On the Analysis of Regulations using Defeasible Rules

Regulations are a wide-spread and important part of government and business. They codify how products must be made and processes should be performed. Such regulations can be difficult to understand and apply. In an environment of growing complexity of, and change in, regulation, automated support for reasoning with regulations is becoming increasingly necessary. In this paper we claim that such automated support can be provided on the basis of defeasible logical rules. We highlight the support that can be provided by this logical tool, and illustrate some aspects using examples from one specific domain: university regulations. Introduction Regulations are a wide-spread and important part of government and business. They codify how products must be made and processes should be performed. Such regulations can be difficult to understand and apply. Even stand-alone regulations can be selfcontradictory, as a result of the incremental process of their development and the lack of a formal drafting process. The problem becomes more difficult when independently developed regulations apply to a situation. For example, when two regulations overlap, it is not clear whether one regulation takes precedence or both regulations apply. Even when regulations are formally drafted, as is often done in the legal domain, problems with the consistency, interpretation and use of regulations still remain (e.g. (Sergot et al. 1986)). an environment of growing complexity of, and change in, regulations, automated support for reasoning with regulations is becoming increasingly necessary. In this paper we discuss issues of using logical methods to analyse regulations. In order to provide adequate support for the analysis of regulations a representation needs to meet some requirements: ̄ Expressive power: The method must be able to represent the basic features of the problem that needs to be modelled and analysed. ̄ Reasoning support: Not only should the features of the problem be representable, but it should also be able to manipulate the representations to derive properties of and conclusions from these representations. This lies at the heart of logical analysis. ̄ Naturalness of expression: It should be possible to represent the problem at hand in a transparent and natural way. Otherwise, even if the method satisfies the previous two requirements, it will fail, because it will be difficult to use the method in concrete situations. We propose to use defeasible rules with priorities as a logical method of analysing regulations. Rules are normally sufficiently expressive to represent single items within a regulation. Since regulations may contradict one another the use of defeasible rules is adequate: they are rules that do not necessarily fire; rather they may be blocked by other rules with contrary conclusions. For more general arguments in favour of defeasible rules see (Schurz 1994). Regulations commonly include exceptions. In addition there may be principles by which one set of regulations is superior to another (for example, a regulation of a higher authority is superior to a regulation of a lower authority; or a more recent regulation might override an older regulation). The priorities on defeasible rules are an adequate and natural way of representing this information. The reasoning support will be provided by a particular logic we will be using: defeasible logic (Nute 1987, Nute 1994, Billington 1993). It is an approach to sceptical nonmonotonic reasoning (Antoniou 1997, Marek & Truszczynski 1993) that has a very distinctive feature: It was designed to be easily implementable right from the beginning, unlike most other approaches. In fact it has an implementation as a straightforward extension of Prolog (Covington 1997). Defeasible logic is a sceptical formalism, meaning that it does not support contradictory conclusions. In46 From: AAAI Technical Report WS-98-13. Compilation copyright © 1998, AAAI (www.aaai.org). All rights reserved. stead it seeks to resolve differences. In cases where it is impossible to decide whether to conclude A or --,A, the logic does not conclude either of them (thus the name "sceptical"). Sceptical reasoning is appropriate for the study of regulations. Users of regulations are mostly interested in getting correct advice without being confronted with conflicting views. Basics of Defensible Logic A set of regulations will be represented as a defensible theory. A defensible theory1 consists of four different kinds of knowledge: facts, strict rules, defensible rules, and a superiority relation. Facts denote simple pieces of information that are deemed to be true regardless of other knowledge items. A typical fact is that Tweety is a bird: bird(tweety). Strict rules are rules in the classical sense: whenever the premises of a rule are given, we are allowed to apply the rule and get a conclusion. When the premises are indisputable (e.g. facts) then so is the conclusion. example of a strict rule is "Emus are birds". Written formally: emu(X) --+ bird(X). Defeasible rules are rules that can be defeated by contrary evidence. An example of such a rule is "Birds typically fly"; written formally: bird(X) :¢, flies(X). The idea is that if we know that something is a bird, then we may conclude that it flies, unless there is other evidence suggesting that it may not fly. The superiority relation among rules is used to define priorities among rules, that is, where one rule may override the conclusion of another rule. For example, given the defensible rules r: bird(X) ~ flies(X) r’: brokenWing(X) ~-,flies(X) which contradict one another, no conclusive decision can be made about whether a bird with a broken wing can fly. But if we introduce a superiority relation > with r ~ > r, then we can indeed conclude that it can’t fly. It turns out that we only need to define the superiority relation over rules with contradictory conclusions. Also notice that a cycle in the superiority relation is counterintuitive from the knowledge representation perspective. In the above example, it makes no sense to have both r > r~ and r ~ > r. Consequently, the defensible logic we discuss requires an acyclic superiority relation. l a knowledge base in Defeasible Logic Examples: University regulations Example 1: Academic misconduct A typical rule found in university regulations is the following, taken from the Griflith University policy on academic misconduct: Where a student has been found guilty of academic misconduct on more than one occasion and has previously been penalised as set out in 3.1 3.3 above, the penalty shall normally be exclusion from the course, unless in the opinion of the relevant Assessment Board there are mitigating circumstances. This is a typical rule with exceptions. In the framework we are proposing, we would represent this rule as follows: rl : guilty, repeat, previouslyPenalised exclude r~ : mitigatingCircumstances ~ -.exclude