Socially Intelligent Agents

This introduction explains the motivation to edit this book and provides an overview of the chapters included in this book. Main themes and common threads that can be found across different chapters are identified that might help the reader in navigating the book. 1. Background: Why this book? The field of Socially Intelligent Agents (SIA) is by many perceived as a growing and increasingly important research area that comprises very active research activities and strongly interdisciplinary approaches. The field of Socially Intelligent Agents is characterized by agent systems that show humanstyle social intelligence [5]. Humans live in individualized societies where group members know each other, so do other animal species, cf. figure 1.1. Although overlap exists, SIA systems are different from multi-agent systems that a) are often only loosely related to human social intelligence, or use very different models from the animal world, e.g. self-organization in social insect societies, or b) might strongly focus on the engineering and optimization aspects of the agent approach to software engineering. In the past, two AAAI Fall Symposia were organized on the topic of Socially Intelligent Agents, in 1997 and 2000. Both symposia attracted a large number of participants. The first symposium gave a general overview on the spectrum of research in the field, and in the years following this event a variety of publications (special journal issues and books) resulted from it1. Also, a number of related symposia and workshops were subsequently organized2. Unlike the 1997 symposium, the 2000 symposium specifically addressed the issue of Socially Intelligent Agents The Human in the Loop. A special issue 2 Socially Intelligent Agents Figure 1.1. Elephants are socially intelligent biological agents that live in family groups with strong, long-lasting social bonds. Much research into socially intelligent artifacts is inspired by animal (including human) social intelligence. of IEEE Systems, Man and Cybernetics, Part A emerged from this symposium which provides an in depth treatment of a few research approaches in that area3. Unlike the special journal issue, this book has a radically different nature: it is intended to be the first definitive collection of current work in the rapidly growing field of Socially Intelligent Agents, providing a useful and timely reference for computer scientists, web programmers and designers, computer users, and researchers interested in the issue of how humans relate to computers and robots, and how these agents in return can relate to them. Each of the 32 chapters is, compared to a journal article, relatively short and compact, focusing on the main theoretical and practical issues involved in the work. Each chapter gives references to other publications that can provide the reader with further detailed information. In the area of software and intelligent agents many other publications are available, e.g. [1], [9], [6], proceedings of the Autonomous Agents and other conferences, just to name a few. However, none of them provide a state-ofthe-art reference book on Socially Intelligent Agents with an interdisciplinary approach including both software and robotic agents. Despite many publications that either a) specialize in particular issues relevant to Socially Intelligent Agents (e.g. robots, emotions, conversational skills, narrative, social learning and imitation etc., cf. [12], [10], [3], [7], [2], [11], [4]), or b) present a small number of in-depth discussions of particular research projects (published in journal issues mentioned above), the field of Socially Intelligent Agents is missing a state-of-the-art collection that can provide an overview and reference book. More and more researchers and PhD students Creating Relationships with Computers and Robots 3 are interested in learning about and participating in SIA research, but at present the only way to learn about the field is to go through and select among a large number of widely ‘distributed’ and often difficult to access publications, i.e. journal issues, books, conference and workshop proceedings etc. Our motivation to edit this book was therefore based on the belief that there is a strong demand for a book that can be used by students, researchers and anybody interested in learning about Socially Intelligent Agents. The main strength of the book is the breadth of research topics presented and the references given at the end of each chapter, so that researchers who want to work in that field are given pointers to literature and other important work not included in the book. The book presents a coherent and structured presentation of state-of-the-art in the field. It does not require the reader to possess any specialist knowledge and is suitable for any student / researcher with a general background in Computer Science and/or Artificial Intelligence or related fields (e.g. Cognitive Science, Cybernetics, Adaptive Behavior, Artificial Life etc.). Also, at present the growing field of Socially Intelligent Agents has no core text that can be used in university courses. This book fills this gap and might be used in different courses for postgraduate studies, and as research material for PhD students, e.g. for studies in Applied Artificial Intelligence, Intelligent and Autonomous Agents, Adaptive Systems, Human-Computer Interaction, or Situated, Embodied AI. 2. Book Structure and Chapter Overviews The remaining thirty-two chapters of this book are organized into two parts. The structure of the book is visually shown in figure 1.2. The first part addresses the theory, concepts and technology of Socially Intelligent Agents. The second part addresses current and potential applications of Socially Intelligent Agents. The first part of the book has twelve chapters organized in three sections covering three major themes, namely relationships between agents and humans, edited by Alan Bond, agents and emotions/personality edited by Lola Cañamero, and communities of social agents, edited by Bruce Edmonds. The second part of the book consists of twenty chapters organized in five sections covering the themes of interactive therapeutic agent systems, edited by Kerstin Dautenhahn, socially intelligent robots, edited by Lola Cañamero, interactive education and training, edited by Kerstin Dautenhahn, social agents in games and entertainment, edited by Alan Bond, and social agents in e-commerce, edited by Bruce Edmonds. The content of the sections and chapters is described in more detail below. Note, that thematically we have strong overlaps between all chapters in this book, the division into thematic sections is mainly of practical nature. This 4 Socially Intelligent Agents Figure 1.2. Book structure, showing the division into two parts and eight sections.