TL;DR
This paper investigates the usability of voice user interfaces (VUIs) in car navigation systems, specifically how age groups interact with a VUI designed to be consistent with a graphical user interface (GUI). They found that while a consistent VUI/GUI approach improves usability, especially for older users, the complexity of voice guidance and system structure needs careful consideration to avoid frustrating experienced users.
What Problem Does This Paper Address?
As in-vehicle information systems become more sophisticated, their interfaces need to be intuitive and accessible to a wide range of users. While Automatic Speech Recognition (ASR) has enabled Voice User Interfaces (VUIs), there's a lack of detailed research on how VUI usability is affected by user demographics, particularly age. Existing studies have either lacked experimental results or focused on limited subject groups. This paper aims to bridge this gap by studying the usability of a VUI that mirrors a GUI (VGUI) across different age groups, with a specific focus on making it accessible to elderly users.
Methods
The researchers developed a prototype in-vehicle information application that could be operated via voice commands or an 8-inch touchscreen. They implemented core functions like dialing by phone number, accessing the phonebook, and searching for addresses or Points of Interest (POIs). A key aspect of their approach was the "VGUI" design, where voice commands were visually represented by virtual buttons on the screen, aiming to provide a consistent user experience between voice and touch interactions.
Usability tests (UTs) were conducted iteratively with 245 Japanese subjects, ranging from their 20s to 60s. These tests involved performing specific tasks, and subjects were video-recorded. When ASR failed or was deactivated, subjects were interviewed about their thought processes and next actions. The prototype was progressively refined based on the problems identified in these UTs. A preliminary study also helped understand user behavior with VGUs, revealing that users often relied on isolated words or connected words rather than full sentences for voice commands.
Datasets
Not discussed in the paper. The study utilized a self-developed prototype and conducted usability tests with human subjects.
Key Insights and Ideas
- VGUI Consistency is Key: A VUI that mirrors the structure and operation of a GUI significantly aids user understanding, particularly for those less familiar with voice interfaces.
- Age Matters for Usability: Older users (60s) experienced a higher number of problems and serious issues, indicating a slower learning curve for new system structures compared to younger age groups.
- Voice Guidance Needs Simplicity: While voice guidance is crucial for explaining system operation, it needs to be straightforward and concise. Overly long or repetitive guidance can become a point of frustration, especially for users who have already learned the system.
- Task Order Influences Perception: The order in which tasks are presented can impact perceived usability. Subjects in this study experienced more problems with the initial tasks, suggesting they were still forming their mental model of the VGUI.
- Experienced Users vs. Novices: A VGUI designed for ease of learning might not satisfy experienced users who quickly grasp the system. This suggests a need for adaptive interfaces that can adjust their behavior based on user proficiency.
- Elderly Users as Valuable Testers: Focusing usability studies on elderly users can be highly efficient, as they tend to reveal more problems and root causes, thus accelerating the design improvement process.
Key Results
- The research achieved its goal of at least 95% task completion rate for at least 95% of users, with one subject completing tasks at 85% and the rest exceeding 95%.
- Subjective interest in continuing to use the in-vehicle information appliance by voice increased from 64% (expectation) to 79% (experience) after the usability tests, indicating a positive impact of the VGUI design.
- Serious problems were disproportionately higher in the 60s age group for the "dialing number w/ GUI" task, with 50% of these serious problems attributed to this group.
- The most common causes of serious problems were related to voice guidance and understanding the system's structure.
Limitations and Future Work
The authors acknowledge that experienced users might find a VGUI designed for ease of learning to be less satisfying. They suggest that future research should focus on developing methods to dynamically assess user proficiency and adapt the VGUI's behavior accordingly. This would involve exploring how to measure a user's understanding and skill level iteratively to provide a more personalized and efficient user experience.
Summary updated