Crowdsourcing interactive fiction games

Procedural generation of games has become an active research field. We present a system that automatically generates an interactive fiction (IF) by learning from crowdsourced corpora of example stories. We ask crowd workers from Amazon Mechanical Turk to write short stories about a given situation with simple language, from which a plot graph is learned, containing plot events, temporal precedence and mutual exclusion relations between the events. The plot graph describes an IF where players and non-player characters choose from executable events as determined by the plot graph. We demonstrate an IF learned from the domain of bank robbery.