A computer-based model for collaborative narrative generation

This paper describes MEXICA-impro, a computational model for collective plot generation. It comprises two agents that work together in order to generate a narrative. Each contributor has its own representation of common-sense knowledge and social beliefs. I present results showing how MEXICA-impro can produce storylines that could not be produced by any of the participants alone, and how such stories generate new knowledge structures that might be employed by the collaborators to progress original plots. These results suggest that MEXICA-impro is capable of developing what I refer to as collectively-creative stories. We have established a solid framework that we hope will be the basis for extensive further research.