Formal Verification of Computer Narratives

We consider modern computer narratives as deterministic finite automata with identified initial, winning, and losing states. We discuss various interesting temporal properties in the context of narratives, which range from simple reachability to more complex concepts such as the degree of “pointlessness”. These temporal properties can be useful in model checking narratives, but first an appropriate representation is needed. We introduce the Narrative Flow Graph (NFG), a class of 1-safe Petri Net, suitable for modelling computer narratives compactly, and the higher-level Programmable NFG (PNFG) language that compiles down to NFGs. We show how NFGs can be automatically encoded as transition relations for the NuSMV model checker, and use high-level knowledge from the PNFG language to reduce the number of boolean variables required. We finally provide an experimental analysis of small, medium, and large narratives.