In this paper we describe a system, called BankXX, which generates arguments by performing a heuristic best-first search of a highly interconnected network of legal knowledge. The legal knowledge includes cases represented from a variety of points of view—cases as collections of facts, cases as dimensionally-analyzed fact situations, cases as bundles of citations, and cases as prototypical factual scripts—as well as legal theories represented in terms of domain factors. BankXX performs its search for useful information using one of three evaluation functions encoded at different levels of abstraction: the domain level, an “argument-piece” level, and the overall argument level. Evaluation at the domain level uses easily accessible information about the nodes, such as their type; evaluation at the argument-piece level uses information about generally useful components of case-based argument, such as best cases and supporting legal theories; evaluation at the overall-argument level uses factors, called argument dimensions, which address the overall substance and quality of an argument, such as the centrality of its supporting cases or the success record of its best theory. BankXX is instantiated in the area of personal bankruptcy governed by Chapter 13 of the U.S. Bankruptcy Code, which permits a debtor to be discharged from debts through completion of a court-approved payment plan. In particular, our system addresses the requirement that such Chapter 13 plans be “proposed in good faith.”

We discuss a theoqy of heurtitic strategies and tactics for making arguments in a domain guided by rules where the primary task is to use previous cases to argue for a particular interpretation of a rule in a new fact situation.Since rules frequently have various problematic aspects, such as unstated exceptions or prerequisites, or use terms that are not clearly defined, the actual interpretation is a matter of debate, and the arguer must use interpretations from precedent cases to form his argument.The argument strategies and tactics actually adopted depend on the arguer's point of view and the complexwn of his case in light of the rules and the precedents.The tactics, called "moves" here, are ultimately expressed in a small set of generic "argument primitives, " such as analogizing and distinguishing.We discuss how these argument strategies, moves, and primitives are used by our mixed paradigm system, CABARET (Roland & Skalak, 1990).We illustrate with examples from an area of U.S. Federal income tax law.

In this paper we discuss a hybrid approach to the problem of statutory interpretation that involves combining our past approach to case-based reasoning (“CBA”), as exemplified in our previous HYPO and TAX-HYPO systems, with traditional rule-based reasoning (“RBR”), as exemplified by expert systems. We do not tackle the fullblown version of statutory interpretation, which would include reasoning with legislative intent or other normative aspects (the “ought”), but confine ourselves to reasoning with explicit cases and rules. We discuss strategies that can be used to guide interpretation, particularly the interleaving of CBR and RBR, and how they are used in an agenda-based architecture, called CABARET, which we are currently developing in a general way and experimenting with in the particular area of Section §280A(c)(1) of the U.S. Internal Revenue Code, which deals with the so called “home office deduction”.

Legal reasoning is often couched in terms of legal classification. We examine how three models of classification — Classical, Probabilistic and Exemplar — are used to perform legal classification. We argue that all three models of classification are implicitly applied by existing AI methods. The CABARET (“CAse-BAsed REasoning Tool”) system is suggested as an architecture that applies all three models. The relative difficulty of revising knowledge in rule form, in HYPO-style dimension form, and exemplar form is considered.