Composing models

We study a new composition operation on (epistemic) multiagent models and update actions that takes vocabulary extensions into account. This operation allows to represent partial observational information about a large model in a small model, where the small models can be viewed as representations of the observational power of agents, and about their powers for changing the facts of the world. Our investigation provides ways to check relevant epistemic properties on small components of large models, and our approach generalizes the use of ‘locally generated models’.