Adaptive Branching Quests Based on Automated Planning and Story Arcs

Abstract

Interactive storytelling in games is a powerful tool to create immersive and engaging experiences for players. In this context, the adherence to a predefined story arc, coupled with adaptation to individual personality traits, are essential to ensure, at the same time, thematic consistency and player involvement in story-driven games. One promising way to meet such requirements is to treat plot composition as an interactive plan-generation problem, and develop a method whereby branching quests can be adequately handled and adapted in real-time. A key feature of the method is the ability to, after evaluating the effects of player decisions, perform the adaptations needed to keep the current story arc in close approximation to the predefined story arc. The underlying player preference model uses a set of artificial neural networks trained, on the basis of the players' responses to a brief Big Five personality test, to classify their preferences for specific quest decisions. This paper presents our quest adaptation method and summarizes the results we obtained through the application of our method in a fully implemented game prototype.