Can VUI Turn-Taking Entrain User Behaviours?
Mridumoni Phukon, Abhishek Shrivastava, Bruce Balentine · 2022
This paper presents a study where human participants converse with a voice user interface (VUI) equipped with a unique turn-taking protocol. The protocol addresses constraints intrinsic to half-duplex systems. It uses tones to sonify the invisible seam of half-duplex turn-taking, encouraging users to adjust to this unnatural seam, and thus allowing turn-taking to perform more reliably. We control the VUI's turn-taking behaviour by manipulating the speech platform's temporal turn-taking parameters that have been applied to its automatic speech recognition (ASR). By analyzing the human participant's behaviour, we show that users do respond to the artificial turn-taking behaviours. Further, over a period, the users tend to entrain to the turn-taking of the VUI.