Introducing Social Machines as web-enabled entities integrating social energies and computational powers into a sociotechnical system (whether purposeful or not) where social dynamics animate communities, this paper proposes a theoretical framework in which to observe them. Attempting to strike a balance between the roles of humans and nonhumans, and aware of the difficulties that this heterogeneity presents, we propose to approach the questions of capturing the social dynamics of a social machine through prosopography. Prosopography is a method, used in particular by historians, that allows to systematically study a collection of biographies, be they of persons, artefacts, infrastructures of groups thereof. Systematization is achieved through designing an appropriate questionnaire to gather homogeneous data across the biographies. Our questionnaire design relies on the identification of five archetypal elements in biographical narratives. Illustrating our method with three examples, we demonstrate how our archetypal narratives have the potential to describe at least aspects of the social dynamics in social machines.