A Transition System for Causality and Strategic Responsibility

Abstract

Causality plays an important role in many day-to-day processes and humans reason about causality all the time. There has however not been much research on causality in (multi-)agent systems. In this work I introduce a way to integrate a structural causal model in multi-agent system models, specifically in a concurrent game structure (CGS). In such a causal CGS, every transition corresponds to an intervention on agent variables of the causal model. The Halpern and Pearl framework of causality is used to determine the effects of a certain value for an agent variable on other variables. This causal CGS allows us to analyse and reason about causal effects of agents’ actions on each other and their shared environment in multi-agent settings. In this work I study and analyse the relation between the derived-multi-agent system model, this causal CGS, and the original structural causal model and I analyse the causal CGS to show a relation between strategic responsibility and causality. This work first gives an overview of the literature on causality, labelled transition systems and concurrent game structures, several temporal logics and several notions of responsibility. After that it defines the notion of a causal CGS, and I show how agent strategies in this CGS relate to causality in the original causal model and show a relation between strategic responsibility and causality in the HP framework. This work can in the future be used for (multi-)agent systems where causal relations play an important role. It can be used to plan strategies or to determine causation and responsibility when things go wrong.