This study proposes an AI system that supports expert assessment by utilizing Generative AI to conduct a multiperspective analysis of factors underlying task non-attainment in developmental assessments, and preliminarily examines its behavior and effects on experts. Conventional assessment methods tend to record a child’s failure to complete a task simply as a “lack of ability.” In practice, however, there are cases where children possess latent abilities but fail to perform due to temperament, anxiety, or environmental factors. Overlooking this “Competence-Performance Gap” carries the risk of leading to inappropriate diagnoses and interventions. In the proposed system, the AI generates hypotheses regarding “why a task was not attained” from the perspectives of developmental psychology, early childhood education, occupational therapy, and speech-language pathology, thereby complementing the expert’s interpretation. We position this work as an exploratory case study (n=6 evaluations; 2 experts × 3 video clips from ESCS sessions), designed not to confirm false-negative reduction directly but to generate indirect process-level evidence through proxy indicators (AI-expert judgment agreement, hypothesis count change, confidence shift, and category shift of hypothesis labels). The observed patterns suggest that the AI and the experts held complementary perspectives. While the AI support did not alter the final judgments themselves, it was accompanied by consolidation of hypotheses (mean 3.3 → 2.0) and slightly elevated confidence in the majority of cases. Conversely, the issue of “over-interpretation” emerged as a challenge for the AI. It was also suggested that the detailed behavioral descriptions, introduced to suppress hallucinations, might have inadvertently encouraged the AI to focus excessively on localized actions. These findings are hypothesis-generating rather than confirmatory; direct verification of false-negative reduction requires a subsequent larger-scale study with longitudinal ground truth.

In general online voice communication environments, participants are not assumed to engage in multiple conversations simultaneously. As a result, opportunities to overhear or participate in other conversations, as well as opportunities for new conversations to emerge during ongoing interaction, are limited. Although many studies have explored online dialogue environments that emulate face-to-face situations using avatars or virtual spaces, relatively little attention has been paid to the discrete network structures inherent in online dialogue environments and their potential as resources for dialogue design. The purpose of this study is to develop an online dialogue environment that enables multiplex participation across multiple dialogues and to explore the characteristics of interaction unique to dialogue environments based on graph structures. We developed an online dialogue environment called Plexa based on a model that represents the participation structure proposed by Goffman (1981) as a graph, and conducted an experimental study involving groups of three participants. The results showed that in all groups, participants within a dialogue space brought external participants into that space based on their own judgment―an interaction that is rarely observed in dialogue environments grounded in physical co-presence. These findings suggest that online dialogue environments based on graph structures can give rise to interaction patterns that are distinct from those found in face-to-face situations.

This study investigates how positive feedback from a trainer agent affects mood and motivation during multi-task training that combines walking and a computerized facial-expression reaction task. Twenty-four older adults (12 women; mean age 71 ± 6.39) completed two weekly sessions (45 minutes each, three 15-minute blocks per session) using a “Walk training robo” equipped with a display for the cognitive task and a tablet-based virtual trainer. We compared a With Feedback (WF) condition―agent greetings plus in-task and post-task positive reactions tied to performance and time prompts―against a No Feedback (NF) condition that retained neutral messages only. Mood, impressions of the agent, and motivation were assessed each week. Wilcoxon signed-rank tests showed that WF increased negative mood items (e.g., “irritated,”“sad”) relative to NF and reduced perceived “ professionalism, ” while increasing training-related satisfaction (trend) and intention to continue (significant). Observations suggested that repeated upbeat reactions from the trainer agent during errors may have conflicted with participants’ self-evaluation, eliciting frustration even as moment-to-moment prompts fostered persistence. These findings indicate that, in cognitively demanding multi-task contexts, overly positive but weakly calibrated feedback can harm affect and perceived expertise, yet still bolster motivational outcomes. Designing feedback that balances positivity with situational accuracy may optimize both affective and motivational effects for older adult training.

Remote asynchronous architectural education offers significantly limited opportunities for peer learning among students. This study positions virtual-tour-based VR as a platform for integrating diverse learning content in asynchronous environments, implementing an environment that combines 3D model displays, drawing panels, and dynamic contents (presentation and critique videos), and reports on its educational effects through two years of operation. An online exhibition of outstanding student works was recreated in a VR space, enabling learners to explore each content while navigating the space. Based on first-year findings, the 3D model display method was transitioned from Object VR to 3D scanning, and dynamic content was added in the second year. Questionnaire surveys indicated that the 3D scanning method was positively evaluated in terms of freedom of viewpoint, and the dynamic content was positively evaluated for supporting understanding of design intent and evaluation perspectives. A total of 70% of respondents gave positive evaluations for enhanced understanding through integrated viewing of multiple contents. As an implementation case study without controlled comparison with alternative content delivery formats, this study presents practical knowledge for utilizing web-based virtual tour VR as an integrated platform for peer learning in asynchronous architectural education.

