Grand Challenge Problems on Cross Cultural. Communication. {Toward Socially Intelligent Agents{ Takashi Kido 1

Transcription

1 Grand Challenge Problems on Cross Cultural Communication {Toward Socially Intelligent Agents{ Takashi Kido 1 NTT MSC SDN BHD, 18th Floor, UBN Tower, No. 10, Jalan P. Ramlee, Kuala Lumpur, Malaysia Abstract. The stage is set for studying cultural inuences on humanhuman, human-agent and agent-agent communications. So far much attention has been paid to create adaptive agents which have adaptive user interface or exible capability to communicate with another software agents. However little eort has been made for examining cultural inuence in communications. I believe that as in the human-human communication, the cultural issues are unavoidable even in the communications conducted by software agents. \Socially intelligent agent", which can adapt itself to ll such cultural gaps between the parties involved in the communication, needs to identify the dierence in forms of social interactions. This paper raises six grand challenge problems on cross cultural communication that CIA community should jointly tackle in creating such socially intelligent agents. 1 Introduction In order to communicate with the foreigner eectively, it is important to understand their cultural background. In this paper, the word \culture" includes every social aspects which aect individuals' decision making and communication styles. The word \individual" here means not only human but also includes robot or software agent. The purposes of this paper are (1) to show the importance of cross cultural communication for AI and Agent studies and (2) to provide grand challenge problems that CIA community should jointly nd answers. 2 Why Cross Cultural Communication now? There are several reasons why studies on cross cultural communication are important in CIA community. (1) The opportunity to meet with dierent cultures has been increasing due to the improvement of information technology and transportation. Therefore, there are great demands for mutual understanding of human-human cross cultural communication. (2) Applying AI and Agent technologies would potentially benet international businesses or international group activities. (3) The study of user interface or software agent has been confronting

2 cultural issues. In order to improve human-machine interaction, it is essential to understand the cultural inuences. In fact, newly introduced human-machine interface can be rated dierently based on the user's cultural background. (4) Future robots and software agents will need to interact with each other and humans, in respect of established social protocols or cultural dierences. We believe that the \socially intelligent agent" study meets the cultural problems. The term \socially intelligent agent" was introduced in AAAI97 Fall Symposium series held in MIT. [5] [2] This symposium was focused attention on studies of human-like social behavior and expertise and also on approaches to design and evaluate articial systems to interact socially with humans in an acceptable manner. Although current robots or software agents are still far from humans or natural social creatures, new forms of social interaction need to be developed in the future. We are now starting new projects in Malaysia under the initiative of MSC (Multimedia Super Corridor) project which is lead by Malaysian government to develop multimedia utopia in the dawn of 21st century. This national project tries to create an environment where multi-national, multi-cultural entities come and produce various kind of software and multimedia products. The projects provide us a good environment for research on cross cultural communication. 3 GrandChallenge Problems We propose the following 6 projects as challenge problems which AI and Agent communities should tackle. 3.1 What is the cross cultural communication?, Case-study Culture underlines every aspect of social behavior and inuence individual communication style, personality, character, motivation, knowledge, and cognitive mechanisms. [3] Actually, there are cross-cultural dierences, in both verbal and non-verbal forms, in expression of emotion. For example, in the American culture, looking directly into the eyes of the person to whom you are talking with is seen as a positive trait; a sign of condence. The Japanese culture, on the other hand, sometimes considers this negative, improper. In order to (1) mutually understand dierent values, communication styles all over the world and (2) create new technology to support cross cultural activities, we propose to organize a place where we can share several case studies for cross cultural communication. 3.2 Linguistic studies for cross cultural communication Cross-cultural dierences in communication styles have been explored extensively in the eld of linguistics and cultural anthropology. The nding that each language aects human's mind and social behavior in a distinctive manner is worth further study. Such kind of work should be the basis for designing intelligent software agents.

