How can we have confidence that self organizing systems actually do what we expect them to? In this position paper we overview some mechanisms at the basis of controlling and predicting the behavior of autonomous and self-organizing systems despite components’ autonomy and non-deterministic behavior. In particular we focus the analysis on the SAPERE ecosystem as an exemplary model to frame the discussion. We identify three main directions with which to gain confidence on the overall system behavior: (i) confidence from layering, (ii) confidence from large numbers, (iii) confidence form the structure and dynamics of the state space. In the paper we describe this ideas and their implication in the design of self organizing applications.