Integration of epistemically uncertain information: modelling consensus

Abstract Increasingly diverse and contrary claims as to what is fact pose challenges for information integration. Formal representation of established fact, ontology, is well studied, but that of how agents establish what they believe to be fact, epistemology, is less so. This paper argues that formal representation of multiple competing claims is required to enable systematic analysis, comparison and judgement in deciding if consensus has been established, and by whom. An epistemological framework is described for assessment of differing beliefs, permitting a view on the emergence of consensus. An agent’s beliefs may change over time, and thus consensus may also change. The framework draws on relevant theories including many possible worlds, two-dimensional (2D) semantics and four-dimensionalism. Chalmers’ 2D semantics is shown to provide a bridge between possible four-dimensional (4D) worlds and descriptions of their parts, four-dimensionalism enabling precision in identifying, aligning and comparing claimed differences and commonalities across worlds, via use of Lewis’s counterpart theory. Nesting of world beliefs is then shown to precisely capture agreement and disagreement. This information in turn is shown to enable quantifiable assessment of consensus, and possible metrics are suggested. The paper concludes with a discussion on implementation and future work towards development of a computable 4D epistemology.