How May I Interrupt? Linguistic Design Guidelines for Proactive In-Car Voice Assistants

Authors: Anna-Maria Meck

Abstract:

Proactivity is a sought-after feature in Voice Assistants (VA) and considered the next sensible step in growing from merely reactive to more conversational interfaces. Creating proactive use cases thereby poses a challenging design task with a multitude of potential pitfalls. Proactive interactions possibly interrupt users who are already conducting important, demanding, and potentially even security-relevant primary tasks, e.g., driving a car. So far, research has identified precise timing and careful consideration of primary task engagement as key to successful proactive interactions. While there is substantial literature on proactive VAs in terms of when to interrupt users, how to interrupt them has received less attention. Previous research has shown that VA users are susceptible to differing formulations of VA system outputs (prompts) though and that they prefer some formulations over others. Syntactical, grammatical, and lexical nuances play a role in how a VA is being perceived. To close the gap of to date insufficient linguistic-driven design guidelines for proactive VAs, a crowdsourcing study was conducted to examine users’ formulation preferences for proactive prompts in an automotive setting. Our findings show concrete syntactical best practices for formulating proactive in-car prompts, thereby allowing for the compilation of hands-on design guidelines.


Year: 2023
In session: Interaction & Dialogue
Pages: 24 to 31