Interaction between autonomous computational entitites in a common environment is arguably the most distinguishing feature of agents and
multiagent systems research. To achieve efficient and effective automation of such interaction, agents must be able to reason about the interactions
they engage in at runtime, and to make rational decisions about their own behaviour, ideally given high-level specifications of their task environment.
In this talk, I will present three different lines of research that we are currently pursuing in Edinburgh, and which follow this spirit by proposing methods
for automated rational reasoning at the macro (automated norm synthesis), meso (argumentation-based conflict resolution) and micro (practical social
reasoning architctures) level. I will also talk about more recent work on trying to unify these different subproblems in a common framework, and discuss
the challenges and potential benefits of embarking on such an enterprise.