Abstract
This paper describes one way in which a precise reason model of precedent could be developed, based on the general idea that courts are constrained to reach a decision that is consistent with the assessment of the balance of reasons made in relevant earlier decisions. The account provided here has the additional advantage of showing how this reason model can be reconciled with the traditional idea that precedential constraint involves rules, as long as these rules are taken to be defeasible. The account presented is firmly based on a body of work that has emerged in AI and Law. This work is discussed, and there is a particular discussion of approaches based on theory construction, and how that work relates to the model described in this paper.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
This paper is in part based on a revised version of work originally presented (Horty 2011a, b). This paper extends those papers by firmly locating the work developed from them in the context of AI and Law, and providing extensive discussion of points of similarity and difference with previous approaches to draw out its contribution to AI and Law.
The term “rule model” is itself due to Alexander, who applies it only to what I describe here as the strict rule model, which reflects his own preferred view; We use the term more broadly to apply to any model in which precedential constraint is thought to be carried by rules.
The idea is also, of course, closely related to the work of Jaap Hage on reason based logic. In Hage (1993) and Hager (1997, 2005), Hage provided a logical account of case-based reasoning on the basis of a reason—rather than a rule-based model. In Hage (1993), conflicts were resolved either through explicitly stated general principles, such as rule1 is more specific than rule2, or using numeric weights. The later books identify precedent cases as one sources of reasons why a set of reasons may be considered stronger than another, but, like Prakken and Sartor (1998) and Roth and Verheij (2004), these are always used to justify a form of a fortiori reasoning. The need to be able to go beyond a fortiori reasoning and explain how conclusions can be drawn when the precedents are not unequivocal is one of the motivations of this paper.
This approach can be found, for example in, Bench-Capon (1991).
See Schild and Herzog (1993) for an approach along these lines.
CATO was designed to support the teaching of reasoning with cases to law students; the two cited papers also report the results of using the system with students.
See Bench-Capon and Rissland (2001) for a discussion of this issue.
The approach was further developed in Bench-Capon (1999), which represented the power sets of pro-plaintiff and pro-defendant factors as separate lattices, and then joined them into a partial order using the precedent relations derived from the priority rules; in this way all factor combinations could be systematically represented. A different approach to a fortiori reasoning, developed in the setting of legal theory, can be found in Horty (2004).
This is not to understate the challenges involved in producing an analysis of cases in terms of factors. Certainly judges rarely if ever explicitly refer to factors, in the sense used here, in their decisions. Sometimes they do use some standard phrases (e.g. ‘imminent danger’ in the series of cases in Levi 1949) which is a good indicator of a factor. For the domain of CATO, Trade Secrets law, treatises such as the Restatement of Torts were used to supplement the decisions and were found to give a good guide to the considerations relevant to the decisions, and hence the factors that needed to be looked for. Automatic analysis of decisions in terms of factors has also been explored by Bruninghaus and Ashley in (2005) and Ashley and Brüninghaus (2009). Although there is no large scale validation of the approach with respect to legal practice, empirical studies in Aleven (1997) showed that the approach was effective in teaching law students how to reason with cases.
For the purpose of this paper, we take it that the rule underlying a court’s decision is plain, ignoring the extensive literature on methods for determining the rule, or ratio decidendi, of a case, which are rarely if ever explicitly stated by judges in their decisions. Validation of such rules would pose problems which are beyond the scope of this paper. We will also assume that a case always contains a single rule, ignoring situations in which a judge might offer several rules for a decision, or in which a court reaches a decision by majority, with different judges offering different rules, or in which a judge might simply render a decision in a case without setting out any general rule at all.
Note that, as in Prakken and Sartor (1998), the factors in the antecedent of the rule all favour the same side.
This general picture is described in much more detail in Horty (2007), which appeals to techniques from nonmonotonic logic to develop a detailed theory of reasons as the premises of default rules; the theory allows us to understand both how reasons can be defeated by stronger reasons and how they can be excluded from consideration, in the way suggested by Raz (1975).
Note that this is a departure from Prakken and Sartor (1998), in which precedents also gave rise to a rule for the losing side, albeit one with lower priority than the rule for the winning side.
