On semantics-based minimal revision for legal reasoning

When literal interpretation of statutes leads to counterintuitive consequences, judges, especially in high courts, may identify counterintuitive consequences and revise interpretation of statutes. Researchers have studied revisions for computational legal representation. Generally, studies on revision usually consider minimal revision to reflect limitation of judges' legislative power. However, those studies tend to minimize the number of operations used for changing rules rather than minimize the changes of semantics (the set of conclusions obtained from the program), which vary among cases. In this paper, we consider minimizing the changes of semantics of a rule-base written in a normal logic program. We consider that each possible fact-base (the representation of a case) has its corresponding semantics and corresponding dominant rule-base, which is a set of Horn clauses obtained from the subset of rule-base that is specific to the considered fact-base. Hence, we present a new sub type of semantics-based minimal revision called a dominant-based minimal revision. Furthermore, we present one guidance to obtain one dominant-based minimal revision by using legal debugging and Closed World Specification. We also compare the dominant-based minimal revision with the syntax-based minimal revision in Theory Distance Metric. As the syntax-based minimal revision minimizes the number of operations used for changing rules, the comparison shows that the syntax-based minimal revision may cause extra semantics changes compared to the dominant-based minimal revision, especially when the rule-base contains multiple rules for the same consequence. We discuss that such extra semantics changes can be considered as unintentional changes caused by the syntax-based minimal revision. Hence, legal reasoning systems can check with the user such extra semantics changes to confirm the user intention of changes.