Navigating social dilemmas transparently is essential for social artificial agents, especially when resource values are subjective. We propose a model where agents act based on three values: self-value, estimated value to another, and the "looking-glass self" (the other’s estimate of the self’s value). We tested this model in a simplified Rummy game by optimizing three agent types (self-oriented: S, socially-oriented: SO, and socially-oriented with looking-glass self: SOL) in competitive and cooperative simulations. Results indicate that SO performed best in competition, while SOL was superior in cooperation. This study demonstrates that a simple, value-based model with a transparent computational process can yield distinct, context-appropriate strategies, providing a promising approach for building more understandable artificial agents.

This study presents a comparative analysis of Virtual Reality (VR) based single and group stress reduction sessions using a Large Language Model (LLM) agent as a virtual therapist. A total of 42 participants were recruited. 21 participants attended a single session, while the remaining participants were divided into 7 groups, with 3 members in each group, and joined the group session. Participants engaged in immersive VR environments designed for stress relief with interactive support from the LLM agent. Quantitative results revealed that both sessions effectively reduced stress, as measured by the Heart rate. However, group sessions demonstrated slightly better outcomes, including 14.7% greater STAI reduction(M=12.95, SD=2.4), 7.47% lower perceived workload(M=50.43, SD=4.2), and 20.53% higher engagement than single session as evidenced by interaction points(M=22.9) and sentiment positivity(M=0.32). Qualitative feedback highlighted peer support and collaborative dynamics in group settings, while single sessions provided a more personalized experience. By exploring how LLM agents perform across these contexts, this research highlights their potential to enhance VR therapy’s scalability, engagement, and effectiveness. These findings inform the design of accessible, impactful mental health interventions, advancing the integration of VR and AI to meet diverse stress management needs.

Although a smile is usually interpreted positively in social settings, it is not always valid with the change in the background knowledge of the situation. This is true especially in situations where one believes that the community in which they are is corrupted and might cause harm in some way. We believe that the same reasoning can be applied to HCI. But does that mean it is better not to smile in the corrupted community? For that, we designed an experiment that contains two conditions, Low-Risk and High-Risk. Our findings show a trend suggesting that friendliness is valued more in low-risk conditions, whereas in high-risk situations, this effect appears diminished. This study contributed to understanding the functionality of a smile in HCI where the risk is accompanied.

This paper presents a novel interactive navigation system using multimodal large language models (LLMs) to deliver intuitive landmark-based turn-by-turn directions through both-way human-agent interaction. Comparative user studies reveal that this cognitive navigation approach enhances driving behavior and user satisfaction over traditional numerical systems. Analysis of steering, acceleration, and braking patterns shows significant improvements (p < 0.01) with large effect sizes (d > 0.7). Subjective feedback indicates that 80% of participants favored the cognitive system, citing reduced stress, improved focus and better anticipation.

As Large Language Model (LLM) become more prevalent and capable, opportunities for influencing human decision-making are increasing. We engage in small groups, and complex and emergent interactions, such as compromises by members and proposals of new ideas, may occur. However, these creative aspects of interaction systems, frequently observed in small groups, have not received much attention. This study focuses on the decision-making process in small groups, considering individual members’ satisfaction with the interaction process and its outcome as variables. These variables were incorporated into the reward system of a reinforcement learning model, and learning was conducted. Furthermore, a linguistic simulation was performed using an LLM. The results suggest that when each member accepts the conclusion reached by the group with a positive attitude, it simultaneously reduces the time required to conclude decision-making. Moreover, the LLM can play a role in promoting the emergence of creative aspects in group decision-making, such as compromises and the proposal of new ideas. The findings highlight the potential of LLM in facilitating emergent decision-making processes in small groups and the importance of considering member satisfaction in enhancing the efficiency and creativity of group decision-making.