Although we focus in this paper on which might be called “pure” common law, we believe the model would also apply to cases in which precedential reasoning is used to interpret statutory or constitutional language, or even in civil law jurisdictions. For the relevance of precedents to such reasoning see, for example, the papers in MacCormick and Summers (1997).
Although preference relations are of central importance in AI and Law treatments, such as those of Prakken and Sartor (1998) and Bench-Capon and Sartor (2003), the preferences defined there typically relate rules (and values), rather than reasons. Expressing preferences on reasons has similar advantages to the use of preferences between reasons rather than rules in the treatment of accrual of arguments; see Modgil and Bench-Capon (2010).
Prakken (2005) gives an example in which an additional fact may weaken a case: such a fact would mean that different factors are applicable to the cases and so this principle is not threatened. Thus, in his example, if hot weather favours jogging and wet weather favours not jogging but hot and wet together favour jogging, we have three factors, hot, wet and hot-and-wet, based on the two facts hot and wet rather than two factors and their combination.
Note that the rejected reason is the same as the antecedent in the losing rule formed using the method of Prakken and Sartor (1998).
Capturing this style of reasoning was the motivation of Bench-Capon (1999).
This example is modeled on some hypothetical cases considered by Prakken and Sartor (1998).
Note that the numbers of factors in a set should not be important. One factor may easily outweigh several factors. It is important that this is so, since the number of factors is always, to a greater or lesser extent, an artifact of the analysis: whether the facts of a case are represented as one factor or several will depend on the factors chosen by the analyst. We are, of course, relying on the factors capturing all significant nuances: for example whether the individual left her job may be detail enough, but we may need to distinguish between resignation and other reasons for termination of contract. This will depend on the analysis of the decisions.
Note that this positivist treatment of precedential constraint in particular does not entail any view of whether the law more generally should, or should not, be understood in positivist terms.
Had the court not considered f π5 to be by itself stronger than f δ5 , then other pro-plaintiff factors could have been included in r 4 to allow a decision for the plaintiff: for example the antecedent could have been {f π1 , f π2 , f π5 }, which would be a way of showing explicitly that the plaintiff’s position was stronger than in c 1, and would present weaker constraints on future decisions.
Although this account of precedential constraint is related to the reason model originally proposed by Lamond (2005), there are at least two significant differences between Lamond’s proposal and the present account. First, at a conceptual level, Lamond sets out his reason model as an alternative to the conventional view of precedential constraint, with its emphasis on rules, while, on the present account, the roles of reasons and rules are integrated: it is reasons that function as the premises of precedent rules, and the preference ordering on rules is derived from the preference ordering on reasons. And second, although Lamond himself does not develop a precise implementation of his reason model, the account set out here does seems to differ from what he has in mind. In particular, Lamond suggests (pp. 18–19) that the meaning of precedents should be understood in terms of “protected reasons” and other forms of exclusionary reasons. Although there may be merit in this idea, we did not, in the end, find it necessary to appeal to exclusionary reasons in developing the present account.
Chorley and Bench-Capon (2005), for example, demonstrates the rarity of finding a theory capable of explaining all the cases in a substantial case base. The difficulty may be further compounded by the analysis requiring some degree of subjectivity in the assignment of factors.
It is this stronger preference relation that is given by the reasoning in Bench-Capon (1999). The use of values, as in Bench-Capon and Sartor (2003), provides a relation of intermediate strength, where transitive reasoning is permitted for factors that promote the same value, but not for factors promoting different values.
Any reader who is unfamiliar with these issues is invited to consult Kornhauser and Sager (1986) and the extensive literature on judgment aggregation spawned by this paper, in addition to the standard work in economics on the derivation of group preferences from individual preferences.
See, for example, the discussion in Cross (1968, pp. 190–195); at times, other, more complicated situations are also referred to as cases of first impression, such as situations in which multiple precedent rules apply yet none clearly trumps the other, or situations in which entirely novel factors are present.
Of course, depending on their standing in the judicial hierarchy, a court may have the further option of overruling the precedent. We will not consider this further option in this paper, but see, for example, Wyner and Bench-Capon (2009).
