In this position paper, we argue that a new methodological paradigm and software platform is required for developing citizen-oriented social computing applications. The platform that we propose is based on the idea of 'Collective Intelligence as a Service', and is grounded in the formalisation of computational models derived from an empirical analysis of psychological processes and social practices. This in turn provides the enablers for developing radically innovative tools for computational sustainability and computer-supported collective action in smart communities (e.g. Smart cities). Our vision is illustrated with two exemplars, one a healthcare application for patients with peripheral arterial disease, and the other an application for collaborative energy conservation to meet targets set out in a city's Sustainable Energy Action Plan.