How many state-based conflict deaths (total of all civilian and combat deaths, including both Ukrainian and Russian combatants) will be reported by ACLED in Ukraine in September 2025?
Here’s what Jeremy’s Multi-AI Oracle predicts today:
Less than 500: 10%
Between 500 and 1,000: 20%
Between 1,000 and 1,500: 30%
Between 1,500 and 2,000: 25%
Greater than 2,000: 15%
~3 minutes read
(Carolyn‘s note: The news reporting function of Perplexity failed today. However, Claude, Mistral and OpenAI, as shown below, provided credible results despite lacking the usual Perplexity news inputs.)
Querying Claude (AI predicts: [“Less than 500”: 0.15, “Between 500 and 1000”: 0.25, “Between 1000 and 1500”: 0.3, “Between 1500 and 2000”: 0.2, “Greater than 2000”: 0.1] – confidence: 4)
Querying Mistral (AI predicts: [“Less than 500”: 0.05, “Between 500 and 1000”: 0.2, “Between 1000 and 1500”: 0.35, “Between 1500 and 2000”: 0.25, “Greater than 2000”: 0.15] – confidence: 4)
Querying OpenAI (AI predicts: [“Less than 500”: 0.1, “Between 500 and 1000”: 0.2, “Between 1000 and 1500”: 0.25, “Between 1500 and 2000”: 0.25, “Greater than 2000”: 0.2] – confidence: 5)
Question Type: Multiple Choice
Confidence: 4
MAPD: Avg: 0.053333333333333, Max: 0.066666666666667
# LLM responses: 3
Explanations of the above statistical measures here —>
Model value:
* Less than 500: 10%
* Between 500 and 1,000: 20%
* Between 1,000 and 1,500: 30%
* Between 1,500 and 2,000: 25%
* Greater than 2,000: 15%
The reasoning behind the prediction of conflict-related deaths in Ukraine for September 2025 relies heavily on historical data and trends reported by ACLED, combined with an understanding of ongoing conflict dynamics. Most analyses suggest that the conflict will continue at similar intensity levels to previous months, with a probable death toll range between 1,000 and 1,500. However, some anticipate variations depending on potential developments such as unexpected ceasefires, peace negotiations, or escalations. The prediction is sensitive to changes in ACLED’s reporting methods, potential geopolitical shifts, and unforeseen developments in the conflict, all of which could lead to significant deviations from the expected figures.
Runtime: 67 seconds.
Past forecasts by Phil’s and Jeremy ’s bots —>
Jeremy’s Ukraine forecasts are a side experiment with the VIEWS competition. The VIEWS Ukraine dashboard is pictured below. (Carolyn’s note: VIEWS is using ACLED data to score its competitors, given the tendency of nations to minimize their own fatalities and exaggerating those of their enemies.)
Source: VIEWS machine forecasting competition https://data.viewsforecasting.org/
Below, from the Institute for the Study of War and American Enterprise Institute, dated Sept. 11, 2025: https://storymaps.arcgis.com/stories/36a7f6a6f5a9448496de641cf64bd375
