Skip to main content Skip to main navigation


ConceptSuperimposition: Using Conceptual Modeling Method for Explainable AI.

Wolfgang Maaß; Arturo Castellanos; Monica Chiarini Tremblay; Roman Lukyanenko; Veda C Storey
In: AAAI Spring Symposium: MAKE. AAAI Spring Symposium (AAAI SSS), Pages 1-6, CEUR Workshop Proceedings, 2022.


Many artificial intelligence (AI) applications involve the use of machine learning, which continues to evolve and address more and more complex tasks. At the same time, conceptual modeling is often applied to such real-world tasks so they can be abstracted at the right level of detail to capture and represent the requirements for the development of a useful information system to support an application. In this research, we develop a framework for progressing from human mental models of an application to machine learning models via the use of conceptual models. Based on the framework we develop a novel ConceptSuperimposition method for increasing explainability of machine learning models. We illustrate the method by applying machine learning to publicly available data from the Home Mortgage Disclosure Act database which contains the 2020 mortgage application data collected in the United States. The machine learning task is to predict whether a mortgage is approved. The results show how the explainability of machine learning applications can be improved by including domain knowledge in the form of a conceptual model that represents a mental model, instead of relying solely on algorithms. Preliminary results show that including such knowledge can help advance the explainability problem.