Integrating XAI for Predictive Conflict Analytics
Macis L., Tagliapietra M., Castelnovo A., Regoli D., Greco G., Cosentini A.C., Pisano P., Carroccetto E.
2nd World Conference on eXplainable Artificial Intelligence (xAI 2024), Valletta, Malta, 2024
We propose integrating a transformer-based Artificial Intelligence Early Warning System (AI-EWS) with Integrated Gradients, an eXplainable AI technique that attributes model predictions to specific input features over time. Applied to the ACLED geopolitical dataset, the approach forecasts conflict fatalities across 168 countries over a 12-week horizon, while providing interpretable explanations of model behaviour to support policymakers in conflict prevention and resolution.