Sudan updates from AI Botmaster Jeremy Lichtman of April 24, 2025

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.

Here’s what Jeremy‘s latest version of his Multi-AI Oracle predicts: (Carolyn’s note: Both Jeremy and I agree, these probabilities are, so far, just an experiment. It is too early to know how good these will turn out to be)
* Less than 1000: 1%
* Between 1000 and 3000: 5%
* Between 3000 and 5000: 9%
* Between 5000 and 8000: 13%
* Between 8000 and 12000: 27%
* More than 12000: 45%

Obtaining outside data.
Checking internet sources, using keywords: Sudan conflict deaths 2025
Results from around the web:

Okay, let’s tackle this query about the Sudan conflict deaths in 2025. The user wants current web results summarized, including any relevant data tables and a time series if possible. They also want URLs listed at the end.
First, I’ll check the provided search results. The most recent one is from March 2025 (source [5]), which is within 2025. The others are from December 2024, February 2025, and January 2025. Since the current date is April 2025, these are the latest available.

Looking at the numbers: source [1] mentions over 28,700 reported fatalities by November 2024, including 7,500 civilians. Source [2] from February 2025 also cites 28,700 deaths with 7,500 from intentional injuries. Source [3] mentions up to 150,000 killed, and source [4] notes 61,000 deaths in Khartoum alone, with 90% unrecorded. The WHO report (source [5]) doesn’t give direct death counts but focuses on health crises and displacement.

The challenge is reconciling the varying estimates. ACLED and Science.org give around 28,700, while CFR and Genocide Watch cite much higher numbers. The WHO report supports that the situation is dire but doesn’t provide exact death tolls. I need to present these discrepancies and possible reasons, like underreporting.

For tables, the user requested data in tabbed or CSV format. The direct numbers from sources [1] and [2] can be tabulated. The time series is tricky because the data points are from different months but not a continuous series. Maybe show the different estimates with their sources and dates.

I should also mention the humanitarian impact from source [5], like displacement and malnutrition, as context for the death toll’s significance. Finally, list all the URLs used. Make sure citations are correctly numbered and attached to each statistic.

The Sudan conflict has escalated into one of the deadliest crises of the 21st century, with staggering human losses and widespread devastation. Key findings from recent reports include:

## Reported Fatalities and Displacement
– **ACLED** recorded over **28,700 reported fatalities** by November 2024, including **7,500 civilians killed in direct attacks**[1][2].
– **LeMonde** estimated at least **150,000 civilian deaths** by November 2024, with **61,000 killed in Khartoum alone** (26,000 directly from warfare)[4].
– Over **12.8 million people** are forcibly displaced, creating the world’s largest displacement crisis[5].

## Humanitarian and Health Crisis
– **30.4 million people** (over half the population) require humanitarian aid, including **16 million children**[5].
– Famine conditions are confirmed in parts of North Darfur, with **4.9 million children and pregnant women acutely malnourished**[5].
– Attacks on healthcare infrastructure have severely limited access to medical services, exacerbating outbreaks of cholera, malaria, and measles[5].

## Conflict Dynamics and Atrocities
– Both the **Sudanese Armed Forces (SAF)** and **Rapid Support Forces (RSF)** are accused of **war crimes**, including mass rape, torture, and using starvation as a weapon[2][4].
– The RSF faces charges of **genocide against non-Arab ethnic groups** in Darfur, echoing atrocities from the early 2000s[2][4].

### Reported Death Toll Estimates

*Note: Discrepancies arise from differing methodologies, with ACLED tracking direct violence and others including indirect deaths (e.g., starvation, disease).*

**Sources Used**
1. https://acleddata.com/conflict-watchlist-2025/sudan/
2. https://www.science.org/content/article/how-many-have-died-sudan-s-civil-war-satellite-images-and-models-offer-clues
3. https://www.cfr.org/global-conflict-tracker/conflict/power-struggle-sudan
4. https://www.genocidewatch.com/single-post/sudan-genocide-emergency-january-2025
5. https://www.who.int/publications/m/item/public-health-situation-analysis–sudan-conflict-(10-march-2025)

OpenAI gives a base rate of fraction of probability:
The question difficulty is rated 9 (0 to 10)
Historical weighted factors include:
0.8
0.7
0.6
A Bayesian calculation could be performed as follows:
Using a weighted sum of historical conflict data and factoring in the current situation: P(Interest|News) = [P(News|Interest)*P(Interest)]/P(News). With high recent reports of death it updates the prior upwards.
Bayesian base rate: 0.88
Sufficient news to provide a good forecast? 1 (0 or 1)
News is relevant, topical and unbiased? 1 (0 or 1)
Question classification: reference_class
Expected time (days) until question close: 180

The following were considered in order to produce this base rate:
The base rate was established by considering the conflict’s intensity, historical data from previous conflicts, and the health crisis impacting mortality. ACLED data provides a baseline, corroborated by humanitarian reports from WHO highlighting worsening conditions.

Ideally, the news feed would contain the following sorts of information for a better forecast:
Access to more recent, independent ground reports or satellite imagery data might help refine estimates. Information from NGOs actively working in affected areas can provide more granular insights.

Some potential divergent considerations that might affect the base rate:
Possibility of international intervention or peace negotiations which could drastically reduce the conflict intensity and thus lower the death toll. Underreporting and difficulty in data verification could mean actual figures remain uncertain.

The following chain of events are necessary for the question to resolve positively: (Carolyn’s note: the Multi-AI Oracle was confused.)
Very high likelihood – High likelihood – High likelihood – High likelihood

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”: 0.1, “Between 5000 and 8000”: 0.24, “Between 8000 and 12000”: 0.3, “More than 12000”: 0.3] – 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”: 0.14, “Between 8000 and 12000”: 0.2, “More than 12000”: 0.5] – confidence: 7)

Question Type: Multiple Choice
Confidence: 6
MAPD: Avg: 0.055555555555556, Max: 0.13333333333333
# LLM responses: 3

Model value:
* Less than 1000: 1%
* Between 1000 and 3000: 5%
* Between 3000 and 5000: 9%
* Between 5000 and 8000: 13%
* Between 8000 and 12000: 27%
* More than 12000: 45%

The various AI analyses highlight a grave humanitarian crisis in Sudan with estimates of fatalities ranging from over 28,700 to 150,000 or higher due to the ongoing conflict, which is causing direct and indirect deaths. The conflict shows no clear signs of resolution, with widespread violence, health crises, and displacement exacerbating the situation. Historical patterns from similar conflicts suggest that mortality rates may remain high or increase. However, there are factors that could potentially mitigate the death toll, such as international intervention, successful peace negotiations, or improvements in humanitarian aid delivery. Additionally, uncertainties in reporting, such as underreporting or misclassification of deaths, could result in a range of outcomes where the actual death toll might be lower than predicted.

Past forecasts on Sudan’s conflict fatalities —>

Guide to Multi-AI Oracle’s Reports (some of these only apply to older versions)
Bayesian calculation: a mathematical technique that helps determine the conditional probability of an event based on prior knowledge and new evidence.
MAPD: Mean Absolute Percentage Deviation calculates the average percentage difference between actual values and predicted values, providing a relative measure of error.
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 Below, map of control in Sudan’s civil war as of April 15, 2025.

Below, before and after pictures of the devastation of the civil war in Sudan.

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/