COVID-19 and technological developments have increased opportunities for remote communication. This has led to a focus on "telepresence avatar robots," a technology that makes people feel as if they are sharing the same place, even from remote locations. The telepresence avatar robot takes the approach of supporting communication by generating a realistic sense of conversation that incorporates technologies such as camera motion and eye contact with the speaker. However, this approach does not enable smooth communication because the robot operator cannot recognize the intention of the local worker’s behavior, since the information transmitted from the remote location is less than that available in face-to-face communication. In this study, we focus on gaze-following motion as an unconscious human behavior and implement it in a robot to verify whether a robot operator can perceive the intention of a remote worker through this autonomous motion. Eye-tracking is an action in which one tracks the other’s gaze in a local conversation, and it is performed unconsciously. The experimental results suggest that this autonomous behavior can generate an estimation of gazing intention Consequently, this approach may be a useful method for achieving smoother communication.

In recent years, communication between humans and interactive robots has increased, even though their conversations are not always one-on-one. Sometimes, robots accept newcomers or actively invite them to participate. In such situations, the robot needs to estimate the internal states of both the existing participant and the newcomer, and determine whether to accept or reject the latter’s participation based on its estimations of their internal states. In this study, we explored how humans estimate the internal states of others based on body torque, which is expressed as a combination of face and body directions in new participant situations. Unfortunately, how humans infer internal states from physical signals such as the body torques of current and new participants and how they behave has not been studied. In this study, we analyze the effects of body torque on human behavior based on the results of field experiments conducted at the National Museum of Emerging Science and Innovation (Miraikan). Our results suggest that the body direction of a robot affects human behavior in new participation situations.

COVID-19 and technological developments have increased opportunities for remote communication, including online meetings. This has led to a focus on “telepresence avatar robots”, a technology that makes people feel as if they are sharing the same place, even between remote locations. The telepresence avatar robot takes an approach to support communication by generating a realistic sense of conversation by incorporating technologies such as camera motion and eye contact with the speaker. However, these approaches involve conscious manipulation by the robot operator of actions that are unconsciously performed by the human operator. Compared to local communication, it is difficult to achieve a sense of presence for the robot operator. In contrast, we implemented autonomous behavior similar to the joint attention that humans unconsciously perform, and verified whether the robot operator feels that he or she is performing joint attention and gains a sense of presence through these autonomous behavior. Joint attention is an interaction in which both parties share the object of attention in a local conversation and is unconscious. We speculate that when the robot performs this behavior autonomously, the robot operator’s awareness of the person confronting the robot and the remote environment increases, creating a sense of presence as the operator gains a sense of immersion in the robot. The experimental results suggest that autonomous behavior generates joint attention gaze intention estimation, and that this approach may be a useful method for robot operators to obtain a sense of presence in the environment and to predict the behavior of others.

The physical signal called body torque, expressed as a combination of face and body directions, are likely to be effective in a newcomer situations in dialogues with three or more participants (multi-person conversation). Newcomers in a multi-person conversation infer the current participants’ internal states from their physical signals and decide when and to what extent they will be allowed join the conversation. However, insufficient research has been conducted on the agent’s design of behavior to encourage or discourage participation in conversations based on the current and new participants. This paper analyzes the effects of body torque on new participants based on the results of a field experiment at the National Museum of Emerging Science and Innovation (Miraikan). Our results suggest that current participants’ head and body directions independently and interactively affect a newcomer’s behavior.

People have always perceived machines to be entities that perform tasks based on explicit instructions from humans, so it was assumed that machines themselves would not behave autonomously according to their own independent purposes. Therefore, people do not understand how to manage interactions with machines when they use information from their surroundings or behave autonomously and purposely without instructions from humans. People are often negatively affected, by experiencing resistance or vigilance, in their interactions with machines that behave like people. In this study, we focused on the interaction between a person and an agent using a helping situation as the context. In the experiment, we observed the cognitive responses of the recipient to the helping agent in the helping situation and evaluated how people viewed the agents. The results showed that people were likely to perceive the agent as a person, even when the helper was an agent. Based on the findings from the experiments, we discussed how people can appropriately handle interactions with agents and how an agent should be designed to achieve good interactions with people.

