Society is progressively moving towards a socio-technical ecosystem in which the physical and virtual dimensions of life are more and more intertwined and where people interaction often takes place with or mediated by machines. The scale at which this is happening and the differences in culture, language and interests makes the problem of establishing effective communication and coordinated action increasingly challenging.

So far, the attention has been mainly devoted to collective adaptive systems (CAS) that provide or impose some form of harmonization or lightweight coordination of meaning and actions where machines do most of the computation and humans are at the periphery and only act as consumers. In these settings, collectives are generally homogeneous and are orchestrated to achieve a common global goal where knowledge adaptation only takes place locally to a single unit.

Our goal is to move towards a hybrid system where people and machines tightly work together to build a smarter society. We envision a new generation of CAS centred on the two foundational notions of compositionality and diversity where humans and machines “compose” by synergically complement each other thus bridging the semantic gap between low-level machine and high-level human interpretation of data and where they interoperate collectively to achieve their possibly conflicting goals both at individual and societal levels. Operationally, peers in the system will implement a continuous unlimited cycle in which data is sensed, interpreted, shared, elaborated and acted upon. Actions are taken on the basis of system suggestions and the way humans react to them. Actions generate new data thus alimenting the cycle ad infinitum.

To meet this very ambitious goal these systems should be hybrid, distributed, open and large scale where multitudes of heterogeneous peers producing and handling massive amounts of data can join/leave dynamically following unpredictable patterns with no central coordination and interoperating at different spatial and temporal scales. By identifying the right incentive schemes and privacy levels, these systems should assist humans in their everyday activities, be able to cope with the diversity of the world in terms of language, knowledge and personal experience and to work in presence of possibly imperfect information.