Reasoning About an Agent Based on Its Revision History with Missing Inputs.

In this paper, we extend on work presented in [1] where we proposed a method for reconstructing an agent's initial epistemic state from an observation on its belief revision behaviour. There, we assumed that the observation is complete in the sense that all revision inputs during the time of observation were known to us. Here, we drop this assumption and investigate the case where there are intermediate inputs we have no information about. The focus will be on determining the core belief of the agent — a belief the agent commits to at all times. [ABSTRACT FROM AUTHOR]