Voice in Human–Agent Interaction

Analysis

Why This Paper Matters

This survey addresses a critical gap in the design of intelligent agents: the role of voice. As social robots, conversational agents, and voice assistants become ubiquitous, understanding how voice shapes human perception and interaction is essential. The paper synthesizes decades of research to show that voice is not merely a communication channel but a social cue that influences trust, engagement, and user experience. Its human-centered perspective is timely given the rapid commercial adoption of voice interfaces.

The meta-synthesis approach is particularly valuable because it integrates findings across diverse agent forms—from embodied robots to disembodied assistants—revealing common principles. The identification of a generational shift in attitudes toward agent voice, linked to the proliferation of mobile assistants like Siri and Alexa, underscores the evolving nature of human-agent relationships. This work provides a roadmap for researchers and practitioners aiming to design more natural and socially acceptable voice interactions.

Technical Contributions

The paper's main technical contributions include:

• A vHAI classification framework that categorizes voice features (e.g., gender, accent, tone) across agent types, contexts, and user demographics.
• Grounding of findings in social psychology theories such as the Computers Are Social Actors (CASA) paradigm and the Uncanny Valley effect.
• A critical analysis of how voice interacts with agent embodiment, including mismatches between voice and physical form.
• Identification of future research directions, such as adaptive voice personalization and cross-cultural voice design.

Results

The survey reports that agent voice significantly impacts user perceptions of social presence, trust, and likability. For example, human-like voices generally increase engagement but can trigger uncanny valley effects if mismatched with appearance. Generational differences show younger users more accepting of synthetic voices, likely due to early exposure to mobile assistants. The paper does not provide quantitative metrics but synthesizes qualitative trends across 235 cited works.

Significance

This work has broad implications for AI practitioners designing voice interfaces. The framework can guide decisions on voice gender, accent, and expressiveness based on target user groups and contexts. It also highlights ethical considerations, such as stereotyping through voice design. As voice becomes a primary interaction modality, this survey offers a foundational reference for creating more effective and socially aware agents.