responded. We query 5 currently, but they have a relatively high API failure rate.
Model value: For normal confidence mode, this is the median. For high confidence, this is the Mellers value. For low confidence, this is either the Mellers or Revers
Background on VIEWS: Near the end of 2024, the latest forecasting competition of the Violence and Impacts Early-Warning System (VIEWS) was launched. On Dec. 4, 2024, during a telecon, a group of machine forecasting experts involved in VIEWS observed Jeremy’s demonstration of his Multi-AI Oracle. Although it was too late for Jeremy to sign up as a VIEWS competitor, they urged him to run a side competition for everyone’s benefit. Jeremy chose their Sudan conflict deaths question.
Background on the Sudan conflict, as evaluated by the team of Jeremy Lichtman and Carolyn Meinel:
(1) The importance of the Sahel, a region in Africa that includes Sudan:
Why the Sahel Actually Matters (Possibly To You, Too)
(2) April 29, 2024: Today, a Russian diplomat is negotiating gold mining with the Sudanese Minister of Minerals.
(3) June 3, 2024: Today the African Union Announced Plans to Mediate in Sudan’s Civil War
(4) Sept. 11, 2024 Over 100K starvation deaths 96% likely in Sudan by 1 Jan 2025
Carolyn has checked Sudan’s news sites below to ensure that they are still in operation. However, given the widespread violence in Sudan, any of these could go offline without warning.
https://www.alnilin.com/
https://www.alrakoba.net/
https://www.altaghyeer.info/ar/
https://www.altaghyeer.info/en/
https://www.alttahrer.com/
https://www.alzaeemsd.com/
https://www.assayha.net
https://www.cover-sd.com/
https://kushnews.net/
https://rakobanews.com
https://www.sudanakhbar.com/
https://sudanjem.com/
https://suna-sd.net/en
https://sudantimes.net/
https://www.sudantribune.net/
https://www.sudantribune.com/
https://sudaneseonline.com/
https://www.sudanile.com/
https://sudanow-magazine.net/
https://www.sudaress.com/
https://www.tagpress.net/
Below, a screenshot from VIEWS forecasting conflict fatalities in Sudan through January 2028. Below that, satellite images of the devastation in Sudan.
Below, before and after pictures of the devastation of the civil war in Sudan.
