Abstract

This study explores how AI instructor voice characteristics, gender (male vs. female) and style (humanoid vs. robotic), influence learning in simulated driving. Drawing on dual-coding theory, we argue that these features shape cognitive effort and emotional responses, affecting both immediate reactions and later performance. A laboratory experiment using a simulator tested these effects. Findings indicate that male and humanoid voices generally enhance outcomes, with male humanoid voices producing faster responses and stronger retention. In contrast, robotic voices triggered quicker short-term reactions but impaired later performance. Mediation analyses showed that cognitive effort explained the benefits of male voices, whereas confused emotions mediated the negative effects of robotic voices on immediate reactions. This study extends AI-HCI research by showing how AI, positioned as an instructor rather than an assistant, can shape learning outcomes through voice design.

Download

Share

COinS