Metaverse growth highlights challenges in avatar-mediated trust due to limited non-verbal cues. This study focused on posture, a common non-verbal cue and examined how an avatar’s forward-leaning, neutral, or backward-leaning posture impacts trust perception by users. In a one-factor, three-level VR experiment, participants interacted with a avatar with one of three fixed postures. Results showed that forward-leaning posture significantly increased cognitive trust compared to other postures, suggesting that it may conveys benevolence and competence in metaverse environments.

The first purpose of this study was to investigate the influence of the type of perceived social support that virtual agents bring to a human user when the user shares one’s negative emotional experience with the agent, on the extent to which the user perceives attachment functions as well as on trust on the agent. The second purpose was to investigate the influence of differences in agency on the extent to which a human user perceives attachment functions and trust on the agent. An experiment with kind of social support and agency as two between-subject factors was conducted with 69 Japanese adult males as subjects. Subjects participated in the experiment one at a time, conversing with a virtual agent in a VR environment. The subjects were asked to recall their own worrisome event and to talk with the virtual agent about the worrisome event they recalled. After subjects shared their worrisome events, the virtual agent gave the subjects statements to make them perceive social support, either instrumental or emotional. The results showed that when the subjects talk about their negative emotional experiences, given the information that the virtual agent is controlled by a computer, the virtual agent that gives the subjects statements eliciting perception of emotional support has positive effects on the the subjects’ perception of secure base, safe haven, maintaining proximity, as well as on cognitive and affective trust.

People respond more positively to anthropomorphic social robots that possess humanlike appearance, behaviors or interaction. To enhance human-robot interaction it is necessary to investigate and understand how people perceive such humanistic behaviors. This study explored the influence of humanlike-listening behavior of social robots on participants’ perception of the robots. We compared three different listening behaviors including; active listening, active empathic listening, and verbal-only empathic listening, and evaluated the effect of each behavior on participants’ perception of social robots in terms of animacy, likeability and perceived intelligence. We focused on emotional-verbal and nonverbal aspects of listening behaviors. The results indicated that a social robot with active-empathic listening behavior is perceived higher than a social robot with active listening behavior in terms of animacy, likeability and perceived intelligence. However, nonverbal aspects of listening behavior did not significantly influence the perception of the robot, when emotional-verbal behaviors were present. Furthermore, emotional-verbal behaviors of listening were more closely associated with perceived intelligence, while nonverbal behaviors affected animacy and likeability.

The development of automated, unobtrusive measurement of pro-social indicators such as trust, empathy, liking and altruistic tendency in daily interactions is an unsolved problem in designing intelligent agents that socially interact with people. This work puts forward and tests a model for using emotional mimicry (matching of emotional displays/expressions) as a potential proxy measurement for the pro-social indicators above. For this purpose, a human-agent interaction experiment was conducted in a social context where the interacting agent belonged to a competitive in-group/out-group with the interacting person. Computer vision and Cross Recurrence Quantification Analysis (CRQA) were used to measure people's mimicry of facial expressions in response to agent emotional stimuli. We found that the emotional mimicry of sadness was positively correlated with the empathy that people felt towards the agent. We could not find sufficient support for the relationship of sadness mimicry with the other pro-social indicators, or happiness mimicry with any pro-social indicators; the reasons for this are investigated. The results suggest that with the accurate measurement of one's emotions, it could be possible to measure emotional contagion/spread in context and, subsequently, one's pro-social attitudes.

Emotional contagion is the process by which emotions unconsciously spread between people, often via emotional mimicry, a process in which we mimic others’ emotional displays, and consequently, converge with their emotions. The social context theory of emotional mimicry in human-human interactions suggests that these processes are modulated by social factors, such as the group membership identities of interactants (whether the interactants belong to the same social group or not). However, the social factors that modulate the emotional contagion between humans and artificial agents are not known yet. Hence, this research attempts to understand the process of emotional contagion in human-agent interaction by conducting an interaction experiment in a competitive in-group/out-group context. Emotional contagion as well as the social attitudes of trust, empathy, liking, bonding, and pro-social orientation were found to be significantly higher in the in-group than in the out-group setting. Furthermore, these social attitudes were highly correlated with emotional contagion, which demonstrates the potential of using emotional contagion as an unobtrusive proxy measurement of people’s social attitudes.

Self-disclosing one's life experiences is known to be beneficial to the psychological health of older adults, from the viewpoint of integrity. It has been shown that people tend to self-disclose more to the people they are fond of. For improving interpersonal liking, the use of the “gain effect” has been shown to be effective. “Gain effect” refers to that people like a person who initially has a negative attitude but gradually develops a positive attitude, compared to a person who consistently has a positive attitude. Based on these previous studies, our study aims to clarify the effect of a change in the robot's listening attitude on the self-disclosure of the elderly. We conducted an preliminary experiment, wherein 15 elderly participants self-disclosed to a robot for approximately 20 minutes. The participants were assigned to one of three groups according to the type of robot they self-disclosed. 1) The CN group where participants interacted with the robot that consistently listened with neutral behavior, 2) the CP group where participants interacted with the robot that consistently listened with positive behavior, and 3) the Ch group where participants interacted with the robot that listened with neutral behavior first and subsequently with positive behavior. The results revealed that the Ch group participants' ratio of self-disclosure utterances was the highest among the three groups. Furthermore, the willingness to self-disclose tended to be higher for the Ch group than for the CN group, in terms of everyday experiences and loss experiences.

