Within the field of multiagent systems, the Belief-Desire-Intention
model of practical reasoning has been highly successfully and has over
the years become the de facto standard for the modelling and design of
deliberative agents. However, despite the fact that interaction between
agents is as important as the design of individual agents in multiagen
systems research, the field has so far failed to come up with an
equally generic model of practical social reasoning that could be used
to view existing methods for reasoning about interaction in a unified
way. In this talk we discuss initial work on such a model which we call
the Expectation-Strategy-Behaviour architecture and which is based on
specifying expectations and the way agents can use those to coordinate
their behaviour with that of others. We discuss its relationship to the
BDI model, present examples of applying its principles from the area of
agent communication, and present a number of promising research avenues
for the future.