Designers Characterize Naturalness in Voice User Interfaces: Their Goals, Practices, and Challenges

Title of the Paper

How Designers Characterize the Naturalness of Voice User Interfaces (VUI): Goals, Practices, and Challenges

Paper Information

• Research Domain: Human-Computer Interaction, Voice User Interface Design, Natural Language Processing
• Keywords: Naturalness, Voice User Interface, Designers, Voice Interaction, Practical Application, Social Conversation, Transactional Conversation, Challenges, Tool Support, Conversational Naturalness

Research Background and Issues

• Identified Issues or Challenges:

  1. Current design guidelines for voice user interfaces (VUI) emphasize achieving "natural human-machine dialogue," but lack in-depth research on how this process is realized.
  2. Designers face specific challenges in creating natural dialogues, such as balancing social and task-oriented interactions and insufficient tool support.
  3. Although there is extensive literature on user interaction with VUIs, the perspective of designers has been significantly undervalued.
  4. Existing tools and technologies for natural language design in VUIs (e.g., SSML) fail to adequately meet design needs.

• Why This Issue Is Important:

  1. Natural voice interactions enhance user experience, improving VUI intuitiveness and accessibility.
  2. Understanding designers' experiences and difficulties is essential for establishing clear and practical design guidelines and developing supportive tools.
  3. With the growing prevalence of intelligent voice assistants (e.g., Siri, Alexa, Google Assistant), promoting more "natural" designs is increasingly urgent.

• Research Motivation and Related Work:

  1. Current research primarily focuses on user experience, with limited attention to how designers define "naturalness" and achieve related goals.
  2. Definitions of "naturalness" in natural language interaction are diverse and fragmented, lacking consistency.
  3. There is an urgent need for systematic research on designers' cognition, methods, and challenges in design practices.

Solutions

• Proposed Methods or Solutions:

  1. Conduct interviews with 20 VUI designers to study how naturalness is defined and achieved in design practices.
  2. Summarize 12 design dimensions of naturalness and classify them into three conversational contexts: social, transactional, and core.
  3. Identify seven major challenges and propose targeted improvement strategies and guidelines.

• Innovative Aspects of the Solutions:

  1. Detailed decomposition of naturalness features and classification based on three interaction contexts.
  2. Identification of challenges in designing natural interactions and linking them closely to real-world design scenarios.
  3. Detailed suggestions for improving existing tools (e.g., SSML).

• Implementation Steps and Key Techniques:

  1. Interview Study:
     • Participants included designers with diverse backgrounds and experiences (20 individuals: 13 professional designers and 7 amateur designers).
     • Semi-structured interviews were conducted to gather insights into their design practices, goals, and challenges in creating natural voice interactions.
  2. Data Analysis:
     • Reflective thematic analysis was conducted using Braun and Clarke's framework.
     • Both inductive and deductive methods were employed to identify new themes and validate the applicability of existing theories.

Research Findings

• Specific Findings:

  1. 12 Naturalness Features:
     • Social Conversations: Expressing empathy, maintaining interest and engagement, showing interest in the user.
     • Transactional Conversations: Proactively assisting users, strong topic handling within task domains, presenting task-adapted roles, delivering information with high speed and accuracy*, remembering user information for personalized services*.
     • Core Features: Using colloquial language, appropriate intonation variation, understanding language changes, collaborative repair of conversational breakdowns.
  2. 7 Major Challenges:
     • Challenges specific to social conversations: balancing task-oriented and social characteristics, lack of expressiveness in voice synthesis, inefficiency and suboptimal results of SSML, inability to accurately capture user emotions.
     • Common challenges: overly abstract design guidelines lacking practical applicability, difficulty in writing text in colloquial form, handling diverse user inputs and their contexts.

• Comparison with Existing Solutions and Advantages:

  1. Provides a more systematic characterization and classification of naturalness features.
  2. Proposes specific improvements for limitations in tools like SSML, such as more intuitive demonstration-based voice editing tools and automated script validation tools.
  3. Addresses details of designers' actual work, bridging the gap between theoretical research and practical application.

• Experimental or Evaluation Results:

  1. Interview data revealed designers' precise understanding of naturalness needs and existing shortcomings.
  2. Thematic analysis of the data provided a comprehensive summary of challenges.

• Limitations and Future Directions:

  1. Limitations:
     • The study did not directly observe designers' work environments, lacking contextual richness in analysis.
     • The 12 naturalness features reflect designers' perceptions, which may not fully align with end-user experiences.
  2. Future Directions:
     • Conduct research based on field observations or contextual investigations.
     • Explore potential differences in perceptions of naturalness between designers and end users.
     • Investigate the support provided by specific tools (e.g., natural voice editing tools with demonstration functions) for design practices.