Studies of personal space between human and robot have been growing. However, little is known about the effects of personal space violation by a robot. Expectation Violation Theory (EVT) explains how people respond to personal space invasion by another person, depending on if the person is perceived as rewarding or punishing. This study aims to extend EVT to HRI and to examine the effects of personal space violation using EVT as an analytical framework. A 2 × 5 within-subject experimental design was employed. The participants were approached by two robots with different reward value (rewarding and punishing) at varying distances (threat, near, norm, slightly far, and far distances). The participants were asked to rate the robots’ perceived warmth, competence, and comfort scores after each approach to evaluate the communication outcomes after the personal space violation occurred. The results suggested that approaching distances had influences on the communication outcomes. As predicted by EVT, how each distance affected the communication outcomes differ according on robot’s reward value. The findings indicated that EVT was partially successful in predicting the effects of personal space violations depending on the reward value and approaching distance of a robot. This study has provided a partial support that EVT can be applied to HRI and contributed to understanding the effects of personal space violation by a robot. When considering personal space in HRI, perceived reward value of the robot plays an important role as it can influence the effects of the personal space violation greatly.

The current study investigated the effect of a robot's gaze direction while not engaged in any tasks (i.e., while "idle") on users' comfort and subjective impressions of the robot. We predicted that users would feel differently when the robot looked at the same objects as the user (i.e., during joint attention between robot and user) or not, even when the robot's gaze direction was the same. We conducted an experiment in which participants watched a video clip five times in five different robot gaze direction conditions. To examine the effects of joint attention, participants were divided into two groups by the distance from the monitor on which the video was presented. Joint attention between the robot and participant occurred in different robot gaze directions between the two groups. We measured participants' comfort, subjective impressions of the robot, and related variables. The results revealed that the robot's gaze direction affected participants' comfort and subjective impressions among participants 75 years old and above. However, we found no effect of joint attention. Additionally, participants felt most comfortable and had the most positive impressions of the robot when it looked at a monitor on which participants did not watch the video clip.

The purpose of this study was to propose a model of development of trust in social robots. Insights in interpersonal trust were adopted from social psychology and a novel model was proposed. In addition, this study aimed to investigate the relationship among trust development and self-esteem. To validate the proposed model, an experiment using a communication robot NAO was conducted and changes in categories of trust as well as self-esteem were measured. Results showed that general and category trust have been developed in the early phase. Self-esteem is also increased along the interactions with the robot.

This study investigated the effect of speech errors made by a robot on people's perception of the familiarity of the robot. Four types of speech errors were implemented for a communicating robot. Subjects' perception of the familiarity of the robot, along with other subjective evaluations of impressions of the robot, were measured in an experiment and compared between conditions with and without speech error. Results showed that speech error improved the familiarity score when the robot did not make any speech errors in the first contact with subjects and then made speech errors in the second contact on a separate day. On the other hand, speech errors lowered people's perception of the sincerity of the robot. These results imply that the robot should not make speech errors in the early stage of engagement with human users, while some speech errors after the users become accustomed with the robot might be effective in improving users' perception of the familiarity of the robot. Results also showed that speech error improved familiarity for people having low-scoring attitudes about the diversity and operation of robots.

Recently, autonomous agents with emotional behavior are developed to improve the human-agent interaction. Users' perceived "believability" of autonomous agents is considered to affect interactions with agents. While human-like emotional behavior is considered to increase believability, actual emotional behaviors differ across cultures. Therefore culture-specific emotional behaviors of agent might increase believability. In this paper, two kinds of emotional behaviors specific in Japanese culture, namely amae and Japanese way of agency appraisal, were implemented in agents. Influences of these behaviors on believability were investigated experimentally. Results showed that both Japanese behaviors increased believability. Especially amae increased personality factor, which is one of the factors of believability.

The purpose of this study was to compare multi-dimensional construct of people's attitudes toward robots across international regions. A questionnaire was developed with 125 elements representing people's attitudes toward robots, and used in an investigation in Japan, China and Taiwan. A factor analysis was conducted on the responses to the questionnaire from all regions, and a twelve-factor structure was extracted. Moreover, a factor analysis was conducted on the responses of the each region, and factor structures were compared with the universal factor structure extracted from all regions. Regional similarities and differences in dimensions of attitudes towards robots were discussed.

The purpose of this study was to investigate dimensions that construct people's attitudes toward robots. In the first phase, investigations using free description and group interview were conducted to extract potential elements for attitudes toward robots. In the second phase, a questionnaire battery was developed based on the elements extracted, and a survey investigation was conducted with the questionnaire. A factor analysis was conducted on the responses to the questionnaire, and nine factors were extracted as dimensions of people's attitudes toward robots.