This study investigates how the grounding process is composed and explores new interaction approaches that adapt to human cognitive processes that have not yet been significantly studied. The results of an experiment indicate that grounding through dialogue is mutually accepted among participants through holistic expressions and suggest that common ground among participants may not necessarily be formed in a bottom-up way through analytic expressions. These findings raise the possibility of a promising new approach to creating a human-like dialogue system that may be more suitable for natural human communication.

We propose an agent model that determines its behavior from an internal state and a spatial relationship with a target to generate approaching and avoiding behaviors in encounter scenes. This model is based on the relationship with an opponent rather than with a scenario. The agent moves to increase the utility value obtained from the preferences for both aggressive and passive involvement. We analyzed the behavioral and utterance data of human–human interactions based only on two-dimensional position information by simple-shaped robots. The rate of participants’ behavior following the model was significantly higher than that of a random walker. Based on this result, we estimated the internal state during the interactions from the behavior of the participants and analyzed it. The words uttered by one member of a pair correlated with the internal state estimated from the behavior of the other member of the pair. The frequency of the internal states observed from the participants who were recommended to interact with the partner was different from that observed from the participants who did not receive such recommendations. These results suggest that a model with two preferences can approximate a human’s internal state in encounter scenes.

Human-Agent Interaction (HAI) research covers a wide range of agents, including the modeling of human cognitive processes and the design of agents? cognitive processes. However, no framework for organizing the wide range of research areas has yet been concretely developed. In this study, we propose a framework for the design of Human-Agent Interaction. This framework is described as a system consisting of a set of variables that affect HAIs and a function that represents the relationship between these variables. We listed what papers were accepted at the international conference on Human-Agent Interaction in 2019 following a framework of variables on which each study focuses. We argued that approaches that connect variables and elaborate models of human cognitive processes are important for future HAI research.

In the field of Human-Agent Interaction (HAI), continuation of interaction is one of the main areas of research. If the behavior of the agent is too predictable, humans stop interacting with it when they get bored. In this study, we aim to build agents that people want to keep interacting with, by employing the Adaptive Control of Thought-Rational (ACT-R) cognitive architecture. As a preliminary step, we attempt to clarify the conditions required to maintain interaction between humans and agents, by modeling the experiences of fun and boredom based on intrinsic motivation.

To construct a general theory in human-agent interaction (HAI), it is essential to discuss how to generalize from models within the research area. In this study, we propose a formulation of HAI for the establishment of axiom systems. The environment for agents, behavior as a time change in the environment, and the internal state as an evaluation of the environment are identified as basic variables. The relations of these variables are expressed by functions. Ideally, HAIs should be explained in one axiomatic system to discuss the similarities and differences among human-human interaction, human-animal interaction, and human-robot interaction. For this purpose, we need to verify the model theoretically in a future study.

Currently, it is difficult for humans and AI agents to cooperate because the agent has incomplete intention understanding. In this paper, we propose the design of cooperative interaction between humans and AI creatures. As an experiment, two creatures learned to lift a heavy box in a virtual environment simultaneously. As a result, one of the creatures was able acquire the behavior of following the other creature automatically. We need to verify whether cooperation between humans and AI can be established through ongoing investigations.

To design agent behavior in the initial phase of communication, we propose an agent model that possesses preferences for a relationship with a target. As an internal state, this model considers two relationship preferences with a target: control and acceptance. The agent's spatial behavior, which is represented by a change in its position, is determined based on its internal state and its positional relationship with the target. We investigated the interactions caused by the combination of the internal states of two agents through a computer simulation. We simplified the combinations of the internal states for the agent to reach its preferred position and clarified the position combinations that become unreachable. Based on these results, we discussed the change of the internal state based on rationality. In this model, since the target's internal state is directly expressed by its behavior, the agent can adjust its own internal state based on rationality without estimating the target's internal state.

