Exploring One-Way, Two-Way, and Proactive Voice Interactions: Usability and Decision Support in In-Truck AI Assistants

Sammanfattning

With rapid advancements in artificial intelligence (AI) and communication technologies like 5G and 6G, voice interaction is increasingly recognized as a promising approach for automotive human-computer interaction (HCI). However, existing research has largely overlooked the professional truck-driving context, focusing primarily on passenger vehicles. To address this gap, this thesis investigates how three distinct in-vehicle voice interaction styles—one-way, two-way, and proactive—affect truck drivers’ situational awareness, usability perceptions, and decision-making processes. A mixed-methods user study was conducted with six professional truck drivers. In a counterbalanced design, participants interacted with prototypes representing each of the three voice interaction modes. Data were gathered via think-aloud protocols, Wizard-of-Oz simulations, the System Usability Scale (SUS) questionnaire, and semi- structured interviews. The results indicate clear trade-offs between the interaction styles. The one-way system, though minimally intrusive, offered limited support for complex decisions. The two-way interaction was consistently favored for its balance between user control and responsiveness, leading to higher SUS scores and greater perceived trust. In contrast, the proactive mode elicited polarized reactions: while some participants appreciated its "smart" and preemptive nature, others found the unsolicited prompts to be intrusive. In terms of situational awareness, the two-way mode helped drivers clarify uncertainties on demand, whereas the proactive mode sometimes heightened awareness through timely alerts but could also disrupt attention at critical moments. Overall, these findings suggest that future in-vehicle voice assistants for the trucking domain must carefully balance proactive support with user autonomy. The design should ensure that drivers receive timely guidance without being either overwhelmed or under-informed.