In modeling multiagent systems for real-life problems, techniques for approximate reasoning about vague concepts and dependencies (ARVCD) are necessary. We discuss an approach to approximate reasoning based on rough sets. In particular, we present a number of basic concepts such as approximation spaces, concept approximation, rough inclusion, construction of information granules in calculi of information granules, and perception logic. The approach to ARVCD is illustrated by examples relative to interactions of agents, ontology approximation, adaptive hierarchical learning of compound concepts and skills, behavioral pattern identification, planning, conflict analysis and negotiations, and perception-based reasoning.