Efforts to promote bicycle use are being undertaken worldwide. In Japan, the rate of bicycle-related accidents and the number of deaths per population are higher compared to other countries. To reduce bicycle accidents, it is important to identify near-miss incidents from the perspective of bicycle users. This paper developed categories of near-miss incidents from the perspective of bicycle users by utilizing bicycles equipped with dashcams. It also collected and analyzed data on bicycle near-miss incidents. Additionally, bicycle near-miss incidents were categorized by road structure - such as singlelane sections and intersections - and their incidence trends were compared. As a result, short-passing distance overtaking on road sections and head-on collisions at intersections accounted for the highest proportion.

Resilience is increasingly discussed in medical safety as the adaptive capacity of organizations and individuals to restore normal operations after unexpected events. Prior research has largely highlighted its benefits, but the burdens of maintaining resilience remain understudied. Semi-structured interviews were conducted with 14 experienced staff members (≥5 years in the night emergency department; ≥3 years of night duty). Using the Functional Resonance Analysis Method (FRAM), the work system was modeled as a set of interacting functions. Next, 33 night-duty nurses rated each function on resource availability, adaptive behavior required, and associated anxiety. FRAM identified 15 functions in the nighttime emergency process. Ratings showed that a subset of functions carried disproportionately high burdens, were performed by only some nurses, and demanded greater adaptation and higher anxiety than other functions. Overall, resilience in this setting appeared to depend on concentrated adaptive efforts at specific upstream functions in the workflow.

To enhance students’ learning, various learning cycles have been proposed to date. Some previous studies have suggested that learning cycles and classroom environments are related to the promotion of students’ learning outcomes and learning motivation; however, these approaches have not been sufficiently generalized. To address this issue, this study focuses on classes conducted based on The Enhanced Learning Cycle proposed at Musashino University and implemented in a classroom with a freely reconfigurable layout. We attempted to evaluate this learning environment by visualizing conversational interactions among participants. The results showed that conversations involving three or more students occurred during the analyzed scenes, and that interactions were not limited to adjacent seats but also took place between students separated by two or more seats. In these scenes, the instructor conducted the class while engaging in conversations with all students. These findings suggest the usefulness of conversational interaction as a variable for estimating the relationship between instructional activities and classroom layout.

This study examines collaboration among two technical high schools, two special needs education schools, and a university under Japan's DX High School initiative. Through change agentry, the university introduced expertise and the social robot LOVOT, which acted as a boundary object, bridging institutional differences through its interpretive flexibility. Unlike consortium models where the university controls activities continuously, the stirring model presented here encourages schools to develop autonomous practices after the initial intervention. The stirring model illustrates how modest, timely inputs can reorganize networks and support sustainable, autonomous collaboration.

In agent-based sales promotion within retail stores, balancing customer attraction with effective attention guidance during information presentation remains a critical challenge. This study investigates the impact of an “adaptive embodiment of an agent”―a Mixed Reality (MR) agent that dynamically adjusts its presence based on the interaction context―on user behavior within a physical space. Specifically, we examined the effects on user behavior of a behavior design in which the agent is physically present to attract customers at the initiation of the interaction and subsequently disappears to present information while directing attention to the product during the sales promotion. Results from a comparative experiment conducted in a retail store confirmed that this approach compensates for the drawbacks of static embodiment (constant presence or absence), enabling the simultaneous achievement of customer attraction and attention guidance. Additionally, a post-hoc exploratory analysis focusing on parent-child groups indicated that parents’ screen-gazing duration tended to be higher in the adaptive embodiment condition, providing preliminary insights for future research on the effects of dynamic embodiment in multi-party interactions.

Although educational robots have proven effective for children’s learning, how to integrate them into home environments without interrupting parent-child interaction is still unclear. We addressed this gap by proposing a semi-autonomous paraoperated robot designed to be partially controlled by a parent. In contrast to autonomous robots, which may reduce effort at the cost of engagement, we posited that active operation would foster parental involvement. We compared three experimental conditions (no robot, an autonomous robot, and our semi-autonomous robot) among parent-child dyads. Our findings demonstrated that the proposed method significantly increased parental verbalizations compared with the autonomous robot condition, without adding to their perceived workload. These results provide evidence that semi-autonomous para-operated robots are a viable and effective tool for facilitating interactive home learning.

