Recently, various types of robots co-exsisting with people have been developed. Such a robot should learn human-friendly behaviors through the experiences with people, because the behaviors of the robot suitable to the people cannot be designed. However, there exists several problems on behavior learning, e.g., the problem on the applicability of the acquired behavior owing to the gap between simulation environment and real environment. Therefore, in this paper, we developed a virtual simulation environment connecting seamlessly with a real environment, and proposed a learning method of interaction of the robot with environment and the people in the virtual simulation environment. Furthermore, we conducted several experiments on the interaction with people and the motion control, and discussed the effectiveness of the proposed method.

View full abstract

Proposed is an algorithm that learns the meaning of action instructions and evaluation instructions through interaction with a navigator. The navigator speaks freely, which requires learning from various kinds of few data with biased frequency and error. Utilizing Fisher's exact test, the algorithm finds the pairs of a phrase and a situation which frequently cooccurs. An experiment shows that the learning accuracy improves compared to an earlier study.

View full abstract

We propose a software platform for applications based on socially intelligent agents (SIAs). SIAs are software agents which manifest social intelligence. Although there are many types of social intelligence, we focus on two types, social intelligence for providing emotional support and social intelligence for providing cognitive support. By applying these types of social intelligence, our SIAs are able to simulate human social behavior. We demonstrate how we created the SIAs and describe the conceptual architecture of our software platform. We also show the mechanism for generating the agents' social behavior, which makes our software platform unique. According to the media equation theory, we expect SIAs to benefit users. To examine the effects of SIAs, we designed and implemented two applications. We provide explanation about how SIAs function in these applications and discuss further implications of social intelligence.

View full abstract

Recently, users publish their comments, diaries and photos on the web services such as weblogs and photo sharing services. In these cases, it is common to manage those contents by tagging some keywords for classifying their characteristics. In this paper, we defined those manual annotations by users that include subjective information such as feelings and intentions, as “subjective annotation”. Currently, linguistically expressed subjective annotation, typified by collaborative tagging has come into widespread use. However, it is difficult to express some feelings by linguistically expressed annotation. Therefore, we proposed the use of avatars as a means of nonverbal expression of subjectivity, and confirmed the consistency of feelings elicited by avatars in two cases; over time for an individual and in a group. In addition, we compared the expressiveness and ease of subjective annotation between collaborative tagging and avatars. The result indicated that the feelings evoked by avatars are consistent in both cases, and using controllable avatars is easier than collaborative tagging for representing feelings elicited by contents that do not express meaning, such as photos.

View full abstract

An experimental investigation of how the appearance of the agents affects interpretations people make of the agents' attitudes is described. At first, we conducted a preliminary experiment to select eight artificial sounds among 44 ones that can make people estimate specific primitive attitudes. We then conducted a psychological experiment where participants were presented the artificial sounds selected in the preliminary experiment from three kinds of agents, e.g., a MindStorms robot, AIBO robot, and a normal laptop PC. Specifically, they were asked to select the correct attitudes based on the sounds expressed by these three agents. The results showed that the participants had higher interpretation rates when a PC presented the sounds, while they had lower rates when MindStorms and AIBO robots presented the sounds, even though the artificial sounds expressed by these agents were completely the same. In case of make people interpret the primitive attitudes like positive/negative, it requires that 1) expressing the imaginable information based on the agents' appearance and that 2) assigning the attributes whether the expressed information deviated from the natural energy flows.

View full abstract

In this study, we focused on body location and orientation of an embodied agent to develop equal social relationship between a user and an embodied agent. We conducted a psychological experiment focusing on body orientation of the embodied agent, and the agent's location based on depth-dimensional virtual space partitioning. Moreover, we also considered gender difference in this experiment since preference of body arrangement in human dyadic conversation should depend on gender. Experiment 1 (N=48) was conducted to examine how body arrangement of an embodied agent influenced task performance of picture recognition task of participants and evaluation of impression of the agent by the participants considering gender difference. Additionally, experiment 2 (N=51) was conducted to investigate the influence of body orientation of the agent on the recognition of user-agent distance. These experimental results suggested that male users should prefer more distantly recognized user distance, being compared with female users. This implication indicates further investigation of design of embodied agents considering influence of the body arrangement on a user due to gender difference.

View full abstract

Our research aims at constructing a dialogue-based instructional agent model. To achieve the goal, we propose an `open communication' model as a novel framework for instructional contents, and analyze the interrelations among verbal and non-verbal modalities (speech/gaze/posture) in manzai dialogues. In particular, focusing on inward or outward-oriented function of open communication, we analyzed two video contents of manzai dialogues and explored how multimodal behaviors and interactions of the two pairs of laughmakers realized either of the communicative functions. As a result, we found that the ways of realizing the communicative orientedness differed between the pairs of laughmakers in manzai dialogue based on the interrelations of their speech and multimodal behaviors.

