Keep it Short: A Comparison of Voice Assistants' Response Behavior

Title of the Paper

Keep it Short: A Comparison of Voice Assistants’ Response Behavior

Bibliographic Information

• Subject Area: Human-Computer Interaction and Voice Assistant Design
• Keywords: Voice assistant, natural language interaction, response behavior, voice interface design, empirical study, efficiency, user acceptance, humanized design, adaptability, technological interaction

Research Background and Issues

• Identified Issues or Challenges: The response behavior of current mainstream voice assistants (e.g., Alexa, Siri, and Google Assistant) predominantly adopts full-sentence formats and attempts to mimic human behavior. This design choice may limit the utility of voice assistants as tools, constraining the design space and reducing interaction efficiency.
• Significance: As voice assistants become increasingly prevalent in homes, smartphones, and cars, how users interact with these devices is critical. Enhancing interaction efficiency and user acceptance is essential to maximize the potential of voice assistants and shape the direction of technological development.
• Research Motivation and Related Work:
  1. Current voice assistants are often designed with human-like characteristics to enhance user satisfaction and social interaction experiences. However, this approach can sometimes lead to unrealistic expectations of system capabilities.
  2. Humanized design may increase the social appeal of voice assistants but can also result in inefficiency and cognitive overload.
  3. To address these issues, further exploration of voice assistant response styles and their impact on user experience is necessary.

Solution

• Method or Solution: The authors designed and implemented a prototype voice assistant supported by a web browser, showcasing three different response styles for user testing: Minimal Confirmation Mode (Minimal), Keyword Mode (Keyword), and Full Sentence Mode (Full Sentence).
• Innovations:
  1. Proposed a non-humanized design concept for voice assistants, exploring whether concise keyword responses can improve efficiency while maintaining practicality.
  2. Developed a prototype system to capture user experiences and preferences across three different response styles, enabling experimental validation.
  3. Designed the experiment to systematically compare the effects of traditional voice assistants versus more tool-oriented response styles.
• Implementation Steps and Key Technologies:
  1. Experimental Design: Developed interaction models featuring eight typical requests (e.g., weather, news, setting a timer) and three response styles, observing user preferences and evaluations in different scenarios.
  2. Technical Implementation: Utilized the WebSpeech API and JavaScript to build voice recognition and synthesis functionalities for the voice assistant, designing a prototype response mechanism with customizable user request-response features.
  3. User Study: Recruited 72 participants and conducted online experiments to collect user ratings (quality, behavior, clarity, efficiency) for different response styles, while tracking their preferences.

Research Results

• Specific Findings:
  1. Conducted a systematic analysis of the response behaviors of three mainstream voice assistants (Alexa, Siri, and Google Assistant), revealing that these assistants predominantly use full-sentence formats to mimic human language.
  2. User research confirmed that concise keyword responses achieved similar usability and acceptability ratings compared to full-sentence responses, while significantly improving efficiency due to shorter response times.
  3. Provided new empirical insights into voice assistant design, such as the need to adjust response styles based on task complexity and emphasizing tool-oriented efficiency in voice assistant design.
• Advantages Over Existing Solutions:
  1. Introduced keyword-based responses and confirmation modes into the voice assistant domain, reducing cognitive load while providing straightforward feedback and enhancing response transparency.
  2. Experimental data supported a critical perspective on the current trend of fully humanized design, proposing new design principles aligned with human language capabilities.
• Experimental or Evaluation Results:
  1. No significant differences were found between concise keyword responses and full-sentence responses in terms of effectiveness and user preference.
  2. Concise keyword responses demonstrated significant superiority in efficiency and ease of understanding, particularly for executing simple commands.
  3. Younger users tended to prefer shorter response styles, while older users favored full-sentence formats.
• Limitations and Future Directions:
  1. The experiment was conducted in an online environment, which may not fully reflect real-world usage scenarios. Future studies should involve long-term research in real-world environments.
  2. The study did not include users who do not use voice assistants, necessitating further exploration of their acceptance and preferences.
  3. Proposed the development of adaptive voice assistants capable of dynamically adjusting to user needs, further enhancing the potential for personalized design.

Conclusion

This study explores a more efficient interaction design possibility by introducing new response styles for voice assistants, while highlighting the conflict between humanized design and efficiency in current voice assistant designs. The authors suggest that future voice assistant designs should prioritize supporting user-customizable response styles to achieve an optimal balance between efficiency and acceptability.