We can assume that the set of precedent rules is finite, so that not every applicable rule can be trumped by another rule; as a result, if any rule is applicable to a fact situation, some rule must be binding—applicable and untrumped.
See Simpson (1961, p. 172), who himself cites Levi.
Lamond (2005) makes exactly this point—that following, as well as distinguishing, a previous precedent rule can change the law. However, the way in which he develops this idea indicates, we believe, a problem with his picture of case base dynamics and the evolution of legal doctrine. On Lamond’s view, the doctrine provided by a precedent rule, or ratio, consists, not only in the particular reason carried by that rule, but in the set of facts that have been judged as insufficient in strength to defeat that reason. Following a rule thus leads to a change in legal doctrine because: “every time a precedent is followed, further facts are added to the list of those regarded as insufficient to defeat the reason provided by the ratio (p. 17; see also p. 20).” To illustrate with our example, the doctrine provided by the case c 2 would, on this view, consist in the rule r 2, or \(\{ f^{\delta}_5 \} \rightarrow \delta, \) along with the list {f π1 , f π2 , f π4 } of factors judged by the c 2 court as insufficient in strength to defeat this rule; the application of the same rule in the case of c 11 would change the doctrine by expanding this list to {f π1 , f π2 , f π4 , f π5 , f π6 }, which includes also the new factors judged by the c 11 court as insufficient to defeat r 2. We feel, however, that the use of a simple list of individual factors that are viewed as insufficient to override a rule is not adequate, and instead rely here on a more general preference relation to represent strength comparisons among conflicting reasons—sets of factors, rather than individual factors—and it is this relation that is then updated as rules are applied. On our view, then, the doctrine of c 2 consists in the rule r 2 together with the information that {f δ5 } is preferred to the reason {f π1 , f π2 , f π4 } or any of its subsets, so that none of these reasons can defeat that rule; the case c 11 then adds the further information that {f δ5 } is likewise preferred to {f π5 , f π6 } or its subsets. But all of this is consistent with the idea that a court might eventually decide that {f δ5 } is not preferred to, say, the reason {f π4 , f π5 }—which contains a pair of potential defeaters, one from each of the two cases, which had not yet been considered in combination. It is hard to see how this latter possibility could be captured in Lamond’s list notation, which refers only to individual factors, rather than reasons, or sets of factors.
Note that it is possible for difference preferences to be expressed in different jurisdictions or at different times.
The terminology of values, in this context, was derived from Bench-Capon (2003).
The last two need to be used with caution since the preferences they add cannot be justified from the existing case base, and so tend to weaken the theory.
See Ashley and Bruninghaus’s IBP (2003). IBP was developed from CATO and represents a third direction of development: it broke down the decision in trade secret cases into a number of issues based on the Restatement of Torts, used the factors relevant to those issues to decide the particular issue at hand, and then combined these results.
Compare this with the model in Lindhal and Odelstat (2006), where the authors, following Ross (1957), view legal reasoning as a two step process: first from facts to intermediate predicates, and then from intermediate predicates to legal conclusions. Effectively, the approach studied in this paper, along with the theory construction approach, performs only the second of these steps; as noted by Ashley and Brüninghaus (2003), factors function as intermediate legal predicates, so that the first step is already carried out once a case has been analyzed in terms of factors. We will look further at work containing chains of reasoning below.
There are, of course, other ways of justifying these rules, such as an analysis of opinions in the cases.
