Cognitive workload using interactive voice messaging systems
archive: archived pipeline: cataloged verified
Summary
Coleman, Turrill, Cooper, & Strayer (2016, HFES Proceedings) report three experiments crossing speech quality (natural human voice vs synthetic text-to-speech) with task type (listen-only vs listen-and-compose-reply) for an interactive voice messaging system. Experiment 1 (N=45, 18-40 yr) ran the tasks in isolation at a desk; Experiment 2 paired them with a driving simulator; Experiment 3 used an instrumented vehicle on residential streets. A peripheral DRT (head-mounted red/green LED, microswitch on left thumb, ISO-2012 protocol) indexed cognitive workload via reaction time and A' sensitivity, with a Wizard-of-Oz operator providing perfectly reliable speech recognition.
Key finding
Composing replies to text/email messages by voice raised DRT reaction time significantly more than passively listening, while natural vs synthetic voice quality made no significant difference - speech production, not speech comprehension or voice fidelity, is the primary driver of cognitive workload in voice IVIS.
Methodology
mixed_methods
Sample size: 45
Quality score: 5 / 5