Recognising that  coordination is a fundamental concern when both analysing
and modelling CAS, and that (ii) CAS often exhibit stochastic behaviours,
stemming from probabilistic and time-dependent local (interaction) mechanisms,
in this talk we argue that (a) measuring expressiveness of coordination
languages, and (b) predicting behaviours of stochastic systems based on
coordination models are two fundamental steps in the quest for designing well-
engineered CAS. 

As a concrete ground where to or discussion, we describe some of our current
works as well as our ideas for further research.