Dynamic Distributed Nonmonotonic Multi-Context Systems∗

Nonmonotonic multi-context systems (MCS) provide a formalism to represent knowledge exchange between heterogeneous and possibly nonmonotonic knowledge bases (contexts). Recent advancements to evaluate MCS semantics (given in terms of so-called equilibria) enable their application to realistic and fully distributed scenarios of knowledge exchange. However, the current MCS formalism cannot handle open environments, i.e., when knowledge sources and their contents may change over time and are not known a priori. To improve on this aspect, we develop Dynamic Nonmonotonic Multi-Context Systems, which consist of schematic contexts that allow to leave part of the information interlinkage open at design time. A concrete interlinking is established by a configuration step at run time, where concrete contexts and information imports between them are fixed. We formally develop a corresponding extension and provide semantics by instantiation to ordinary MCS. Furthermore, we develop a basic distributed configuration algorithm and discuss several refinements that affect the resulting configurations, in particular by means of optimizations according to different quality criteria. This discussion is complemented with experimental results obtained with a corresponding prototype implementation.