Abstract
As social robots transition from laboratory settings to real-world applications, understanding the factors that influence human trust and perception is critical. This study investigates the interaction between a robot's grammatical fluency and its task competency. Using a 2x2 experimental design, researchers analyzed how these variables affect perceptions of intelligence, reliability, and teamwork. The findings reveal that while language fluency is the primary driver of perceived verbal skill, actual task performance is the dominant factor in determining a robot's effectiveness as a teammate.
Introduction
The integration of social robots into homes and workplaces necessitates natural communication and effective task execution. While "language bias"—prejudice against non-native speakers—is well-documented in human-human interaction, its impact on human-robot interaction (HRI) remains under-explored, particularly when coupled with varying levels of task performance. This research addresses a fundamental question: does a robot's ability to speak fluently compensate for technical failures, or does broken grammar undermine even a highly efficient machine?
Methodology
The researchers conducted a large-scale HRI study involving 60 monolingual English-speaking participants. The experimental setup featured a NAO humanoid robot engaged in a "What Am I?" animal guessing game.
The study employed a 2x2 factorial design:
- Language Fluency: The robot either spoke with native-level English grammar (Fluent) or incorporated common non-native grammatical errors (Disfluent).
- Task Competency: The robot either correctly identified the user's animal (Competent) or incorrectly rejected a correct guess (Incompetent).
After three rounds of interaction, participants rated the robot on a 5-point Likert scale across six dimensions: verbal competence, intelligence, reliability, expectations, teammate quality, and willingness to work together again.
Results and Analysis
Statistical analysis using Multivariate ANOVAs (MANOVA) highlighted distinct patterns in how humans perceive robotic agents based on their communication style and performance.
The Dominance of Fluency in Verbal Perception
Language fluency was the overwhelming factor in judging the robot's verbal ability. Fluent robots received high scores (~4.5/5) regardless of their task success, while disfluent robots were rated significantly lower (~2.0-2.6/5). This suggests that users decouple linguistic skill from functional utility when assessing communication.
Competency as the Foundation of Teamwork
In contrast to verbal ratings, perceptions of the robot as a "teammate" and its ability to "meet expectations" were driven almost entirely by task competency. Users were significantly more forgiving of a disfluent robot that performed correctly than a fluent robot that failed the task.
The "Halo Effect" of Intelligence
Perceived intelligence emerged as a hybrid metric. While task success improved intelligence ratings, fluent speech provided a significant boost even when the robot failed. This "halo effect" suggests that polished communication can partially mask technical incompetence in the eyes of the user. Conversely, the "disfluent-incompetent" condition suffered a "double-whammy," receiving the lowest scores for both intelligence and reliability.
Discussion and Limitations
The study underscores that while developers should strive for linguistic naturalness, functional reliability remains the cornerstone of a positive user experience in collaborative settings. However, several limitations warrant further study:
- Participant Diversity: The focus on monolingual English speakers may not generalize to multilingual populations who might be more empathetic toward disfluent agents.
- Scope of Fluency: The study focused on grammar; future research should examine the roles of accent, prosody, and speech rate.
- Stakes of Interaction: The low-stakes nature of a guessing game may influence the "Willingness to Work Again" metric, which showed no significant differences across groups.
Conclusion
Effective social robotics requires a balance of social grace and functional precision. While fluent speech enhances the perceived intelligence and verbal skill of a robot, it cannot substitute for task competency when a robot is viewed as a collaborative partner. For developers, ensuring high task reliability is the most critical factor for fostering trust and perceived teamwork.
Summary updated