This study investigates the influence of agents’ behavioral characteristics on human decision-making and the manner by which humans evaluate and treat agents, specifically in situations where they interact with agents adopting different strategies. Based on the ultimatum game with egocentric, exocentric, and adaptive agents, the findings suggest that inequity aversion is most likely to be expressed when the agent adopts an egocentric strategy. Human decision-making in interactions with agents is shaped not only by a preference for fairness but also by a self-serving tendency to prioritize one’s own benefits, provided that the opponent accepts the proposal. Furthermore, when the distribution is more favorable to the participants, they tend to evaluate the agent’s intelligence more negatively.

Using the Miracle Question method, this study comparatively evaluated how older adults and university students were affected by interacting with an active listening counseling robot for three weeks. The research focused on two aspects: (1) evaluating changes in perceptions of the robot and (2) how conversational content influenced these impressions. The results showed that older adults initially rated the robot higher for Anthropomorphism, with increased ratings for Perceived Intelligence and Anthropomorphism by Week 3. Both groups rated Perceived Safety the highest, linked to the prevalence of positive remarks during the counseling session conversations. These findings highlight the potential of positive dialogue strategies for counseling robots to engender trust and emotional safety across these age groups.

This paper presents collaborative concept-mapping tutor (CoCot ver.2), a collaborative learning support system that integrates concept maps with a conversational agent. CoCot ver.2 features two agents: a teacher agent and a student agent. The teacher agent acts as a human instructor, engaging in conversations with learners, aiding their metacognition, and summarizing the discussion content. The student agent learns from the learners' concept map creation and generates their own concept map knowledge. These agents are developed using (1) a cognitive architecture (ACT-R) for knowledge generation for the agents' concept map and (2) GPT 3.5 for part of the language processing for agent-based feedback.

This study presents a systematic literature search and narrative meta-review of the current state of research on conversational agents (CAs), including embodied CAs, chatbots, and social assistive robots (SARs).The investigation identifies 1,830 academic articles, of which 315 articles satisfied the inclusion criteria for the review.Systematic reviews across various fields are reported, including mental disorders, neurodevelopmental disorders, dementia/cognitive impairment, other medical conditions, elderly support, health promotion, mental health, education, industrial applications, agent characteristics, and robot characteristics.The study highlights challenges in current CA research, such as the scarcity of high-quality comparative studies and the acceptance of CAs by users and caregivers, particularly in elderly support.The article also categorises ethical discussions into nine elements: privacy, safety, innovation, user acceptance, psychological attachment, care philosophy, evaluation, social systems compatibility, and rule development.It also offers insights into the development of future guidelines.The role of CAs in fostering human relationships through their conversational function is emphasised to provide guidance for subsequent CA research and social implementation.As advancements in CA technology and research continue to progress, there is an increasing demand for sophisticated psychological investigations addressing relationships, emotions, and the self.

In this report, we investigate the factor of social presence of a robot by using an actual robot and comparing it with a CG robot studied in our previous study. A laboratory experiment is conducted using a simple jury decision-making task, where participants play the role of a jury and make decisions regarding the length of the sentence for a particular crime. During the task, a robot with expert knowledge provides suggestions regarding the length of the sentence based on other similar cases. Results show that participants who engaged with an actual robot showed higher conformity with the suggested length of a sentence compared to the participants who engaged with a CG robot presented through a computer monitor. This study shows results that are consistent with those of previous studies in that interacting with physically aware robots is more engaging and also shows its effects on decision-making in a court.

Studies in collaborative learning have shown that explanation activities drawing on diverse perspectives facilitate deeper understanding and metacognition. However, it is difficult to develop an explicit understanding of others’ perspectives and knowledge through communication in a computer-mediated environment. The present study investigated the use of a visually shared concept map interface, expected to facilitate dyadic awareness of different perspectives and thus improve learning performance during explanation activities. In this study, each dyad built a concept map about a key technical term in psychology, and generated explanations of the term and generated explanations of the term on mutually accessible concept maps. We predicted that learners would be able to (1) gain deeper knowledge through the shared explanations; and (2) explain the key term from different perspectives or knowledge sets. Twenty-six university students participated in this experiment, and we assessed their performance through free recall tests before and after they used the concept mapping tool. Our findings showed that learners were able to (1) gain learning performance and (2) explain a concept based on different perspectives. We discuss the implications of our findings and suggest directions for further research on the development of learning support systems.

