AAAI 1991 SpringSymposium Series Reports

Traditionally, research approaches to building models of argument knowledge and belief have differed in terms of objectives, scope, and methodology. For example, in rhetoric, the groundswarrant-claim model has been used to analyze the structure of arguments. In psycholinguistics, researchers have analyzed the discourse structure of expository text by applying theories of discourse and schema coherency. In the field of logic, the emphasis has been on establishing axiomatic systems for deducing consistent beliefs. In AI, researchers have developed systems involving truth maintenance and evidential reasoning as well as models of legal reasoning, structural models of argument and discourse, and knowledge-based models of argument comprehension and generation. A forum for multidisciplinary discussion of these approaches was provided last spring by the Symposium on Argumentation and Belief. The objective of the symposium was to further understanding of the issues, current status, and future directions of research in argumentation and beliefs. To facilitate the exchange of ideas, the presentation of each paper was followed by a commentator’s response and a question-answering session. Papers were organized into five sessions that yielded insight into the following areas: representation of belief and argument knowledge, processes of argument comprehension and argument generation, the role of domain knowledge in argumentation, the role of memory in argumentation, methods for assessing relative strengths and weaknesses of arguments, process of persuasion and belief change, the role of planning in argumentation, the role of case-based reasoning in argumentation, causal reasoning, analogical reasoning, and the teaching of argumentation skills. The symposium also included five panel discussions that dealt with issues involving the following areas: logical and philosophical models of argumentation and belief, methodology of research in argumentation and belief, representation of belief and argument knowledge, belief formation and learning through argumentation, and comprehension and generation of natural language in arguments. The research described at the symposium demonstrates that designing systems capable of arguing requires characterizing interrelationships that exist between planning, reasoning, language comprehension, and language generation. For example, people reason about whether a goal should be achieved and use reasoning to justify beliefs about the efficacy, side-effects, and cost of selecting and executing plans. People also possess planning information on how to select and apply argument strategies. At the same time, human experts are able to not only present and justify their beliefs on possible courses of action but also understand opposing beliefs and argue persuasively for or against various positions. Given the complexity of these interrelationships, it is all the more important to continue developing and experimenting with whole, functioning prototypes, so that both the strengths and weaknesses of the proposed theories of argumentation and belief can be tested and revealed. The major benefit derived from the symposium was the multidisciplinary understanding of fundamental issues in knowledge representation, knowledge organization, and knowledge application that must be addressed within any intelligent computer system capable of understanding and engaging in arguments.