References
Aleven V (1997) Teaching case-based argumentation through a model and examples. PhD thesis, Intelligent Systems Program, University of Pittsburgh
Aleven V, Ashley KD (1997) Evaluating a learning environment for case-based argumentation skills. In: Proceedings of the sixth international conference on artificial intelligence and law (ICAIL-97). The Association for Computing Machinery Press, pp 170–179
Alexander L (1989) Constrained by precedent. South Calif Law Rev 63:1–64
Alexander L, Sherwin E (2001) The rule of rules: morality, rules, and the dilemmas of law. Duke University Press, Durham
Ashley KD (1990) Modeling legal argument: reasoning with cases and hypotheticals. The MIT Press, Cambridge
Ashley KD, Brüninghaus S (2003) A predictive role for intermediate legal concepts. In: Bourcier D (ed) JURIX 2003: the sixteenth annual conference on legal knowledge and information systems. IOS Press, pp 153–62
Ashley KD, Brüninghaus S (2009) Automatically classifying case texts and predicting outcomes. Artif Intell Law 17(2):125–165
Bench-Capon TJM, Sartor G (2001) Theory based explanation of case law domains. In: Proceedings of the 8th international conference on artificial intelligence and law. ACM, pp 12–21
Bench-Capon TJM (2003) Persuasion in practical argument using value-based argumentation frameworks. J Logic Comput 13(3):429–448
Bench-Capon TJM (1991) Practical legal expert systems: the relation between a formalisation of law and expert knowledge. In: Bennun M, Narayanan A (eds) Computers, law and AI. Ablex Publishers, Norwood, pp 191–201
Bench-Capon TJM (1999) Some observations on modelling case based reasoning with formal argument models. In: Proceedings of the 7th international conference on AI and Law, pp 36–42
Bench-Capon TJM, Modgil S (2009) Case law in extended argumentation frameworks. In: Proceedings of the 12th international conference on AI and Law, pp 118–127
Bench-Capon TJM, Rissland EL (2001) Back to the future: dimensions revisited. In: Verheij B, Lodder A, Loui RP, Muntjewerff A (eds) JURIX 2001: the fourteenth annual conference on legal knowledge and information systems. IOS Press, Amsterdam, pp 41–52
Bench-Capon TJM, Sartor G (2003) A model of legal reasoning with cases incorporating theories and values. Artif Intell 150(1–2):97–143
Berman DH, Hafner CL (1993) Representing teleological structure in case-based legal reasoning: the missing link. In: Proceedings of fourth international conference on AI and Law, pp 50–59
Berman DH, Hafner CL (1995) Understanding precedents in a temporal context of evolving legal doctrine. In: Proceedings of fifth international conference on AI and Law, pp 42–51
Brüninghaus S, Ashley KD (2003) Predicting outcomes of case-based legal arguments. In: Proceedings of the ninth international conference on AI and Law, pp 233–242
Brüninghaus S, Ashley KD (2005) Reasoning with textual cases. In: Muñoz-Avila H, Ricci F (eds) ICCBR, Lecture notes in computer science, vol 3620. Springer, Berlin, pp 137–151
Chorley A, Bench-Capon TJM (2005) Agatha: using heuristic search to automate the construction of case law theories. Artif Intell Law 13(1):9–51
Chorley A, Bench-Capon TJM (2005) An empirical investigation of reasoning with legal cases through theory construction and application. Artif Intell Law 13(3–4):323–371
Cross R (1968) Precedent in English law, 2nd edn. Oxford University Press, Oxford
Hage J (1993) Monological reason-based logic: A low level integration of rule-based reasoning and case-based reasoning. In: Proceedings of the fourth international conference on AI and Law, pp 30–39
Hage J (1997) Reasoning with rules. Kluwer, Dordrecht
Hage J (2005) Studies in legal logic. Springer, Dordrecht
Horty JF (2004) The result model of precedent. Leg Theory 10:19–31
Horty JF (2007) Reasons as defaults. Philosopher’s Imprint 7(3):1–28
Horty JF (2012) Reasons as defaults. Oxford University Press, Oxford
Horty JF (2011) Reasons and precedent. In: Proceedings of the 13th international conference on artificial intelligence and law, pp 41–50
Horty JF (2011) Rules and reasons in the theory of precedent. Leg Theory 17:1–33
Kornhauser L, Sager L (1986) Unpacking the court. Yale Law J 96:82–117
Lamond G (2005) Do precedents create rules. Leg Theory 11:1–26
Levi E (1949) An introduction to legal reasoning. The University of Chicago Press, Chicago
Lindahl L, Odelstad J (2006) Intermediate concepts in normative systems. In: Goble L, Meyer J-J (eds) Deontic logic and artificial normative systems, Lecture notes in computer science, vol 4048. Springer, Berlin, pp 187–200