3 3.3 User Interface and new media for cross cultural communication It is important for software developers to understand cultural dierences in human-human communication and to apply them to user interface design. Because, in many cases, the rating for new design and interface greatly depends on the cultural background of the users. [4] For instance, there is a report that U.S. users do not tend to favor anthropomorphic agents while Japanese users prefer them. [3] This report says that Microsoft's \Bob" did not stimulate to much interest in the U.S. marketplace, while Japanese users seem to openly embrace human-like agents. Although further surveys and analysis are necessary to conclude this issue, it is true that cultural dierences have great inuence on human-computer interaction. It is a great issue to be discussed on how new media and design can aect the human mind and the culture. 3.4 Negotiation strategy among dierent cultural agents History shows that understanding cultural dierences is very important for successful negotiations in the eld of international business and diplomatic aairs. In order to obtain best results through negotiation, the negotiators should understand the background and the cultures of their counterpart. In cyber age, a software agent may resolve some conicts or maximize their own utility by negotiating with other agents. Some eective negotiation strategies adopted by rational agents are proposed based on the game theory. [7] If the agent has its own personality and behaves like a human, non-rational and culture dependent negotiation may be required to success in the negotiation. This story is realistic when many service agents and information agents emerge on the internet world. 3.5 Internet technology and its impact for cross cultural communication The rapidly expanding global infrastructure of the Internet has given great impacts on the world wide communications and it also provides several opportunities for application, to validate and to shape the new technology for cross cultural communication. [9][1] For example, virtual meeting place, where people from all over the world meet on Internet, provides cross-cultural environment for various information exchanges. In designing a way to locate people with similar interests and also to introduce them to each other, intelligent systems should be developed. [8] Another possibility is adaptive web sites: sites that record user access logs and automatically improve their organization and presentation based on user access data. [11] Culturally adaptive web sites requires the function in detecting dierent forms of communicative behaviors and responding to them appropriately by using user access data. Machine learning technology will play a prominent role for this purposes.

4 3.6 The origin and evolution of communication, social norms and culture To understand the origin of language and meaning is a long term vision of communication science. [6] Each living creature has its own communication style which is dierent from that of other species. This communication style is formed during the process of natural evolution. Axelrod shows how cooperative behavior emerges from the competitive and selsh agent's world. [13] One of the assumption is that a culture is a by-product of natural evolution. If this assumption is true, we can simulate future possible cultures under various conditions. The interesting question is \how do team work or culture depended behavior evolve in social dynamics?" or \how do social relationships develop?" For this purpose, one of the possibilities is to provide a common simulation environments for evaluation of various theories, algorithms, etc. (RoboCup is one of the good examples for this common task [12] ) Further discussions are expected in this workshop. 4 Conclusion In this paper, I emphasized the importance of cross cultural communication studies for AI and Agent communities and provided grand challenge problems that CIA community should jointly tackle. These challenge problems on cross cultural communication include (1) The case study of several cross cultural communications (2) Linguistic approaches (3) The eect on Human-Computer Interaction (4) Negotiation strategy (5) Internet Application (6) Evolution of communication, social norms and culture. Further discussion is expected in the following workshop and also on other activities. References 1. Takashi Kido: A Vision of AL-Spider: -Toward Adaptive Information Gathering-. IJCAI97 Workshop on The Future of AI and the Internet. (1997) 2. Takashi Kido: Is Social Intelligence necessary for Agent-Agent Interaction? -Toward Socially Intelligent Software Agents AAAI Fall Symposium on Socially Intelligent Agents. (1997) 73{74 3. Patricia O'Neill-Brown: Setting the Stage for the Culturally Adaptive Agent AAAI Fall Symposium on Socially Intelligent Agents. (1997) 93{97 4. Chisato Numaoka: Innate Sociability: Sympathetic Coupling AAAI Fall Symposium on Socially Intelligent Agents. (1997) 98{ Kerstin Dautenhahn: Ants Don't Have Friends {Thoughts on Socially Intelligent Agents AAAI Fall Symposium on Socially Intelligent Agents. (1997) 22{27 6. Luc Steels: Perceptually grounded meaning creation. Second International Conference on Multi-Agent Systems (ICMAS96). (1996) 338{ Matthias Klusch: A Polynomial Kernel-Oriented Coalition Algorithm for Rational Information Agents. Second International Conference on Multi-Agent Systems (ICMAS96). (1996) 157{164

5 8. Leonard N. Foner: Yenta: A Multi-Agent, Referral-Based Matchmaking System. Second International Conference on Multi-Agent Systems (ICMAS96). (1996) Alexander Franz and Hiroaki Kitano: Internet-Based Information Systems: Papers from the 1996 AAAI Workshop. AAAI Press. (1996) 10. Todd Siler: Think Like A Genius. Bantam Books. (1997) 11. Mike Perkowitz and Oren Etzioni: Adaptive Web Sites: an AI Challenge. Proceedings of Fifteenth International Joint Conference on Articial Intelligence (IJ- CAI97). (1997) 16{ Hiroaki Kitano, et al.: The RoboCup Synthetic Agent Challenge 97 Proceedings of Fifteenth International Joint Conference on Articial Intelligence (IJCAI97). (1997) 24{ R. Axelrod: The Evolution of Cooperation. Basic Books. (1984)