Narrative structures for multi-agent interaction

Abstract

Multi-agent interaction involves extended exchanges of asynchronous messages among the conversing agents and requires supporting technology in order to manage and keep track of the progress of the conversations. We outline our current work that extends our previous efforts in the area of improved agent interaction protocols by incorporating techniques inspired by narrativity analysis. We introduce in this context the notion of narrative nets, a representation based on coloured Petri nets that models progress in the current narratively-based concerns of the communicating agent. The basic ideas behind narrative nets are presented and an example system modelled after commodities trading is discussed.