Computational sociology models social phenomena using the concepts of emergence and downward causation. But the theoretical status of these concepts is ambiguous; they suppose too much ontology and are invoked by two opposed sociological stands, namely, individualistic and holistic interpretations of social phenomena. In this paper, we propose a theoretical alternative that not only might clarify those concepts, but also keep their heuristic and interpretative value for computational sociology. We do so by advancing two proposals. Firstly, we suggest a non-ontological framework that allows modellers to identify emergent processes. This framework asserts the macro level and micro level as the emergent by-products of an instrumental prompting (the very modellers’ act of distinguishing). Secondly, in order to support analytically the modellers’ simulations, we link this non-ontological framework with the theory of self-referential social systems. This theory gives an account of the emergence of the social realm from the bottom-up as communication and describes the process by which society limits the possible selections of individuals. These two proposals are well-positioned to overcome some epistemological drawbacks, although they also generate new challenges to computational sociology.