Emotion Discourse as Design Heuristic: Creating Emotional Intelligence for Virtual Narrative Agents

A pressing task for researchers who seek to build systems capable of emulating intelligent narrative agents is to develop a basis for ascribing situationally appropriate (and thus believable) emotional states to participants in storyworlds. In this paper, we use a natural-language narrative to examine discourse processes by which storytellers in face-to-face communication construct themselves as emotionally involved agents in narrated events, arguing that closer study of these processes may afford heuristics for system design. Drawing on techniques for analyzing emotion discourse developed in the field of discursive psychology, we describe how narratives both recruit from and contribute to emotionologies, or systems of emotion terms and concepts deployed by participants in discourse to ascribe emotions to themselves as well as their cohorts. Studying how such terms and concepts figure in the discourse of expert storytellers can provide a basis for enhancing the emotional intelligence of virtual narrative agents, while suggesting how end-users’ emotional states might also be factored into system design.