View full abstract

This paper reports empirical results showing the difference between human-human and human-artifact interactions, and proposes human interaction strategies that are useful in designing human-agents based on human behavioral models. Our empirical study uses a PC operation task as a corpus collection experiment, and reveals human-artifact interaction strategies by analyzing how verbal/nonverbal behaviors are allocated in human-human and human-artifact interactions. First, as basic characteristics in human-human and human-artifact conversations, we found that in human-artifact interactions, the number of utterances and frequency of acknowledgements and other responses are smaller than those in human-human conversations. Then, we propose human-human verbal/nonverbal behavior allocation rules, and examine how these rules are violated in human-artifact interactions, suggesting that these violations are complementary behaviors by the listener that displays understanding of the utterance to the speaker without using a verbal response.

View full abstract

The paper describes two related studies that investigated skill transfer between the same workgroups of a bank over a period of 3 months. During this period the smaller group trained the larger group in customer relationship management skills. At the beginning and the end of this period, group members had to describe their interpersonal and intergroup relationships in diagrams using predefined diagrammatic items. The analysis of the diagrams showed that interpersonal characteristics such as friendship, respect, good manners, and interpersonal contact correlate higher with skill transfer (significance level p< 0.01) than intergroup characteristics, which had no significant correlation with skill transfer. A comparison of the two sets of diagrams after two weeks and three months also showed a polarization effect, in that after three months most relationships became more harmonious or more conflict-laden. The latter relationships led to a reduction in skill transfer. This result modifies the contact hypothesis, which emphasized the benefits of close intergroup contact. Consequences for intergroup skill transfer are discussed.

View full abstract

Much research has been conducted on community support using Web assistance systems that do not constrain users temporally and spatially, and onsite systems which aid real world activities such as event spaces or offline meetings. To use the Web assistance system by a user device onsite, it is necessary to link the user device ID to the user ID of the Web system. This paper describes such a device linkage method that aims to achieve linkage based on social relations among users. A device linked user can recommend and invite non-linked friends to the onsite system. This method requires no previous actions by prospective onsite system users, and reduces the burden of onsite managers. In this paper, user scenarios are analyzed and the device linkage system is implemented, being carefully designed to maintain reliability based on social relationships. A field-test of a conference demonstrates its effectiveness.

View full abstract

This paper proposes a system to introduce users to others for social matching. Introduction is one method of social matching based on social context. It requires consideration of social relationships among three persons: a person who wants to be introduced, a person who is the target of an introduction, and an introducer. The social matching process does not simply entail a person finding target persons, but must also include that they find a way to introduce themselves. In this paper, we discuss possibilities for introduction assistance for social matching, having performed experiments using a prototype system. The results show the usefulness of introduction assistance: the function of generating drafts of referral works well for reducing an introducer's burden; it is valuable to show an introducer reasons why a referral client wants to meet with a target person; and social relationships among three persons have an effect on the type of introduction assistance.

View full abstract

In this paper, we explain about SOBA Framework, a software platform for creating remote communication tools, and a knowledge management system. Then, we explain the mechanism of knowledge transfer, and validate the evidence to obtain how much effect is achieved. In the mother factory system, to provide technical assistance for each subsidiary factory in foreign country, a factory in home country will be assigned as a mother factory. Then, the mother factory will be in charge for providing technical assistance and dispatch engineers for its subsidiary factory. This method is widely employed by companies in automotive industry. However, if there is a trouble and the engineers in charge at the subsidiary factory can not solve it, it will be necessary to ask the experts in the mother factory for help by visiting the subsidiary factory. This needs an enormous cost. Moreover, the trouble is left until the expert arrives at the subsidiary factory. In addition, it leads to the decrease in productivity by restraining the experts for a long time. Then, we developed an online knowledge transfer system. This system is based on SOBA Framework, a software platform for creating remote communication tools, and a knowledge management system.

View full abstract

A trick art drawing system based on a torus figure generating algorithm is constructed. The proposed algorithm is based on the shapes and positions of corners of torus figures. The proposed system is able to generate an arbitrary n-bar torus type possible or impossible figure using interactive mouse-clicking operations on a figure generated from n+1 initial inputs only. It is possible to draw a trick art figure by painting each face of the obtained graph. The proposed system can serve as basic tool for experiments on visual psychology and an educational support tool for discrete mathematics classes.

View full abstract

To optimize large-scale distribution networks, solving about 1000 (around 40 cities) TSPs (Traveling Salesman Problems) within an interactive length of time (max. scores of seconds) is required. Yet, expert-level (less than 3% of errors) accuracy is necessary. To realize the above requirements, a Multi-step intelligent GA method was developed. This method combines a high-speed GA with an intelligent GA holding problem-oriented knowledge that is effective for some special location patterns. If conventional methods were applied, solutions for more than 20 out of 20,000 cases were below expert-level accuracy. However, the developed method could solve all of 20,000 cases at expert-level.

View full abstract