TL;DR
This study investigates how the voice of humanoid agents impacts their acceptance across different age groups, with a particular focus on seniors. The findings reveal that voice is a critical factor for elder acceptance, influencing their willingness to interact and their perception of the agent's gender, a preference not observed in younger demographics.
What Problem Does This Paper Address?
As ambient assistive technologies increasingly incorporate humanoid agents into daily life, understanding user acceptance is paramount. While research has explored factors like appearance, less attention has been paid to the specific role of an agent's voice, especially across diverse age groups. This paper addresses the gap in knowledge regarding how voice influences the acceptance of humanoid agents by elders, young adults, and adolescents, and how this differs from their acceptance of mute agents or voice-only interfaces. The authors aim to inform the design of more effective and accepted assistive technologies, particularly for vulnerable populations like seniors.
Methods
The researchers conducted experiments involving 316 participants divided into seven groups: three groups of elders (65+), two groups of young adults (22-35), and two groups of adolescents (14-16). Participants were presented with video clips of four virtual humanoid agents (two male, two female) with and without synthetic Italian voices. A key methodological innovation was the inclusion of a group of elders who only listened to the agents' voices, allowing for a direct comparison of voice-only interaction versus visual-only or combined visual-vocal interaction.
Participants completed the Virtual Agent Acceptance Questionnaire (VAAQ), which assesses willingness to interact, pragmatic qualities, hedonic qualities (identity and feeling), and attractiveness. Statistical analysis, including repeated measures ANOVA and Bonferroni post hoc tests, was used to compare scores across age groups and conditions (speaking vs. mute agents, voice-only for elders).
Datasets
Not discussed in the paper. The study utilized virtual agents selected from the website BOTLIBRE and synthetic voices generated by Natural Reader.
Key Insights and Ideas
- Voice is King for Seniors: For elders, the presence of a voice dramatically increases their willingness to interact with humanoid agents and significantly influences their perception of the agent's gender. This suggests that voice is not just a feature but a fundamental requirement for elder acceptance.
- Age-Related Differences in Voice Impact: While voice is crucial for seniors, younger adults and adolescents show no significant preference for speaking over mute agents. Their acceptance is less dependent on auditory cues.
- Gender Preference Tied to Voice: Elders exhibited a preference for female agents, but this preference was only evident when the agents were speaking. When agents were mute, this gender preference disappeared, highlighting the interplay between voice and perceived gender.
- Voice as a Humanizing Element: The authors hypothesize that for elders, voice might compensate for a perceived lack of human-like features in virtual agents, making them more approachable and trustworthy, especially in assistive contexts where a human-like interaction is desired.
- Pragmatic and Hedonic Benefits of Voice: For elders, voice-only interaction was preferred pragmatically over mute agents, and voice also played a role in hedonic identity, suggesting that auditory feedback enhances the perceived quality and appeal of the agent.
Key Results
- Willingness to Interact: Seniors showed a significantly higher willingness to interact with speaking agents (mean=4.13) and voice-only agents (mean=4.50) compared to mute agents (mean=5.7). Young adults and adolescents did not show such strong preferences.
- Gender Preference: Elders favored female speaking agents over male speaking agents. This preference was not observed for mute agents or for younger age groups.
- Pragmatic Qualities: For elders, voice-only agents were perceived as more pragmatic than mute agents.
- Hedonic Qualities (Identity): Elders attributed better hedonic identity scores to agents' voices compared to speaking or mute agents.
- Hedonic Qualities (Feeling) & Attractiveness: No significant differences were found among elders for these dimensions when comparing speaking, mute, and voice-only agents.
Limitations and Future Work
The authors acknowledge that the study focused on specific age groups and a particular cultural context (Campania, Italy). Future work could investigate the degree to which voice-only interaction affects the acceptance of interactive technologies by adolescents and young adults. Additionally, exploring a wider range of voices and languages, as well as the impact of voice prosody and emotional expression, could provide further insights. The study also did not delve into the specific cognitive or psychological reasons behind the elders' strong reliance on voice, which could be a fruitful area for future research.
Summary updated