Method for evaluating negative aspects of intelligent system solutions in explanation generation tasks

The subject of research is the process of evaluating negative aspects of intelligent system solutions when forming user mental models to create balanced explanations. The objective is to develop a method for evaluating negative aspects of intelligent system solutions to ensure the formation of balanced explanations based on the integration of significant negative characteristics into user mental models. Tasks: develop a hybrid approach to mental model construction that combines importance analysis and contextual analysis of solution relevance for comprehensive evaluation of negative properties of user-selected solutions; develop a method for evaluating negative aspects of user solutions in mental model construction for automated identification and organization of negative characteristics of these solutions. The following methods were used: principles of user mental model construction for intelligent systems, SHAP analysis method, sentiment analysis methods, contextual analysis methods. The following results were obtained. A hybrid approach to mental model construction based on combined evaluation of negative properties of intelligent system solutions is proposed. A method for evaluating negative aspects of intelligent system solutions is developed. Conclusions. The scientific novelty of the results obtained consists of the following. A method for evaluating negative aspects of intelligent system solutions is proposed, which includes stages of preprocessing input unstructured data, SHAP analysis of negative property importance, contextual analysis of negative properties, and formation of combined solution evaluation. The method provides the possibility for informed solution selection by users considering key negative properties and creates conditions for building understandable explanations and increasing user trust.