Probabilistic Models for Detection of Causal Relationships in Data Sequences

Some possibilities of using causal dependencies in AI applications are discussed. Such causal dependencies are usually initially hidden in empirical data, however, they can be recovered from the available data in the process of appropriate analysis. It is shown how explanations formed on the basis of causal dependencies recovered from the analyzed data can be used to increase confidence in the conclusions and results generated by AI systems. Some examples of popular applications are given where the use of cause-and-effect dependencies implicitly presented in the analyzed data makes it possible to enforce the success of using appropriate AI systems. The simplest case of identifying deterministic cause-and-effect relationships in the presence of random properties that do not carry information about the analyzed causes and effects is considered.