TL;DR
While engineers often focus on a robot’s visual design or functional AI, the unintentional "consequential sounds" of its motors and joints are a major hurdle for human acceptance. This study demonstrates that these whirrs and clicks significantly increase human anxiety and distraction while decreasing the willingness to share a workspace, even if the robot’s physical appearance is well-liked.
What Problem Does This Paper Address?
As robots transition from isolated factory floors to shared public and domestic spaces, Human-Robot Interaction (HRI) research has primarily focused on visual aesthetics, social cues, and intentional speech. However, robots are inherently noisy. These "consequential sounds"—the byproduct of actuators and mechanical movement—are often overlooked during design but are omnipresent during operation.
The authors address a critical gap: we lack a quantitative understanding of how these unintentional noises affect human perception. If a robot sounds "scary," "annoying," or "distracting," will people actually want to live with it? This paper investigates whether these sounds create a barrier to the long-term cohabitation and adoption of robotic technology.
Methods
The researchers conducted a between-participants study with 182 participants (totaling 858 trials).
- Conditions: Participants were split into two groups: a "Sound" condition (viewing videos of robots with their natural operating noise) and a "Control" condition (viewing the same videos, but silent).
- Stimuli: 20-second videos of five diverse robotic platforms: a Quadrotor (drone), Unitree Go1 (quadruped), SoftBank Pepper (humanoid), ClearPath Jackal (wheeled), and UBTECH Yanshee (small humanoid).
- Measurement: Participants rated the robots on 11 Likert-scale questions across four thematic dimensions:
- Associated-Affect: Feelings of anxiety, agitation, or safety.
- Distracted: The robot's perceived ability to interrupt focus.
- Colocate: Willingness to share a home, workplace, or public space.
- Like: General likability of the robot’s appearance and movement.
- Statistical Approach: The team used least-squares linear regression with deviation coding. This allowed them to isolate the specific effect of the sound by comparing it against the visual-only baseline provided by the control group.
Datasets
The study utilized a custom-built set of video stimuli featuring five distinct robotic platforms to ensure a broad coverage of form factors and sound profiles (rotary buzzing, gear whirring, etc.). The resulting quantitative data from the 182 participants serves as the primary evidence base for the paper’s conclusions.
Key Insights and Ideas
- The "Likeability" Paradox: Interestingly, the presence of sound did not significantly impact how much people "liked" the robot's physical appearance or movement. However, it did significantly impact how the robot made them feel. This suggests that a robot can be aesthetically pleasing but still psychologically taxing to be around.
- Sound as a Stressor: Consequential sounds correlated with higher negative "associated affects." Qualitative feedback revealed that loud or "clanky" noises made participants feel anxious and less able to appreciate the technology.
- The Quadrotor Outlier: While most robots followed a similar trend, the Quadrotor (drone) showed a borderline interaction effect, suggesting its high-pitched, consistent buzzing might be uniquely detrimental to human comfort compared to ground-based robots.
- Universal Distraction: Participants found all robots somewhat distracting, but the addition of sound made the "distraction" scores significantly worse across the board.
Key Results
The study confirmed Hypothesis 1: Consequential sounds have a significant negative effect on human perception.
- Associated Affect: Sound caused a statistically significant drop in positive affect ($p < .001$).
- Distraction: Sound significantly increased the perceived level of distraction ($p < .001$).
- Willingness to Colocate: Participants were significantly less willing to share space with a robot when they could hear it ($p < .001$).
- Regression Coefficients: The "Condition: Consequential Sounds" variable showed negative coefficients across all scales (e.g., $\beta = -0.525$ for Colocation), indicating a consistent downward trend in perception when audio was present.
Limitations and Future Work
- Online vs. Reality: The study was conducted via online videos. The authors acknowledge that in-person interactions—where sound has spatial properties and physical presence—might result in even stronger negative reactions.
- Environmental Masking: The videos used a quiet background. Future work could explore how ambient office noise or "sound augmentation" (adding pleasant sounds like birdsong to mask motor noise) might mitigate these negative effects.
- Neurodiversity: The authors suggest that future research should account for individual differences in sound sensitivity, such as misophonia or sensory processing differences associated with autism, which may drastically change how robot noise is tolerated.
- Sound Design: The paper concludes that "quiet" isn't the only goal; engineers should consider the quality of the sound, aiming for lower frequencies or more "organic" acoustic profiles.
Summary updated