The purpose of this paper is to examine the influence of lexical entrainment while communicating with a conversational agent. We consider two types of cognitive information processing:top-down processing, which depends on prior knowledge, and bottom-up processing, which depends on one's partners' behavior. Each works mutually complementarily in interpersonal cognition. It was hypothesized that we will separate each method of processing because of the agent's behavior. We designed a word choice task where participants and the agent described pictures and selected them alternately and held two factors constant:First, the expectation about the agent's intelligence by the experimenter's instruction as top-down processing; second, the agent's behavior, manipulating the degree of intellectual impression, as bottom-up processing. The results show that people select words differently because of the diversity of expressed behavior and thus supported our hypothesis. The findings obtained in this study could bring about new guidelines for a human-to-agent language interface.

Designing embodied agents that are empathic and positive towards humans is important in Human Agent Interaction (HAI) and design factors need to be instigated based on experimental investigation. Agent design specificity, in which less specific animated designs are better than realistic designs, is one of the key factors that facilitate positive emotions during interactions. Focusing on this point, this study investigated the effects of a deformed embodied agent during a collaborative interaction task with the objective of understanding how subjective interpersonal states and emotional states change when deformed embodied agents are used instead of non-deformed agents. This was accomplished by developing an interactive communication task with the embodied agent and collecting subjective and emotional state data during the task. The results obtained indicate that deformed agents evoke impressions of closeness and produce higher arousal states.

Studies in collaborative learning show that learners with fewer social interaction skills have difficulty during collaborative activities with others. The present study investigates how such a learner's skills influence the performance of an online concept-learning tutoring task with a Pedagogical Conversational Agent (PCA). During the task, the learners were guided by a PCA that helped them formulate their explanations of a key concept taught in a large-scale class. The study assesses the degree of learner's social skills based on the autism spectrum quotient (AQ) and investigates how this level skills influences learning with the PCA. Results show that learners with higher social skills performed better on explanation activities with the PCA. This shows that social cognitive skills influence learning performance during communication-based learning support systems. The results of this study suggest that there is a need to design efficient online tutoring systems that consider a learner's social skills.

The goal of this study was to investigate the efficient use of role-taking embodied conversational agents in the facilitation of creative cognition during collaborative activities. Two factors were investigated through an experimental design addressing the number of conversational agents (one vs. two), and method of communication (voice vs. text). Participants engaged in a simple interpretation game with embodied conversational agents. The agents made suggestions on the quality of the participants' interpretations. We investigated how the two factors enhanced the quality and quantity of interpretations in collaborative activities. Results showed that the use of single agents and text-based interfaces enhanced the quantity of the creative interpretations, and the synergy created by the use of multiple agents along with a voice communication method enhanced the quality of creative interpretations. These results suggest that the number of agents and the method of communication are important in the effective use of embodied conversational agents as collaborating peers.

In designing pedagogical agents, it is important to understand what factors stimulate the learner's affect to enhance learning motivation. To do so, the present study investigated the influence of affective mood on gender types of the (1) learner and the (2) pedagogical agent through an online tutoring activity. Affective states were ascertained using a questionnaire constructed on the basis of Russell's (1980) two-dimensional affective model data for 16 times during the tutoring activities. The participants were 290 psychology students, who were made to perform a homework activity. The results of two experiments consistently revealed that male students were sensitive to the pleasantness bipolar and female students were sensitive to the activation bipolar.

Since neural networks have superior information processing functions, many investigators attempt to model biological neurons and their networks. Furthermore, a number of studies of neural networks have recently been made with the purpose of applying engineering to the brain. In this study, we investigate the effect of STDP on the ability to extract phase information buried in fluctuation. We focus on spike timing dependent synaptic plasticity (STDP), and we construct neural networks from a pulse-type hardware neuron model using STDP. We show that phase information buried in fluctuation is revealed by the effect of STDP, making it possible to decode the synaptic weight. Moreover, we show that it is possible to extract the phase difference buried in fluctuation representing the reinforcement part of the synaptic weight, using neural networks with STDP.