On the Problem of Explaining the Results of Intelligent Data Analysis

The possibilities of evaluating the acceptability of results generated by artificial intelligence (AI) systems in the process of intelligent data analysis (IDA) are evaluated. The explanations generated by the IDA system are one of the convenient tools for such an evaluation. Some known approaches to the generation of explanations are considered. Particular attention is paid to causal explanations, which are based on empirical causal dependences extracted from data in the IDA process. Some algorithmic properties of causal explanations are given.