Examining the influence of the domestic social robot anthropomorphism on users’ perceived affinity via intelligent design process: robot social presence as a mediator

Recent advancements in artificial intelligence and robotics have accelerated the integration of social robots into daily life, necessitating a deeper understanding of how human-robot interaction (HRI) can be optimized. A new study delves into the intricate relationship between robot anthropomorphism—the extent to which robots are designed with human-like characteristics—and users’ perceived affinity, with a particular focus on the mediating role of social presence. The research, conducted by a team of academics from East China, employed an innovative intelligent design process to develop robot stimuli and utilized a mixed-factorial experimental design to analyze user responses.

The study, published in Frontiers in Psychology, investigated how two primary dimensions of anthropomorphism—appearance anthropomorphism (AA) and interaction anthropomorphism (IA)—influence users’ perceived affinity and, consequently, their intention to use domestic social robots. Findings indicate that while interaction anthropomorphism plays a significant role, the uncanny valley effect remains a critical factor influencing user perception, particularly concerning highly human-like designs.

The Evolving Landscape of Social Robots

The trajectory of intelligent robots has shifted dramatically from single-purpose industrial tools to sophisticated service robots capable of complex tasks in both commercial and domestic settings. As these robots become more prevalent, from food delivery drones to automated cleaning systems, a key challenge emerges: user familiarity and acceptance. Early research consistently pointed to functionality as a primary driver of adoption. However, with increasing human-robot interaction, the social and emotional aspects—such as the robot’s appearance and the user interface’s intuitiveness—are now recognized as critical design elements, crucial for cohesive robot integration.

This paradigm shift has propelled robot anthropomorphism to the forefront of HRI research. By imbuing robots with human-like features, such as facial expressions, emotions, and behaviors, designers aim to enhance user affinity and foster more meaningful interactions. Examples like the emotionally expressive robot Cozmo, the social interaction-focused Pepper, and the lifelike robotic head Flobi underscore this trend. The underlying human desire is to create robots that resonate emotionally, facilitate intuitive interactions, and ultimately improve user acceptance and intention to use.

Investigating Anthropomorphism: Appearance vs. Interaction

Prior research has largely suggested that higher levels of anthropomorphism, whether in appearance or interaction, can boost users’ perceived affinity and intention to use. For instance, a highly anthropomorphic appearance in a grocery service robot has been shown to positively influence adoption intent. Similarly, the mimicry of human facial expressions can evoke strong emotional responses. However, a significant gap existed in systematically examining the influence of anthropomorphism on domestic service robots and, critically, the potential interaction effects between appearance and interaction anthropomorphism.

This study addressed this gap by adopting a 2×2 mixed-factorial research design. Appearance anthropomorphism (low vs. high) was treated as a between-subjects factor, meaning participants were exposed to only one level. Interaction anthropomorphism (low vs. high) was a within-subjects factor, where participants experienced both levels. Perceived social presence was identified as a key mediator, and users’ perceived affinity served as the primary dependent variable.

The Intelligent Design Process: Crafting Realistic Stimuli

A novel aspect of this research was the utilization of an "intelligent design process" to develop the robot stimuli. This approach leverages advanced technologies such as Artificial Intelligence (AI), augmented reality (AR), and sophisticated application programming interfaces (APIs) to enhance design quality and efficiency. For the appearance anthropomorphism manipulation, AI image generation tools like Stable Diffusion, in collaboration with human researchers, were employed to create high-fidelity robot images. This human-AI collaboration allowed for rapid iteration and refinement of visual designs, producing a set of robot images with distinct levels of human-likeness.

Similarly, for interaction anthropomorphism, an intelligent design method involving iPhone’s Memoji App was used to generate realistic facial expressions. By scanning human faces, the app could instantly produce detailed animated expressions. These were then adapted for the study. For lower levels of interaction anthropomorphism, basic shapes were used in Adobe Illustrator to create simpler facial cues. This process ensured that the stimuli accurately reflected the intended levels of anthropomorphism, a critical step for validating the experimental design.

