Eliciting and Analysing Users’ Envisioned Dialogues with Perfect Voice Assistants Sarah Theres Völkel Daniel Buschek Malin Eiband
[email protected] [email protected] [email protected] LMU Munich Research Group HCI + AI, LMU Munich Munich, Germany Department of Computer Science, Munich, Germany University of Bayreuth Bayreuth, Germany Benjamin R. Cowan Heinrich Hussmann
[email protected] [email protected] University College Dublin LMU Munich Dublin, Ireland Munich, Germany ABSTRACT 1 INTRODUCTION We present a dialogue elicitation study to assess how users envision Voice assistants are ubiquitous. They are conversational agents conversations with a perfect voice assistant (VA). In an online sur- available through a number of devices such as smartphones, com- vey, N=205 participants were prompted with everyday scenarios, puters, and smart speakers [16, 68], and are widely used in a number and wrote the lines of both user and VA in dialogues that they imag- of contexts such as domestic [68] and automotive settings [9]. A ined as perfect. We analysed the dialogues with text analytics and recent analysis of more than 250,000 command logs of users in- qualitative analysis, including number of words and turns, social teracting with smart speakers [1] showed that, whilst people use aspects of conversation, implied VA capabilities, and the influence them for functional requests (e.g., playing music, switching the light of user personality. The majority envisioned dialogues with a VA on/off), they also request more social forms of interaction with as- that is interactive and not purely functional; it is smart, proactive, sistants (e.g., asking to tell a joke or a good night story).