Dealing with Input Uncertainty in Automotive Voice Assistants

Abstract

After their success in the smart home, voice assistants are becoming increasingly popular in automotive user interfaces. These voice assistants are traditionally designed to provide a human-like dialog with the user. Thus, when processing voice input, especially dealing with uncertainty is an important factor that needs to be considered when designing system responses. While state-of-the-art voice assistants offer responses based on their certainty of what they understood, these response-thresholds are largely under-explored. In this work, we close this gap by providing a user-centered approach to investigate which responses are acceptable for voice input users depending on input certainty. Through findings from semi-structured online interviews with 101 participants, we provide insights about designing voice user interface responses based on system certainty. Our findings reveal a sweet spot for executing a task versus requesting additional user input. Further, we provide data-driven guidelines for different in-car voice assistant behaviors.