Probabilistic causality in Markovian models
Session Chair: José Proença
As modern software systems control more and more aspects of our everyday lives, they grow increasingly complex. Therefore, the goal of modern IT science does not only lie in the development of powerful and versatile systems, but also in providing comprehensive techniques to understand these systems. This motivates research on formal notion of cause-effect relations in operational models that enhance the understanding why properties hold or not, and which system components are mostly responsible for the satisfaction or violation of properties.
The talk will present recent work on formal concepts for cause-effect reasoning in Markov decision processes. It will present formalizations of causality based on the probability-raising principle and related algorithms for checking cause-effect relationships and finding “good” causes for a given effect.