aalto1 untyped-item.component.html

LLM-based voice assistants in vehicles: The interplay of response time, driving task and interaction task on user acceptance

Loading...
Thumbnail Image

URL

Journal Title

Journal ISSN

Volume Title

School of Science | Master's thesis

Department

Mcode

Language

en

Pages

65

Series

Abstract

AI-enabled voice assistants (VAs) are increasingly integrated into vehicles, allowing more natural human-machine-interaction. However, little is known about whether users hold the same performance expectations for AI-enabled VAs as they do for traditional systems. Here, system response time (SRT) is crucial as poorly timed responses can disrupt the conversational flow. In a simulator study, the interplay between SRT (0.5s to 6s), driving task complexity (low-medium-high) and interaction task complexity (low-medium-high) was investigated. 28 participants interacted with a VA-wizard, across all conditions in a within-subject design. Results reveal significant differences in user acceptance for SRT, with 0.5s to 1s being accepted across all conditions, whereas delays exceeding 2s are largely unaccepted. However, higher task or driving complexity makes users more tolerant of slower SRTs. Compared to literature on traditional VAs, this study suggests that users are more critical regarding SRT when interacting with AI-enabled systems and provides guidance on required SRTs.

Description

Supervisor

Nieminen, Mika P.

Thesis advisor

Blattner, Andreas
Dittrich, Monique

Other note

Citation

Endorsement

Review

Supplemented By

Referenced By