Case-based reasoning in the law. A formal theory of reasoning by case comparison

Bram Roth

Summary

Bart Verheij, Spring 1999 - Click for an older picture (Summer 1996, at WashU by Ron Loui) This thesis is about case-based reasoning in the law. Case-based reasoning is presented as a method of adhering to decisions taken in settled cases, on the basis of a comparison with the case at hand. If the case at hand provides at least as much dialectical support for the conclusion as a settled case, then the settled case can be followed and its conclusion can be drawn. The thesis sets forth a formal theory of the method of reasoning by case comparison, and it compares the theory to existing approaches to case-based reasoning.

The work has aimed at answering roughly two groups of research questions. The first group concerns the method of reasoning by case comparison itself:

  • How can case comparison be formalised to determine whether a settled case can be followed? Which conclusions can be drawn by comparison with settled cases? Along which patterns of reasoning do the conclusions follow?
The second group of research questions concerns the proposed model’s relation with existing approaches to case-based reasoning in Artificial Intelligence and Law (AI & Law):
  • How does the method of case comparison relate to other approaches to case-based reasoning? To what extent is case comparison formalised in these approaches? Is there an account of the conclusions that follow?
Both groups of questions have been addressed in the five chapters of the present thesis, the contents of which can be summarised briefly as follows.

In Chapter 1 the research context is sketched within which the present work on case-based reasoning was done. The work is positioned within the field of AI & Law, a field where a variety of legal activities is investigated from an Artificial Intelligence perspective, ranging from legal decision making to drafting bills.

In Chapter 2 some issues concerning case-based reasoning in the law are discussed that are especially relevant for the present thesis, both from a legal-theoretical and an AI & Law perspective. Two methods of adhering to decisions are presented, namely case comparison and rule extraction. In the case comparison method the conclusion of a settled case is copied directly after a comparison with the case at hand. In the rule extraction method the decision is first summarised into a general rule, after which the rule is applied to the case at hand.

The chapter goes on to discuss some relevant issues concerning the case comparison method. One issue is which conclusions can be established on the basis of this reasoning strategy. Another issue is along which patterns of reasoning one arrives at the conclusions. Patterns of reasoning typically found in case-based reasoning are ‘analogising’ and ‘distinguishing’. The former involves pointing out similarities between a decided case and the case at hand, while in the latter the relevant differences between the cases are stressed.

Another relevant issue that is discussed in Chapter 2 is, that the outcomes of case comparison are determined by the set of case features that are deemed relevant for the purpose of comparing cases. In particular, it is important at which level of generality the relevant case features are to be stated.

In Chapter 3 a formal theory is developed for reasoning by case comparison. The main objective of the theory is to specify precisely the conclusions that follow on the basis of the case comparison method. A secondary aim is to account for reasoning patterns involved in case comparison, such as analogising and distinguishing.

In the chapter, formal conditions are stated for when there is an analogy between a problem case and a settled one. Intuitively, these conditions hold that the problem case provides at least as much support for the disputed conclusion as the settled one. Then by a variant of reasoning a fortiori, the conclusion can hold in the problem case as well.

To determine whether the settled case can be followed in the problem case, a comparison is made of the argumentation relevant for the conclusion. This argumentation is made explicit by what is called a dialectical argument. Dialectical arguments include statements of fact upon which conclusions can be founded, but also intermediate conclusions. Moreover, in contrast to other work on case comparison, the model does not take for granted which statements are reasons for or against conclusions. Accordingly, it is possible to support or attack that one statement supports or attacks another, which results in a kind of entanglement of the dialectical arguments. In the proposed model this entanglement is also taken into account in case comparison, which becomes apparent in the use of an entangled factor hierarchy.

An important feature of the present model is, that it is acknowledged explicitly that in the law it depends on a contingent choice which case features are relevant in comparing cases. In accordance with this, the relevant case features are given by what is called a comparison basis, and the comparison outcomes are stated relative to this comparison basis.

In Chapter 4 it is discussed how the proposed model relates to other approaches to case-based reasoning. One issue thereby discussed is whether the conclusions that follow from settled cases are defined, another issue is whether there is an account of the reasoning patterns along which these conclusions follow. A third issue is whether adherence to decisions is treated as rule extraction or as case comparison. As seen, none of the discussed approaches presents a formal account of the case comparison method that includes the both the conclusions that follow, and the reasoning patterns along which they follow. A fourth issue is whether it is acknowledged that in the law it depends on a contingent choice which case features are relevant for case comparison. As seen, none of the discussed approaches explicitly deals with the contingency of the set of relevant case features.

In Chapter 5 the theory is critically evaluated. A strong point of the model is that the conclusions that follow by case comparison are specified. The chapter discusses a number of reasoning patterns that are not yet captured by the proposed model. Among these one has, for instance, reasoning on the basis of the hierarchy among courts. The evaluation also discusses how such patterns of reasoning may be dealt with within the model proposed. In addition a number of subjects for future research are recommended, like modelling reasoning with hypothetical cases.

This concludes a brief summary of the contents of the present thesis. The main contributions of the thesis can be summed up as follows.

  • Case comparison is formalised as a method of adhering to decisions. The outcomes of case comparison are formally defined, and there is an account of the conclusions that follow by case comparison.
  • The dialectical structure of cases is made explicit by tree-like structures called dialectical arguments. Dialectical arguments involve statements supporting or attacking other statements. In addition to this, it is possible to support or attack that one statement supports or attacks another, giving rise to a kind of entanglement of the dialectical arguments. It is an innovation of the present model that this entanglement plays a role in case comparison, which becomes apparent in the use of an entangled factor hierarchy.
  • The model explicitly acknowledges that in the law it depends on a contingent choice which case features are relevant in case comparison. Formally this is done by introducing a comparison basis as a contingent set of factors relevant for case comparison, and by defining the comparison outcomes and the derivable conclusions relative to the comparison basis.
  • Typical reasoning patterns in case comparison are accommodated, namely analogising, distinguishing, downplaying and emphasising. The latter two patterns are accounted for as a change in the comparison basis, whereby it is an innovation is that downplaying and emphasising are not only possible for distinctions, but for similarities as well.
Reference:
Roth, B. (2003). Case-based reasoning in the law. A formal theory of reasoning by case comparison. Dissertation Universiteit Maastricht.

Further information (and links to downloadable version): Maastricht University, NARCIS

Full text (19 MB)

Promotor: Prof. dr. H.F.M. Crombag
Co-promotor: Dr. B. Verheij
Beoordelingscommissie: Prof. mr. N.H.M. Roos
Prof. dr. K.D. Ashley (University of Pittsburgh, PA, USA)
Dr. F.M.P. van Dun
Dr. mr. J.C. Hage
Prof. dr. H.J. van den Herik



Bart Verheij's home page - research - publications