Automated Methods for Modeling Thinking and Reasoning Based on the Partial Model Theory

The article is devoted to the problem of modeling natural human intelligence in order to develop explainable artificial intelligence. The problems of modeling thinking, reasoning, consciousness, reflection and self-reflection are solved. For this purpose, formalization of images of perception and thinking, representations and concepts is proposed. The formal presentation of reasoning and consciousness is based on I.P. Pavlov’s studies of the first and second human signaling systems. To solve problems of explainable artificial intelligence, the integration of deep machine learning technologies and logical-semantic methods is proposed. At the same time, neural networks must implement perception, and logical-semantic methods must model reasoning. The mathematical basis of the developed methods for modeling thinking, reasoning and consciousness is the theory of partial models. Embeddings and ontological homomorphisms of partial models are used. For syntactic formal representation we use fragments of atomic diagrams.