Scéalability: Assessing Multi-Agent Storytelling Performances With Amazon’s Alexa

Good storytelling should be captivating. Some tellers immerse their audience in a suspenseful tale by enacting its key actions; others invite their audience to make suggestions as to where a story can go next. It is up to the teller what kind of narrative they chose in order to deliver their story. This thesis introduces Scéalability, a multi-agent system of cooperative actors with unique affordances that share the task of delivering a tale in different forms. To coordinate its storytelling agents, Scéalability’s blackboard architecture permits a low-dependency/high-coherence coupling of agents, ranging from embodied robots that enact a story with space and gesture to speech devices that narrate events and give voice to individual roles of characters with their own vocal features. Stories can thus be presented in different forms described by single or double act combinations of implemented agents. In addition to sketching the Scéalability architecture, this dissertation presents several empirical user studies to explore how audiences perceive and rate the interactivity and enactment features of the system in terms of usability and enjoyment of the story and narrative performances. Furthermore, it explores the value of artificial agents both as narrators and as actors, and considers how design techniques from Human-Computer Interaction (HCI) can mitigate the weaknesses of the underlying story-generation system.