Loui RP, Norman J (1995) Rationales and argument moves. Artif Intell Law 3(3):159–189
Loui RP, Norman J, Olson J, Merrill A (1993) A design for reasoning with policies, precedents, and rationales. In: Proceedings of the 4th international conference on AI and Law, pp 202–211
MacCormick DN, Summers R (1997) Interpreting precedents: a comparative study. Dartmouth Publishing, Hanover
Modgil S (2009) Reasoning about preferences in argumentation frameworks. Artif Intell 173(9–10):901–934
Modgil S, Bench-Capon TJM (2010) Integrating dialectical and accrual modes of argumentation. In: Baroni P, Cerutti F, Giacomin M, Simari GR (eds) Computational models of argument. Proceedings of COMMA 2010, Frontiers in artificial intelligence and applications, vol 216. IOS Press, Amsterdam, pp. 335–346
Prakken H (2005) A study of accrual of arguments, with applications to evidential reasoning. In: Proceedings of the tenth international conference on artificial intelligence and law, pp 85–94
Prakken H, Sartor G (1998) Modelling reasoning with precedents in a formal dialogue game. Artif Intell Law 6:231–287
Raz J (1975) Practical reasoning and norms. Hutchinson and Company (second edition with new Postscript printed in 1990 by Princeton University Press, and reprinted by Oxford University Press in 2002; pagination refers to the Oxford edition)
Raz J (1979) The authority of law. Oxford University Press, Oxford
Rissland EL, Ashley KD (1987) A case-based system for trade secrets law. In: Proceedings of the first international conference on AI and Law, pp 60–66
Rissland EL, Xu X (2011) Catching gray cygnets: an initial exploration. In: Proceedings of the 13th international conference on AI and Law, pp 151–60
Ross A (1957) Tu-tu. Harvard Law Rev 70:812–825
Roth B, Verheij B (2004) Cases and dialectical arguments—an approach to case-based reasoning. In: Meersman R, Tari Z, Corsaro A (eds) OTM workshops, Lecture notes in computer science, vol 3292. Springer, Berlin, pp 634–651
Schauer F (1989) Is the common law law. Calif Law Rev 77:455–471
Schauer F (1991) Playing by the rules: a philosophical examination of rule-based decision-making in law and life. Oxford University Press, Oxford
Schild U, Herzog S (1993) The use of meta-rules in rule based legal computer systems. In: Proceedings of the 4th international conference on AI and Law, pp 100–109
Simpson AWB (1961) The ratio decidendi of a case and the doctrine of binding precedent. In: Guest AG (ed) Oxford assays in jurisprudence. Oxford University Press, Oxford, pp 148–175
Wyner AZ, Bench-Capon TJM (2009) Modelling judicial context in argumentation frameworks. J Logic Comput 19(6):941–968
Wyner AZ, Bench-Capon TJM, Atkinson KM (2011) Towards formalising argumentation about legal cases. In: Proceedings of the 13th international conference on artificial intelligence and law, pp 1–10
Author information
Authors and Affiliations
Corresponding author
Appendix: Observations and proofs
Appendix: Observations and proofs
Observation 1
Let \(\Upgamma\) be a case base containing two precedent cases of the form \(\langle X, r, s \rangle\) and \(\langle X, r', \overline{s} \rangle. \) Then \(\Upgamma\) is inconsistent.
Proof
Suppose the case base \(\Upgamma\) contains the cases \(c = \langle X, r, s \rangle\) and \(c^{\prime} = \langle X, r^{\prime}, \overline{s} \rangle. \) By our coherence constraints on rules and cases—particularly the requirements that the premise of a rule can contain only factors supporting its conclusion along with the requirement that the rule of a case must be applicable to its fact situation—we have (1) \(Premise(r) \subseteq X^s\) and (2) \(Premise(r') \subseteq X^{\overline{s}}\), and obviously (3) \(Premise(r) \subseteq Premise(r)\) and (4) \(Premise(r') \subseteq Premise(r'). \) From (2) and (3), we have Premise(r′) < c Premise(r) by Definition 2, and likewise \(Premise(r) {<}_{c^{\prime}} Premise(r^{\prime})\) by (1) and (4). We therefore have both \(Premise(r') {<}_{\Upgamma} Premise(r)\) and \(Premise(r) {<}_{\Upgamma} Premise(r')\) by Definition 3, and so Definition 7 tells us that \(\Upgamma\) is inconsistent. \(\square\)
Observation 2
Let \(\Upgamma\) be a consistent case base, and suppose X is a fact situation in which none of the rules from \(Rule(\Upgamma)\) are applicable. Then \(\Upgamma \cup \{ \langle X,r,s \rangle \}\) is also consistent, where r is any newly formulated rule applicable in X and supporting s as an outcome.
