We report a mixed-methods field experiment of a conversational service robot deployed under everyday staffing discretion in a live bedding store. Over 12 days we alternated three conditions--Baseline (no robot), Robot-only, and Robot+Fixture--and video-annotated the service funnel from passersby to purchase. An explanatory sequential design then used six post-experiment staff interviews to interpret the quantitative patterns.

In recent years, the demand for social robots has grown, requiring them to adapt their behaviors based on users’ states. Accurately assessing user experience (UX) in human-robot interaction (HRI) is crucial for achieving this adaptability. UX is a multi-faceted measure encompassing aspects such as sentiment and engagement, yet existing methods often focus on these individually. This study proposes a UX estimation method for HRI by leveraging multimodal social signals. We construct a UX dataset and develop a Transformer-based model that utilizes facial expressions and voice for estimation. Unlike conventional models that rely on momentary observations, our approach captures both short- and long-term interaction patterns using a multi-instance learning framework. This enables the model to capture temporal dynamics in UX, providing a more holistic representation. Experimental results demonstrate that our method outperforms third-party human evaluators in UX estimation.

Promotional media, such as paper posters and digital signage, are installed in shopping malls to recommend products and services. However, it has been reported that many customers tend not to be interested in these promotional media and do not receive the information. When product information is not communicated effectively, advertisers are unable to convey the information they wish to share with customers, and customers miss the opportunity to receive valuable information. To address such issues, a lot of methods have been proposed to make people aware of the presence of media; however, there are not many methods that take into account the delivery of product information to customers. In this study, we propose Popping-Up Poster, a pin-based poster device designed to capture customer attention and convey information through dynamic shape changes. To verify the effectiveness of the proposed system, field experiments were conducted in a café, where its promotional effects were compared with those of traditional promotional media, including paper posters and digital signage. These results show that Popping-Up Poster has the potential to be more effective in recommending products and influencing customer product choices compared to conventional promotional media.

Research on self-recommending product agents is under way. Compared with product promotion by humanoid robot agents, self-recommending agents can call a passing cus-omer's attention to product. As the customer's interest is focused on the product, the self-recommending agent can then give instructions, such as “pick me up.” It has been found that customers tend to follow such instructions, and touching the product is known to effectively support sales promotion. From two experiments in a convenience store, a self-recommending agent in this study was successful in attracting customer interest, which included handling the products. However, we also found that, after being picked up, the products' monologue resulted in customers leaving the product behind on the shelf. Herein, we examine the reasons why.

Robot agents are increasingly used for user services, and society is becoming increasingly familiar with such robots. Most robots are of the humanoid type, which are easy to recognize as interaction partners. By interacting with a user, these agents may be able to generate user interest in a product and successfully sell it. However, according to previous studies, it has been suggested that users who are interested in the movement and appearance of the agent may focus only on these aspects and not listen to the recommendation. Therefore,we hypothesized that by making the product the agent, which we call a Self Recommendation Agent(SRA), the attention of the user would be focused on the product itself. Hence, if a user is interested in the agent, it is the same as paying attention to the product. Therefore, we expect that such an agent will be able to gain more attention from users than conventional agents while providing recommendations. To investigate the effectiveness of this agent, we set up a store in a shopping mall and conducted sales experiments. In this field study, we conducted an experiment to compare sales between a SRA and a robot agent.As a result, the SRA was able to make recommendations to many more users than the robot agent, and the users who received recommendations from the SRA remembered more about the recommended product than those who received recommendations from the robot agent.Based on these results, we confirm the possibility that the SRA is effective for advertising.

In this study, we developed POP Cart that uses a shopping cart and an agent to recommend products. The agent on POP Cart is designed to call the customer by name, chatting and making recommendations to the customer casually like a friend. POP Cart has two advantages: first, the agent can form a positive relationship with the customer is in the store, which could be beneficial for a high recommendation success rate, as found in previous studies. Second, the agent can make multiple recommendations for different products according to the position of the customer while they are shopping in a large store. To evaluate the effectiveness of the recommendations made by the POP Cart agent, we conducted a field experiment in a real supermarket in Japan, where 49 participants shopped under three cart conditions. The results revealed that having an agent on a shopping cart is an effective way to recommend and sell products.

In this paper, we discuss a classification method of nursing-care texts using the word2vec. The word2vec is a tool which provides the continuous bag-of-words and skip-gram implementations for realizing word vectors. We have tackled to classify nursing-care texts, which are freestyle Japanese texts, for improving nursing quality in several years. Several machine learning methods have been used for classifying such texts. To train a machine learning method, we used a word list which contains words appeared in the training data. Since the word list is a mere list, the relation among words is not considered. Also the length of the list depends on the number of words. Word vector representation realized word representations in arbitrary dimensional space. We use the word2vec as a alternative word list in this paper. And we propose a new feature vector definition which is based on dependency structures in a text. From experimental results, we compare the proposed definition with our previous works.

In our past works, a standard three-layer feed forward neural network based human activity estimation method has been proposed. The proposed method aims to record the subject activity automatically. The recorded data by MEMS based monitoring devices include raw accelerometer data of his/her activity. From these data, we need to determine what the subject person was doing. In our conventional methods, some numerical datasets of accelerometer which are measured for every subject person were needed to train neural networks. In this paper, we propose an estimation method of subject behavior using fuzzy neural networks. The proposed fuzzy neural network based method can be trained by using fuzzy if-then rules which represent action primitives instead of numerical datasets from subject person.