Counterfactual Generation considering Causal Relationships using Fuzzy Cognitive Maps

D. J. Benito Gutiérrez, J. L. Aguilar Castro, R. E. Lillo Rodriguez

Counterfactual explanations are a key tool in explainable artificial intelligence, typically framed as multi-objective optimization problems balancing validity, proximity, sparsity, and plausibility. Most heuristic methods rely on distance-based objectives and overlook the causal structure of the data. We propose a causal formulation for generating counterfactuals using Fuzzy Cognitive Maps (FCMs), which encode dynamic causal relationships. We introduce two causal objectives: causal coherence, which evaluates whether counterfactual changes propagate consistently through the causal graph, and dynamic causal influence, which measures how counterfactuals alter influence patterns during FCM inference. Experiments on synthetic and real-world datasets show that our method produces counterfactuals that satisfy causal objectives and achieve competitive proximity compared to a causal baseline while maintaining or improving causal compliance.

Keywords: Counterfactual explanations, Explainable Artificial Intelligence (XAI), Fuzzy Cognitive Maps (FCMs), Multi-objective optimization,

Machine Learning
September 2, 2026  12:40 PM
Aula 22

W. Feng, D. Wozabal, C. Ruiz

D. Rodriguez, C. Lalanne, J. Loubes

S. Rodríguez Santana