Proof
Assume that \(\Upgamma\) is consistent and that X is a fact situation in which none of the rules from \(Rule(\Upgamma)\) are applicable. Now suppose for contradiction that \(\Upgamma \cup \{c\}\) is not consistent, where \(c = \langle X,r,s \rangle\) with r a newly formulated rule supporting s. Since \(\Upgamma\) is consistent but \(\Upgamma \cup \{c\}\) is not, the preference relation derived from c must conflict with the preference relation derived from some other case already belonging to \(\Upgamma\)—that is, there must be reasons A and B such that (1) A < c B and (2) \(B {<}_{c^{\prime}} A,\) where \(c^{\prime} = \langle Y, r^{\prime}, \overline{s} \rangle\) is a case from \(\Upgamma. \) From (1), we have (3) \(A \subseteq X^{\overline{s}}\) and (4) \(Premise(r) \subseteq B\) by Definition 2, and likewise, from (2), we have (5) \(B \subseteq Y^s\) and (6) \(Premise(r') \subseteq A. \) From (3) and (6), together with the fact that \(X^{\overline{s}} \subseteq X\) we have \(Premise(r') \subseteq X, \) and of course r′ belongs to \(Rule(\Upgamma), \) so that some rule from \(Rule(\Upgamma)\) is applicable to X, contrary to assumption. \(\square\)
Observation 3
Let \(\Upgamma\) be a consistent case base with \(Rule(\Upgamma)\) the derived set of rules, and suppose X is a fact situation in which some rule r from \(Rule(\Upgamma), \) supporting the outcome s, is binding. Then the case base \(\Upgamma \cup \{ \langle X,r,s \rangle \}\) is also consistent.
Proof
Assume that \(\Upgamma\) is consistent with X is a fact situation in which the rule r from \(Rule(\Upgamma), \) supporting the outcome s, is binding. Now suppose for contradiction that \(\Upgamma \cup \{c\}\) is not consistent, where \(c = \langle X, r, s \rangle. \) Since \(\Upgamma\) is consistent but \(\Upgamma \cup \{c\}\) is not, the preference relation derived from c must conflict with the preference relation derived from some other case already belonging to \(\Upgamma\)—that is, there must be reasons A and B such that (1) A < c B and (2) \(B {<}_{c^{\prime}} A, \) where \(c^{\prime} = \langle Y, r^{\prime}, \overline{s} \rangle\) is a case from \(\Upgamma. \) From (1), we have (3) \(A \subseteq X^{\overline{s}}\) and (4) \(Premise(r) \subseteq B\) by Definition 2, and likewise, from (2), we have (5) \(B \subseteq Y^s\) and (6) \(Premise(r') \subseteq A. \) From (3) and (6), together with the fact that \(X^{\overline{s}} \subseteq X, \) we have \(Premise(r') \subseteq X, \) so that the rule r′ is applicable in the situation X. From (4) and (5), we have (6) \(Premise(r) \subseteq Y^s, \) and obviously (7) \(Premise(r') \subseteq Premise(r'). \) By Definition 2, then, (6) and (7) tell us that \(Premise(r) {<}_{c^{\prime}} Premise(r^{\prime}), \) from which it follows that \(Premise(r) {<}_{\Upgamma} Premise(r'). \) Since the rule r′ from \(Rule(\Upgamma), \) which supports a different conclusion, is both applicable in X and stronger than r, it follows from Definition 5 that r is trumped. By Definition 6, therefore, the rule r is not binding, contrary to assumption. \(\square\)
Rights and permissions
About this article
Cite this article
Horty, J.F., Bench-Capon, T.J.M. A factor-based definition of precedential constraint. Artif Intell Law 20, 181–214 (2012). https://doi.org/10.1007/s10506-012-9125-8
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10506-012-9125-8