Publication Details

Download

Michael Braun, Florian Alt
Identifying Personality Dimensions for Characters of Digital Agents
In: Character Computing (2020). Human-Computer Interaction Series. Springer, Cham. (bib)

More and more digital services rely on natural "speech-first" user interfaces. With this trend arriving across industries, character design for digital assistants becomes relevant and along with it arises the fundamental problem of finding suitable personality dimensions. Classic personality models, like the Big Five Inventory or the MBTI, contain too many dimensions to be practicable foundations for many design use cases. This chapter introduces a method to distill use case specific personality features from user interactions with broadly diverse characters. In particular, users converse with characters inspired from popular media figures and rate their personalities as well as their user experience and fit to a certain task. As one use case, we demonstrate how fixed parameters and major dimensions for dynamic assistant personalities in an in-car environment can be identified. The method can be used to find out use case-dependent requirements to an assistant personality as well as dynamically customizable character features for personalization purposes.