Opportunities for Machine Learning to Impact Interactive Narrative

An interactive narrative is an education, training, or entertainment experience, based on a narrative that is guided by authorial intent. Interactive narratives have two important characteristics: some form of authorial intent as represented by the narrative structure and an affordance for autonomy of player characters. Technologies for constructing engaging interactive narratives have been springing up more and more rapidly in recent years (see [4] for a recent survey). As this field gains momentum, new and exciting opportunities arise for the application of novel machine learning methods. In this paper, we will present a brief overview of interactive narrative, formalizing the problem in a manner amenable to computational inquiry. Then, we will survey recent work based on the use of statistical machine learning methods to solve various aspects of interactive narrative problem. We will also describe certain aspects of interactive narrative that present new and exciting challenges for the machine learning community such as non-stationarity and new definitions of optimality. Lastly, we will describe the aspects of the interactive narrative problem that we feel will both benefit the most from the application of machine learning methods as well as provide some interesting challenges for the machine learning community.