Getting clues about how to engage with others is quite difficult when the persons involved are mutually unknown and unrelated to each other: strangers, in other words. In fact, those who distribute leaflets or free samples of various goods to strangers in towns or on city streets seem to have a special manner of interacting with others. They can naturally experience success by handing out something without any extra precaution. This study focuses on such interaction that naturally starts by engaging with others. We explored the physical interaction in an experimental field between a person (P1) and an engrossed person (P2) by asking this question: how should P1 approach P2. Our result shows that P1 changed his/her route to approach P2 based on how engrossed P2 is with a book. This result indicates that getting clues for engaging with others must consider the mutual attitudes about how each person wants to interact with another based on two-dimensions: controlling the another and being accepted by another.

Communication, which begins by a consignor who verifies the signal of the communication needs and a receiver who recognizes it, is one element in the construction of relationships with others. Expression through language is effective for beginning to communicate with others. However, such expressions might produce social risk, because language expressions are explicit signals that cannot often ignore the receiver's communication needs. For example, the fatigue caused by social networking is one social risk that obstructs the construction of amicable relations. Instead, people often communicate by nonverbal expressions that provide vague information to reduce such risks. In this study, we analyze physical interaction based on mutual communication needs and build a model that will establish more sustainable relationships. We investigated the adjustment of the necessary communication needs to make a model in this experiment. We tested in a situation where we coordinated the communication needs using a mannequin. In this experiment, participants approached it, went around it, and started communication from a position that showed more intimacy to a partner when the partner's communication needs are low.

In multi-participant conversation, speaker and listener pay attention to all participants to carry out smooth turn-taking. However, in video conversation it is difficult to perceive the timing of when to take turns talking because participants cannot perceive each other's eye-gaze. Consequently, one participant often misses the sign of when another person wants to speak, and continues to speak longer than appropriate. Considering this problem, in this study we propose a video-conference environment in which a robot symbolizes and unifies the eye-gaze motion of several participants. As a result of an experiment to investigate its efficacy and usefulness, it was suggested that the proposed video-conference environment suppresses inappropriate speech compared with a general video-conversation environment.

Humans can communicate because they adapt and adjust their behavior to each other. We hypothesize that developing a relationship with others requires coordinating the desire to communicate and that this coordination is related to agency identification. To model this initial phase of communication, we created an experimental environment to observe the interaction between a human and an abstract-shaped robot whose behavior, moving on the floor and rotating, was mapped by another human. The participants were required to verbalize what they were thinking or feeling while interacting with the robot. At present, we do not have a sufficient number of participants, and experiments and data analysis are ongoing. We must verify the effects of interaction patterns and inspect what type of action and reaction are regarded as signals that enhance interpersonal interaction.

Humans can communicate because they adapt and adjust their behavior to each other. Developing a relationship with a robot, on the other hand, is difficult, especially with a robot that humans have never met before. However, little investigation has addressed interaction when no information about the interaction partner has been provided and with no experimental task. We must clarify how people perceive unknown objects as agents through interaction. We created an experimental environment to observe the interaction between human and a robot whose behavior was actually mapped by another human. The participants verbalized what they were thinking or feeling while interacting with the robot. The results of our experiment suggest that the timing of the movements was used as a cue for interaction development. We need to verify the effects of other interaction patterns and inspect what kinds of actions and reactions are regarded as signals that enhance interpersonal interaction.

Humans can communicate because they adapt and adjust their behavior to each other. Developing a relationship with an unknown artifact, on the other hand, is difficult. To address this problem, some robots utilize the context of the interaction between humans. However, there has been little investigation on interaction when no information about the interaction partner has been provided and where there has been no experimental task. Clarification of how people perceive unknown objects as agents is required. We believe that a stage of subconscious interaction plays a role in this process. We created an experimental environment to observe the interaction between a human and a robot whose behavior was actually mapped by another human. The participants were required to verbalize what they were thinking or feeling while interacting with the robot. The results of our experiment suggest that the timing of movement was used as the cue for interaction development. We need to verify the effects of other interaction patterns and inspect what kind of action and reaction are regarded as signals that enhance interpersonal interaction.

