Causal Modeling of Explanations in Intelligent Systems

The subject of the research is the process of building explanations in intelligent information systems. The purpose of the research is to develop a causal approach to building explanations in intelligent systems in order to represent in the form of explanations the causes of the obtained decision, as well as the causes of actions in the decision-making process and constraints on these actions, which makes it possible to present the operation of the intelligent system in a form understandable to users. To achieve the goal of the work, the following tasks are solved: development of a causal explanation model; development of a method for building explanations based on the causal model. Conclusions. A causal explanation model with a hierarchical three-level structure is proposed, which structures the representation of explanations at global, process, and local levels in temporal and causal aspects. The developed model at the global level contains possibilistic causal dependencies between input data and the system's decision considering hidden factors, at the process level reflects causal dependencies between actions of the decision-making process based on temporal rules of «Next» and «Future» types, and at the local level sets deterministic constraints on the sequence of process actions. A temporal-causal method for building explanations is proposed, which includes five stages: identification of internal and external users, building explanations at the global level with calculation of possibility and necessity, formation of explanations at the process level through temporal rules and causal dependencies, construction of local constraints and their verification, integration of three-level explanation. Experimental verification of the method was conducted on real data of the incident management process.