Visual and Cognitive Demands of a Large Language Model-Powered In-vehicle Conversational Agent

Monk,Chris; Ayala,Allegra; Yu,Christine S.P.; Fitch,Gregory M.; Gruber,Dara · 2026 · arXiv

DOI: 10.48550/arXiv.2601.15034

URL: https://arxiv.org/abs/2601.15034

archive: indexed pipeline: discovered

Abstract

Evaluation of attention and mental effort drivers need when using an advanced conversational AI system (Gemini Live) while driving. The study involved 32 licensed drivers performing tasks while researchers measured cognitive workload through response time tests, monitored eye movements, and collected subjective feedback. Gemini Live interactions shared similar levels of cognitive load to hands-free calls; mean glance durations remained well below the 2-second safety threshold, indicating low visual demand. Voice-based conversational AI systems demonstrated cognitive and visual demands comparable to established, low-risk hands-free benchmarks.

Access

Route: arxiv_oa

Publisher: arxiv

Topics