Recently, it has become increasingly important to evaluate in advance the implications of AI systems on stakeholders and to take appropriate measures accordingly. In this study, we aim to identify ethical risks of an AI system used for employee dialogue training through a questionnaire survey of 12 evaluators and a developer. As a result of analyses of qualitative responses, several issues were identified, including the influence of confirmation bias, methods for explaining the effectiveness of dialogue training, and the pros and cons of using dialogue training results in personnel evaluations. By conducting a follow-up questionnaire on these countermeasures, we examined the safety of the system. In addition, through this study, process-related challenges in the risk management process, particularly those concerning questionnaire responses and aggregation methods, were identified.

This study aimed to elucidate the characteristics of upper-limb movements during touching interpersonal shoulder touching to convey emotions. Twenty-two young adults participated in a task in which they touched the shoulder of a person standing in front of them with their palm under five emotional conditions: “happiness”, “soothing tension”, “easing sadness”, “greeting”, and “initiating contact”. Three-dimensional motion analysis revealed emotion-specific features in the joint dynamics of the shoulder, elbow, and wrist, such as the magnitude and timing of the peak flexion and extension. Particularly, the happiness condition presented with a longer preparation phase, while slower and more gentle movements were observed in the easing sadness and initiating contact conditions. These findings deepen our understanding of natural human nonverbal communication and suggest potential applications for implementing empathic touch behaviors in caregiving and educational robots.

This study aims to clarify the effects of back recline angles （sitting， semi-sitting, and supine positions） on VR interaction characteristics, in response to the growing demand for VR use in reclining positions, such as VR sleep. An experiment was conducted using a Meta Quest 3, in which participants performed a pointing task that required clicking targets distributed omnidirectionally within the virtual environment. Reaction time, head movement, and subjective comfort were evaluated as performance measures. The results of Experiment 1 revealed that reaction times were significantly delayed in the supine position compared to the sitting position, with a particularly pronounced decline in operability for targets located in the lower visual field. Experiment 2 examined short-term practice effects under the supine condition; however, no substantial improvement was observed in reaction time or head rotation magnitude. Based on these results, this paper discusses the necessity of optimizing user interface layouts and alternative input modalities according to posture, and design implications for VR interface design specifically for the supine position.

In this study, we propose and evaluate a motion control method inspired by the “entrainment phenomenon of breathing,” aiming to enhance incabin comfort during future autonomous highway cruising. Specifically, pitch motion designed to reduce the perception of vertical base vibrations was applied at a frequency set to 0.8 times the occupant’s breathing frequency. As a result, we observed an increase in the highfrequency （HF） component of heart rate variability. These findings suggest that active seat control synchronized with breathing rhythms may contribute to improving physiological comfort.

This study investigated limitations in depth perception characteristics in virtual environments using head-mounted displays (HMDs) by comparing visual behavior between real and virtual environments. Twelve healthy university students participated in experiments conducted under identical visual stimulus conditions in both environments. Participants performed a visual recognition task using targets presented at different distances. Binocular disparity during fixation on the visual target and reaction time for target recognition were used as evaluation indices. The results showed that in the real environment, binocular disparity tended to change with increasing viewing distance. In contrast, in the virtual environment, no clear change in binocular disparity was observed across different presented distances. In addition, reaction time for target recognition tended to increase with viewing distance in the virtual environment, whereas it remained relatively stable in the real environment. These findings suggest that in HMD-based virtual environments, changes in presented distance do not sufficiently function as physiological depth cues, and the visual system may rely more on alternative cues such as apparent size to perform the task. This study suggests an ergonomic limitation in interpreting perceptual and behavioral data obtained from HMD-based virtual environments and emphasizes the need for caution when applying such data to the design and evaluation of real-world systems.