For the design of AML we started by looking for common semantic concepts among argumentation tools and arguments, used to support different technical approaches and different fields of application. We captured these core semantic concepts in a common argument representation. This representation distinguished between uninstantiated argument models (templates in our terminology) and instantiated arguments (arguments in our terminology). In addition, templates combined both a question hierarchy (or network in Bayesian net terminology), as well as an aggregation rule attached to each question (node) in the question hierarchy (network).
Rather than use technical terms for the elements of AML (e.g., variable, condset, node, arc), we decided to use legal terminology that is more readily understood (e.g., argument, evidence, exhibit, rationale, relevance). Thus, while AML can represent Bayesian networks, it does so using very different terminology than other popular schemas for Bayesian networks (e.g., MSBNx). AML can represent the same things (and more), but is more easily understood by those not versed in Bayesian networks or probability theory. In addition AML is a relatively open XML schema that can be extended for use by other argumentation tools by incorporating tool-specific information.
|Copyright © 2000-2006 SRI International, 333 Ravenswood Ave., Menlo Park, CA 94025 USA. All rights reserved.|