Virtual Group Knowledge on Topological Evidence Models
Djanira dos Santos Gomes

Abstract:
In this thesis, we study notions of group knowledge and group belief in topological-evidential semantics. We define a multi-agent extension of the topological evidence models defined in [Özg17]. On these models, we present the evidence-based semantics of virtual group knowledge: this extends from individuals to groups the topological semantics of evidence-based belief and fallible knowledge, using the join topology (generated by the union of the individual topologies). As a result, group knowledge is non-monotonic with respect to inclusion. In contrast to distributed knowledge, this notion of virtual group knowledge is more realistic, since it pre-encodes the knowledge that a group can actually achieve after sharing all the individual evidence. We extend the language with a corresponding dynamic operator, which models the intented act of communication. We axiomatise the multi-agent language of evidence, as well as a knowledge-belief fragment of the language that restricts knowledge and belief to individual agents and the full group of agents. We bring these notions into practice by implementing symbolic and explicit approaches to Haskell-based model checking of the language of evidence. We compare the resulting model checkers with respect to performance.