Methodological Aspects on Integrating Fuzzy Systems with Explainable Artificial Intelligence

Today we have increasingly more sophisticated applications of artificial intelligence. However, they have also aroused ethical and legal problems because their models are often opaque or too complicated by nature. Hence, the European Union and United Nations, among others, are in the process of drafting new directives and regulations for these models by presupposing that they should be more transparent and explainable in the future. In brief, we should apply explainable artificial intelligence instead. We will suggest that the fuzzy systems may respond well to these challenges because they are operating with the human-like linguistic and approximate entities and reasoning. We may thus construct transparent and explainable models effortlessly. The criteria for explainable artificial intelligence are at first presented. Then, the basic principles for constructing the typical fuzzy models are considered at a general methodological level. The justifications for applying these models to explainable artificial intelligence are also provided.