Efficient Total Causal Effect Computation in Fuzzy Cognitive Maps for Scalable Explainable Artificial Intelligence

Building trust in AI systems requires transparent and interpretable decision-making, underscoring the importance of eXplainable AI (XAI). Fuzzy Cognitive Maps (FCMs) are a promising XAI method because they combine an intuitive causal structure with multilevel interpretability and hybrid intelligence capabilities. However, the traditional approach to causal effect analysis for globally interpreting FCMs is computationally inten sive and impractical, even for moderately sized networks. To ad dress this limitation, this letter introduces the Total Causal Effect Calculation for FCMs (TCEC-FCM) algorithm, which enables the efficient computation of total causal effects in large-scale FCMs for the first time. By combining binary search and graph traversal techniques, TCEC-FCM avoids the exhaustive exploration of causal paths and substantially reduces computational overhead. Experiments on synthetic FCMs with different sizes and densities demonstrate that the algorithm effectively overcomes key scala bility challenges, thereby extending the applicability of FCMs to large-scale XAI scenarios.