With anxiety disorders on the rise in recent years, early detection and assessment of anxiety states is extremely important. Conventional approaches have mainly used self-reported questionnaires, but they have reliability problems. Therefore, this study aims to generate facial images that include minute changes in facial expression using deep learning techniques to more accurately and simply assess subjects’ anxiety states. This method quantitatively evaluates the degree of anxiety state by analyzing the subject’s facial expression recognition of the generated images. The validity of this method was verified through correlation analysis with the results of the State-Trait Anxiety Inventory (STAI), an existing anxiety assessment tool. The results showed a significant correlation between the facial expression recognition score and the STAI score, indicating that this method is effective in assessing anxiety states. This study aims to innovate the evaluation method of anxiety state by integrating deep learning and psychology.

In this study, we developed a prototype of a blink-based communication device called the “Blinkcommunicator” to verify the feasibility of creating a simple and practical system for selecting picture card messages focusing on blinking. Users of this device blink to select a target from among eight rapidly flashing options (picture cards). The blink-related changes in eye potentials are detected in real-time using a pattern recognition method (referred to as the “blink switch”). Assuming this system as a cognitive task, we investigated whether participants, consisting of 14 healthy individuals, could accurately select the target specified by the blinking switch (target decoding accuracy based on pattern recognition). As a result, the patterns of blinks associated with task performance (particularly peak amplitude and latency) were consistent within individuals, leading to a remarkably high decoding accuracy averaging over 95%. This further enhances expectations for the practical application of this system.

Texture is considered to be one of the factors that determines the perception of palatability. In this study, samples of milk chocolate with different particle size were used to evaluate the subjective well-being perceived from the quality of the chocolate, and EEG and autonomic nervous system were measured as physiological indices during and after consumption of the samples. The results showed that subjective well-being increased with consumption of both samples, which was higher for the finer particle size sample. In addition, differences in postprandial EEG and autonomic indices were also observed depending on the particle size of the milk chocolate. These results suggest that the smooth texture of milk chocolate may have influenced the difference in subjective well-being, which could also be captured by physiological indicators.

We conducted laboratory and field experiments to evaluate the road rage prevention effectiveness of using a “You are being recorded” sticker on a private passenger vehicle. For our first laboratory experiment, 30 college students viewed photographs of three types and rated the depicted vehicle’s likelihood of being a victim of road rage. Results indicated that cars with a “You are being recorded” sticker were regarded as less likely to be victims of road rage than cars with no sticker or cars with an “After you” sticker. Then for a field experiment, we asked 10 college students to use the stickers on their own private vehicles to assess their effectiveness at preventing road rage. Results suggest that cars with a “You are being recorded” sticker, compared to cars without stickers, have some effect on the distance of following vehicles and on feelings of safety while driving.

The aesthetic appeal of a shape is a critical factor in product evaluation. Previous research has suggested that aesthetic evaluations of shapes improve when individuals recognize additional regularity. In this study, we hypothesized that this enhancement in aesthetic evaluation is attributed to the interest sparked by the transition in the recognition process. We conceptualized this hypothesis using the free energy principle, grounded in dual-process theory. Furthermore, to address the motivation behind considering additional regularity, we formulated concepts of both the surprise elicited by the additional regularity and the cognitive effort required to recognize this regularity. To test the validity of these constructs, we conducted experiments with various combinations of primitive shapes. Our results revealed that both our aesthetic indicator and the formulated surprise construct were significantly positively correlated with subjective evaluations upon recognizing additional regularity. Leveraging the findings of this study could enable the design of shapes incorporating additional regularity, thus enhancing aesthetic appeal or creating captivating surprises.

In this paper, we introduce the concept of altruistic regulation through active inference. Our approach involves incorporating models of others and minimizing the combined expected free energy of oneself and others. We apply this concept to agent passing interactions to investigate its impact on fluency. The results demonstrate that this approach enables the estimation of others’ purposes and the generation of altruistic behavior. Furthermore, we show that the proportion of summation regulates the balance between altruism and selfishness. Simulation results suggest that moderate altruism contributes to minimizing the time required to reach a goal and indicate that the appropriate degree of altruism varies depending on the disparities in abilities and environmental factors among agents. These findings highlight that the degree of altruistic behavior can be adjusted through the weighting of summed expected free energy. Moreover, this adjustment impacts the fluency of agent passing interactions.

To extract features of the listing data for purchasing motivation of potential buyers, we focused not only on textual information but also on product images. The target of this study was shoes and we performed a predictive analysis using XGBoost with 1,000 items of past listing data. The dependent variable was whether the item was sold or not, and the explanatory variables were 17 features extracted from the basic information of the item, the characteristics of the item description, and the characteristics of the item image. The results were showed that it was important for preparing item images to have a large number of item images, to use a single background color and to photograph shoes at a 45-degree angle. The results of a questionnaire survey with 240 people revealed that the listing information reflecting the extracted characteristics enhance the purchasing motivation of infrequent users of flea market apps.

In recent years, the usefulness of rough set theory has been gaining attention in Kansei engineering. The authors defined “core” as combination of attribute values that appear frequently based on decision rules and developed a notation method that comprehensively consolidates decision rules containing core. However, there were challenges such as the difficulty in interpreting the importance of attributes other than core and the impact of each core on the results. Therefore, this study aimed to develop a notation system for decision rules based on rough sets with TF-IDF applied to solve these challenges, and a system to visualize the structure of decision rules containing core using formal concept analysis. Furthermore, this system was applied to analyze attribute data of individual investors and its usefulness was verified. As a result, significant attribute values could be discovered, and cores could be classified into three types based on their impact on the results.

The composition has a strong influence on the overall of the work in the quality of arrangement and the impression. While the type of composition makes an impression, the beginners may struggle to place the human subject in the initial stage of sketches. Therefore, we propose a method for the improvements to the placement of the human object in an illustration at the initial sketch stage in this paper. We implemented a system to generate the output image in two basic compositions by using human objects in based on the object detection method by using the deep learning. We conduct quantitative experiments to show the effectiveness of our method.

In recent years, the demand for a technology that estimates how much interest users have for an article in social media is growing because the number of marketing cases using social media is increasing. Especially, to distribute the advertisements on social media that match the interests of each user, a technology that forecasts how a user would react/behave for the post when s/he looks at a post are required. This paper develops a method that forecasts whether a user bookmarks a post that s/he would look at, by analyzing the posts that the user has bookmarked in social media based on Gradient Boosting Decision Tree that is one of the ensemble learning methods in artificial intelligence. Moreover, we evaluate the performances of multiple forecast models (XGBoost, LightGBM, and CatBoost), and analyze the factors that cause users to bookmark a post.

We propose an interactive design system that optimizes the color scheme of a bouquet. The optimization is performed by Interactive Evolutionary Computation, which repeats presentations of candidate designs to a user, evaluations of their scores by the user, and creations of new candidates based on the evaluations. For dealing with the affective information of a user, we believe that how the candidate solutions can appeal to the user’s sensibility is an important issue. Therefore, to provide candidate design images more appeal power, the illustrated designs are transformed into the realistic ones by using the image generative AI, Stable Diffusion. To examine the effectiveness of our system, we performed experiments on 60 subjects. We found our proposed system had an advantage that the user can image the realistic designs clearer during the optimization process.