Fable Reborn: Investigating Gameplay Experience between a Human Player and a Virtual Dungeon Master

Abstract

Dungeons & Dragons (D&D) is a role-playing game typically played face-to-face with participants seated around a tabletop. Online technologies have enabled the conduct of live sessions on virtual platforms, allowing players to continue enjoying the game despite geographic distances. Playing D&D relies heavily on a dungeon master who is tasked with preparing the fictional world and facilitating the interaction between players who each control a single story character. In this paper, we investigate the impact of employing AI techniques in designing a virtual dungeon master (VDM) on the gameplay experience of D&D players. The VDM utilizes Open AI's GPT2 model to generate story prompts in combination with an input plausibility checker, a dice rolling feature, and a candidate story prompt scorer. Feedback from D&D players revealed that the VDM is capable of producing believable fantasy stories but lacks consistency in the quality of its generated stories. Our findings point to the potential abilities of an artificially intelligent agent in engaging human D&D players in role-playing games; however, the VDM is still not capable of truly replicating the experience of playing with a human DM.