Key Findings: Interaction Dominates, Uncanny Valley Looms

The study’s results revealed several significant findings:

• Interaction Anthropomorphism (IA) is Key: A significant positive main effect of interaction anthropomorphism on users’ perceived affinity was observed. This indicates that how a robot interacts and expresses itself emotionally plays a crucial role in how likable users find it. Specifically, lower levels of interaction anthropomorphism were generally preferred, suggesting the influence of the uncanny valley effect. Participants found robots with less complex or overly human-like expressions more comfortable.
• Appearance Anthropomorphism (AA) Less Impactful: Contrary to some prior research, the study found no significant main effect of appearance anthropomorphism on perceived affinity. This suggests that the mere visual human-likeness of a robot’s form may not be as influential as its interactive capabilities in shaping initial user perception, especially when static images were used as stimuli.
• The Uncanny Valley Effect Confirmed: The preference for lower levels of interaction anthropomorphism aligns with the well-documented uncanny valley effect. This phenomenon describes the dip in affinity and comfort experienced when robots become almost, but not perfectly, human-like, evoking feelings of eeriness.
• Significant Interaction Effect: A notable interaction effect between appearance and interaction anthropomorphism was identified. This suggests that the combined effect of a robot’s look and its interactive behavior is more complex than their individual impacts. Intriguingly, in high interaction anthropomorphism conditions, users’ perceived affinity was higher for robots with high appearance anthropomorphism compared to those with low appearance anthropomorphism. This hints at a potential for highly anthropomorphic robots in specific domestic contexts, provided their interactive design is well-executed.
• Social Presence as a Crucial Mediator: The study strongly supported the mediating role of perceived social presence. This means that interaction anthropomorphism influences users’ perceived affinity through their sense of the robot’s social presence. When a robot interacts in a more human-like manner, users feel a greater sense of its presence, which in turn enhances their positive feelings towards it. While perceived emotional interdependence was initially considered, its strong correlation with social presence and lack of significant independent influence led to its removal from the final model, highlighting social presence as the primary mediating factor.

Implications for Future Robot Design

The findings of this study offer significant theoretical and practical implications for the development of future domestic social robots.

Theoretical Contributions:
The research validates the applicability of the Robots’ Social Presence Modified Model and the Theory of Planned Behavior in understanding user perceptions of social robots. It underscores the critical mediating role of perceived social presence in linking robot design characteristics to user attitudes and intentions. The study also highlights the nuanced impact of the uncanny valley effect on interaction anthropomorphism, urging further exploration of its boundaries.

Practical Recommendations:
For robot designers and manufacturers, the study offers actionable insights:

1. Prioritize Interaction Design: Given the strong influence of interaction anthropomorphism, designers should focus on creating intuitive and emotionally resonant interaction capabilities. Initial adoption may benefit from employing lower levels of IA to avoid the uncanny valley, opting for simpler, clearly interpretable expressions.
2. Leverage Social Presence: Actively design robots to enhance perceived social presence. This can be achieved through multi-sensory interactions, clear communication of emotional states, and designs that foster a sense of companionship and co-presence.
3. Consider the Uncanny Valley: Designers must carefully navigate the uncanny valley. While highly anthropomorphic designs might be aspirational, they require meticulous execution to avoid triggering user discomfort. Further research into multi-level anthropomorphism is needed to map out the sweet spot between machine-like and overly human-like.
4. Embrace Intelligent Design: The successful application of intelligent design processes in creating stimuli demonstrates its value for rapid prototyping and high-quality visual asset generation in robot design. This approach can significantly accelerate innovation and reduce development costs.
5. Future-Proofing with High Anthropomorphism: The observed interaction effect suggests that as users become more accustomed to robots in their homes, highly anthropomorphic designs (both appearance and interaction) may eventually yield superior user acceptance, provided the uncanny valley is carefully managed.

Limitations and Future Directions

The researchers acknowledge several limitations. The use of static images as stimuli, while efficient, may not fully capture the dynamic nature of HRI. Future studies could benefit from using dynamic stimuli, animations, or even physical robot prototypes to provide a more ecologically valid experience. The study’s reliance on a student sample also limits the generalizability of findings to broader demographics, particularly older adults or individuals with varying technological literacy. Longitudinal studies are recommended to understand the long-term development of social presence and affinity in domestic settings. Furthermore, exploring a wider range of anthropomorphism levels and investigating other potential mediating factors beyond social presence could enrich our understanding of human-robot relationships.

In conclusion, this research provides valuable insights into the complex interplay of robot anthropomorphism, social presence, and user affinity. By employing innovative design processes and rigorous experimental methods, the study illuminates pathways for creating social robots that are not only functional but also socially engaging and accepted within the home environment.