For future generations of autonomic systems it is becoming increasingly evident that pre-programmed models of behaviour and communication will be inadequate. The environment, in which the systems and their autonomic components are situated, is most likely to be the real world and so open-ended. It is not possible to foresee all system eventualities or actions, that may be needed to be performed, nor the methods by which novel observations may be detected or communicated. Additionally these systems are going to have to adapt to human requirements and so develop shared conventions of communication between themselves and human users in order to cope with novel meanings and requests from unknown sources. This paper proposes the use of cognitive systems reasoning over models of self-organising emergent behaviour as one method to address this problem. The actions of individual participants in the system form Markov Chains, which allows the autonomic observer systems to gain, or be endowed with, a grounded definition of some emergent behaviour. Furthermore the parameters of the associated Markov Decision Problem can be adjusted to the required outcome. It is further proposed that mathematical logic is the most natural specification medium for these cognitive systems. Thus a dialect of First Order Logic, Situation Calculus, is described with facilities to handle sensing and knowledge and stochastic actions. This is then shown as a specification for a known observed, naturally occurring emergent phenomenon that can be solved as a Markov Decision Problem.