A Precise Semantics for Vague Diagrams

Informal vague causal diagrams (VCDs) are a common technique for illustrating and sharing expert intuitions. Normally, VCDs are viewed as precursors to other modelling techniques which necessitates further knowledge acquisition. Here we explore what semantics can be granted to VCDs, without having to request more information from the expert(s) or the domain. The impreciseness of VCDs typically makes them indeterminate. VCD inferencing must assume multiple possibilities and manage mutually exclusive possibilities in separate worlds. Given a library of known behaviour of the entity being modelled, we can use exhaustive abduction over VCDs to prove what behaviours are categorically impossible; i.e. we can use VCDs for knowledge acquisition.