What type of intelligent cockpit can help more in traffic accident scenarios? Considering the effects of gender and anthropomorphism in voice interface, and their interaction with text length

The increasing automation of vehicles has reduced traffic accidents. However, accidents cannot be entirely eliminated, so postaccident assistance is critical for minimizing casualties. Voice interaction with intelligent cockpits (ICs) can offer crucial emergency support. Limited research has focused on the design of ICs for traffic accident scenarios, particularly concerning voice features and linguistic content. In accident scenarios, this study investigated the effects of gender (female, male), anthropomorphism (human-like, machine-like) in the voice interface of IC on comprehension performance (comprehensibility, intelligibility), user experience (persuasion, trust, attractiveness, usefulness, acoustics quality and calmness), and their interaction with text length (short, medium, long). A driving simulator experiment with a three-factor repeated measures design revealed that ICs with female voice interfaces significantly enhanced comprehension performance (comprehensibility increased 5.3%, intelligibility increased 4.7%) and user experience (increases across dimensions: 3.1–8.1%). ICs with human-like voice interfaces significantly enhanced comprehension performance (comprehensibility increased 7.6%, intelligibility increased 3.8%) and user experience (increases across dimensions: 6.0–16.3%). ICs with short text significantly enhanced comprehension performance (comprehensibility increased 14.7%, intelligibility increased 36.5%) and user experience (increases across dimensions: 4.7–8.7%). Additionally, interaction effects were observed. ICs with female voice interfaces mitigated the negative impact of increased text length, whereas ICs with male voice interfaces amplified it. ICs with human-like voice interfaces performed better in short text, although this advantage diminished with longer text. The findings suggest prioritizing ICs with female and human-like voice interfaces and optimizing the match between voice features and text length. The design implications of ICs for accident scenarios are discussed.