Part of the power of social computation comes from using the collective intelligence of humans to tame the aggregate uncertainty of (otherwise) low veracity data obtained from human and automated sources. We have witnessed a surge in development of social computing systems but, ironically, there have been few attempts to generalise across this activity so that creation of the underlying mechanisms themselves can be made more social. We describe a method for achieving this by standardising patterns of social computation via lightweight formal specifications (we call these social artifacts) that can be connected to existing internet architectures via a single model of computation. Upon this framework we build a mechanism for extracting provenance meta-data across social computations.