Humans estimate the emotions of others based on voices, expressions, and gestures, even though such expressions are comprised of many complicated elements. If a simple existence moves, how we estimate its emotional states remains relatively unclear. This paper investigates how humans suppose emotional states by the parameter changes of simple movements. We use a simple disk-shaped robot that only moves on the floor and expresses its emotional states by movements and movement parameters based on Russell's circumplex model. We observed the physical interaction between humans and our robot through an experiment where our participants were seeking a treasure in a given field and confirmed that humans infer emotional states by movements that can be changed by simple parameters. This result will contribute to the basic design of HRI.

This paper reports an analytic finding in which humans inferred the emotional states of a simple, flat robot that only moves autonomously on a floor in all directions based on Russell's circumplex model of affect that depends on human's spatial position. We observed the physical interaction between humans and a robot through an experiment where our participants seek a treasure in the given field, and the robot expresses its affective state by movements. This result will contribute to the basic design of HRI. The robot only showed its internal state using its simple movements.

Humans can interact with strangers because they can communicate with each other. On the other hand, developing a relationship with an unknown artifact is difficult. In order to address this problem, existing studies have explored various approaches to the artifact's behavioral design. However, little research has been done on interaction where there is no information about the interaction partner and where there has been no experimental task. Clarification of how people regard unknown objects as interaction partners is required. We believe that a stage of subconscious interaction plays a role in this process. We created an experimental environment to observe the interaction between a human and a robot whose behavior was actually mapped by another human. We observed this interaction under two different conditions; where the participants knew and where the participants did not know that the robot could interact. Both sets of participants approached or avoided the robot, but differences in the interaction property for each condition was confirmed. The results of our experiment suggest that a stage of subconscious interaction does exist for recognition of artifacts as interaction partners.

In this study, we suggest a method to coordinate turn-taking and talking in multi-party conversations by the gaze of a robot that participates on the side. Also, we use the experimental paradigm named “Cooperative Turn-taking Game in Non-verbal Situation”, which is a simplified multi-party conversation environment. We investigated whether designing eye-gazes for such robots can coordinate turn-taking and talking in multi-party conversations, and we found the robot's gaze could coordinate turn-taking and talking in multi-party conversations. Our study is expected to effectively encourage desirable talking in such multi-party conversations as collaborative learning scenes.

This paper investigates that human tend to percept of humanness and individuality to what features of human motion. We analyze from experiments based on test subjects how recognize of individuality by comparing a developed own self motion with other motion made by motion capture using Kinect sensor. The results showed except for the subjects could not perform the movement extremely distinctive, many subjects to recognize the movement of their own. In addition, the respondents also recognized with a video camera an objective of its own motion, the subjects of some of their own and do not match the image of the inside of me. It is difficult to percept humanness based on less-accurate expression of human motion.

In face-to-face communication, we can confidently communicate through our bodies. Recently, agents have widely surfaced as existences that interact with humans, and several studies have investigated the formation of social relations with such agents. This study focuses on the possibility that the impression of embodied agents and the way of interpreting information offered by them changes owing to different environments where humans and agents exist as well as the method through which instructions are given. We examined the effects of sharing an environment in two cooperative task experiments in real/virtual worlds. Our result shows that the reliability of the information provided by the agent is improved by sharing the same environment with the agent through the body. The method of giving the instructions must also correspond to the task environment.

Our study aims to determine how an agent should be designed to perform cooperative work smoothly. From previous experiments, it has been suggested that a human does not acknowledge an agent as a real participant. We considered how we could confirm the above possibility with a cognitive experiment. We designed an experiment with a game format to determine whether subjects would acknowledge an agent's action and existence if the experimental situation presents mutual purpose and timing of entities. The results of our experiment clearly show that work efficiency improves by the agent's behavior. The results of these experiments will influence the model of Human-Agent Interaction in the future.

This paper investigates how people attribute individual autonomy to a remotely operated robot. An experiment was conducted in which participants remotely operated a goalkeeper robot to defend its goal from the kicker robot. Participants were assigned to two types of experimental conditions. Participants assigned to the first condition watched video images that captured the motion of the kicker robot from behind the goal. Participants assigned to the second condition watched video images of the kicker robot from the position of the goalkeeper robot. The result suggests that people are not concerned with the avatar's embodiment but assume their own body when they are focused on the avatar's (robot's) situation.