This study examined how different media, embodied robots and immersive VR (Virtual Reality), affect psychological stress, future-oriented discourse, emotional expression, and agent impressions in Solution-Focused Brief Therapy (SFBT) based active listening counseling for older adults through a three-week longitudinal study. Thirty-four community-dwelling older adults (mean age: 69.30 years) were assigned to either robot (n=18) or VR (n=16) conditions and participated in weekly dialogue sessions centered on Miracle Questions. Results showed that psychological stress significantly decreased over time in both conditions, with no significant differences between conditions. Future-oriented discourse and positive emotional expression exhibited different temporal patterns: VR showed early advantages, while robot conditions increased in later weeks. Regarding agent impressions, anthropomorphism and animacy improved in both conditions, while perceived intelligence was rated higher in VR. Under the experimental conditions of this study, these findings suggest that VR and robots facilitate future-oriented discourse and emotional expression through distinct pathways. The results suggest that, under the experimental conditions of this study, social presence, information richness, and embodiment of media influence SFBT effectiveness through different psychological processes. This study provides design guidelines for conversational support systems for older adults.

Individuals with developmental disabilities who experience communication difficulties often face challenges in conveying their needs to healthcare providers during time-constrained medical interviews, which can impede access to appropriate medical care. Although patient participation in healthcare has become increasingly important, communication barriers may limit such involvement. This study employed semi-structured interviews with nine individuals experiencing communication difficulties and their regular support providers. Narratives were analyzed using an approach informed by thematic analysis and organized around three perspectives: "medical interviews," "expectations for medical care," and "patient participation in healthcare." The findings highlighted a multilayered structure of difficulties in medical interviews, including challenges in recalling, verbalizing, and self-regulating expression under time constraints; multiple meanings of medical care beyond symptom treatment; and the importance of support providers as "translators" mediating between patients and clinicians. Drawing on these experiences, the study discusses design strategies to foster collaboration between healthcare and welfare services. Proposed strategies include mechanisms for externalizing and summarizing concerns prior to appointments and interfaces that facilitate supplementary or proxy explanations by support personnel, redesigning patient participation as a relational and structural process rather than relying on individuals' linguistic competence.

This study investigates how response timing and nodding motion in gaze-driven interfaces affect impression and empathetic feedback with childcare support discussed as one potential application. Two factorial experiments were conducted: Experiment 1 manipulated response timing (0ms, 300ms, 1000ms, and no response) to evaluate its impact on impressions such as familiarity, activeness, and sincerity, as well as empathy and reassurance. Experiment 2 examined how different types of nodding motions influence the perception of response timing and overall impression. Results showed that a 300ms response delay was rated highest for empathy and reassurance, while activeness was rated highest for immediate (0ms) responses. The physical characteristics of nodding motions (depth, frequency, duration) significantly affected impression ratings, with certain motions perceived as more familiar or sincere. Importantly, the results suggest that impression characteristics may mediate the effect of nodding motion on empathetic feedback, suggesting that design should prioritize the user’s perceived impression rather than simply optimizing motion parameters. These findings provide foundational knowledge for signing low-burden, empathetic nonverbal interactions, with childcare support interfaces as a representative application, emphasizing the importance of combining optimal response timing with nodding motions that evoke positive impressions.

Help Mark and Maternity Mark are pictograms promoted and distributed by prefectural and municipal governments, as well as the Ministry of Health, Labour and Welfare. These marks are designed to let others know that the bearer requires social consideration or assistance. For example, they encourage actions like offering a seat on public transportation. However, a survey [3] revealed that 78.7% of respondents have experienced hesitation when deciding whether to offer their seat on a train. Another survey [4] found that 67.3% of train passengers spend their time using smartphones. To address these situations, this paper proposes and develops an application called Digital Help Sign. The purpose of this app is to encourage helpers to take supportive actions. The app allows users of Help Mark or Maternity Mark to signal their need for assistance through their smartphones. This signal is then received as a notification by potential helpers nearby. The application uses Bluetooth Low Energy technology and is designed to operate while protecting personal information. This paper evaluates the utility of the Digital Help Sign application.

In text-based chat, where nonverbal cues are absent, it is difficult to accurately interpret the speaker ’s intentions and emotions. In contrast, during face-to-face communication, the speaker ’s emotional state can often be inferred from their speech rate. This study introduces sequential text output in text chat, where characters are displayed one at a time to represent speech rate. To verify the impression and impact of this approach, we conducted two experiments: one examined whether such sequential character output is perceived as reflecting the sender ’s speech rate, and the other examined whether variations in output speed influence message interpretation. The first experiment showed that sequential text output at a constant speed with temporary pauses was more likely to be perceived as representing speech rate compared to output with only sequential display or output mimicking actual speech rate. In the second experiment, using a chat system incorporating constant sequential output and pauses, we investigated how variations in output speed affect emotion perception. The results showed that faster sequential output speeds led to perceptions of higher emotional arousal, while emotional valence was not significantly affected. These results suggest that output speed mainly influences perceived emotional arousal rather than emotional valence.