Educational Human-computer Debate: a Computational Dialectics Approach Tangming Yuan and David Moore and Alec Grierson Abstract. Theories of learning suggest that dialogue is important in shaping conceptual development. However, there is widespread de- bate as to the forms of dialogue and which are effective in an educa- tional context. In addressing these issues, we have analysed current knowledge concerning dialectics in philosophy and education. We propose to adopt a computational dialectical approach to study the issues related to the development of an intelligent debating system, which is argued to have potential educational benefit. This approach focuses on using models of dialogue developed in the area of in- formal logic, which prescribe rules to regulate the evolving dialogue. Our proposed research concerns three main issues in the area of com- putational dialectics: dialogue model, debating heuristic theory and dialectical relevance. 1 Introduction The recent development of Computer Based Learning Systems and the emergence of the World Wide Web and the Internet have changed the study life of many people. However, the usual assumption un- derlying these computer based educational systems is that the com- puter does all the informing, the student being merely a passive re- ceiver of the information. The type of teaching interaction, that is, may become unduly didactic [13]. There is therefore a need for dia- logue within interactive computer systems. Further, theories of learn- ing have long suggested that dialogue has an important role to play in shaping conceptual change and developing reasoning skills [18]. There are many different uses of dialogue in an educational con- text. For example, Grasso et al.’s [5] ”Daphne”, a computational agent conducts an advice giving dialogue with the user to provide healthy nutrition education. Maudet and Moore’s [10] human com- puter debate prototype will enable a student and computer to con- duct a fair and reasonable debate on a controversial issue. Raven- scroft and Matheson [17] introduce two kinds of asymmetric dia- logues to support learning. One is the computer being a ”facilitating tutor” and the student the ”explainer”: the tutor raises some ques- tions, students answer the questions, and the tutor solves the con- tradictions of the student’s commitments and helps the students to reach the correct answer rather than directly tell them. Ravenscroft and Matheson’s second dialogue type is similar to the first, but in- cludes further didactic features. Bench-Capon et al. [3] investigated the computer mediated dialogue in legal educational context, which is explanation based, both participants adopting symmetric roles [2]. Pilkington’s study of simulation-based learning identified two types of dialogue, an inquiry dialogue with asymmetric roles and a more collaborative game generating cognitive conflict and reflection ([15], School of computing, Leeds Metropolitan University, Leeds, LS6 3QS, United Kingdom [16]). However, there is widespread debate as to the forms of dia- logue in general and which are effective in educational contexts in particular. We therefore review two approaches to characterising dia- logue types, that of Walton and Krabbe [21] and Baker [1], and then, we make a proposal for human computer debate using a dialectical approach. 2 Dialogue Typology 2.1 Walton and Krabbe’s typology Type of dia- logue Initial situa- tion Participant’s goal Goal of dia- logue Persuasion Conflict of opinion Persuade other party Resolve or clarify issue Inquiry Needs to have proof Find and ver- ify evidence Prove (dis- prove) Negotiation Conflict of in- terest Get what you most want Reasonable settlement that both can live with Information- seeking Need infor- mation Acquire or give informa- tion Exchange in- formation Deliberation Dilemma or practical choice Co-ordinate goals and actions Decide best available course of action Eristic Personnel conflict Verbally hit out at opponent Reveal deeper basis of con- flict Figure 1. Walton and Krabbe’s dialogue typology The most influential dialogue typology is probably Walton and Krabbe’s [21] dialogue model developed in the area of argumenta- tion theory. This model provides a broad typology of dialogue types and their rationale. It is based on three factors: ”(i) the initial situa- tion, (ii) the private aims of the participating agent, (iii) the joint aims to which all participants implicitly subscribe”. Six dialogue types are included in this model: persuasion, negotiation, inquiry, deliberation, information seeking and eristic. See figure 1 (citing from [21]). Reed examined the above dialogue model in some depth in agent commu- nication research [19]. He suggests that ’eristic’ dialogue is unlikely to play a significant role in current computer science research. He also suggests that persuasion, inquiry and information-seeking dia- logues handle belief, while negotiation dialogue raises a contract and c 2002 T. Yuan, D. Moore & A. Grierson Workshop on Computational Models of Natural Argument Edited by Giuseppe Carenini, Floriana Grasso and Chris Reed ECAI 2002. 15th European Conference on Artificial Intelligence