Embodied interaction does not necessarily require direct face-to-face situation, but it can be organized by sharing the third party indirectly common to the participants. In this paper, an experiment is conducted to examine how two persons orally coordinate their intention and action in order to achieve a given experimental task. As a result, it is suggested that the co-embodied interaction efficiently works, even when the participants are grounded in different situations or have different views toward the target.

In this paper, we focus on co-embodied interaction between two persons who occupy a differentiated situation in the same environment. An experiment is conducted to examine how two persons orally coordinate their intention and action in order to achieve a given experimental task. As a result, it is suggested that the co-embodied interaction efficiently works, even when the participants are grounded in different situations or have a different view toward the target.

This study aims to establish an accessible interaction model that resembles barrier-free human communication, which is examined through the eye movement. It utilizes human communication as a social function in human-agent interaction. This has two advantages: (1) decreasing user's cognitive load to communicate with other people, (2) avoiding special training to access directly to computers. This psychological experiment examined the effect of CG eye movement of human-like agent on human social impression. In addition, this experiment presented a CG character to subjects. As a result, subjects were able to respond naturally to the eye movement aided by the CG humanlike agent

In the scene of a face to face conversation, people exchange not only verbal messages but also non-verbal information such as facial expressions, which express such mental conditions as their mood, emotion, or attitude. The purpose of this study is to examine the degree of sympathy affected by human toward an interactive agent displaying CG generated facial expressions through a psychological experiment. Our experiments have shown us that prosocial behavior was observed when, in response to a negative mood induced by horrible or frightful pictures unconditionally shown, subjects gave more assists to the agent whose facial expression expresses the mood congruous to that of the subjects

This paper examines whether humans perceive additional meaning in computer utterances with different voice prosody and change their decisions based on such interaction if the prosodic differences carry informational significance. We conduct a route selection experiment in which participants were asked to find a route to a goal in a 3-D maze generated by computer graphics. The maze system occasionally outputs a confirmation in response to the participant’s choice of a route at a junction. The prosodic characteristics of the confirmation utterances reflect whether the route selected is right for reaching the goal in a normal alley or wrong, leading to a dead end. In this experiment, participants are able to recognize the prosodic meaning and successfully make a decision that leads to a route through the maze. This result suggests that participants can perceive the differences in pitch range of the confirmation voice, and used the pitch range in reselecting their route. Furthermore, wide pitch range confirmation voice indicates a dead end resulted in a higher reselection ratio. This suggests that participants gave a negative interpretation to wider pitch range responses.

We discuss the design of a life-like agent interface that considers the social aspects in human-agent interaction. The underlying hypothesis of the design is that human social behaviors toward life-like agents are on a par with those toward humans. Humans tend to sympathize with and follow other humans based on their affiliation needs. Therefore, it should be with life-like agents. We have designed and incorporated life-like guide agents into an exhibition guide system based on the hypothesis. The system was used at the 12/sup th/ ATR Open House, where we had the opportunity to study the behaviors of its users. We observed that the behaviors of the users were directed toward maintaining their good social relationships with their life-like agents, thereby supporting our hypothesis. The results indicated the importance of incorporating an agent's capability to induce the effect of human affiliation needs in a system design.

The paper presents an evaluation method to assess impressions toward a computer-generated creature through human-computer interaction. The aim of our study is to construct a mechanism on a computer to realize empathic interaction between humans and creatures. In this paper, we examine the e ect on personality cognition of a virtual creature when the creature replies with di erent timing to a human utterances. The creature has an abstract appearance and a non-linguistic voice. We control the following three conditions for timing of the creature's response: a) alternation, b) overlap, and c) neutral condition. As a result, we found that the evaluation value of the alternation condition shows more positive behavior than that of the overlap and neutral conditions when the creature has the above features.