Arguing with Dimensions in Legal Cases

Abstract

In this short paper we introduce some novel arguments based on legal cases represented as sets of dimensions. Although there has been much work on arguments based on factors, there has been relatively little recent work which has exploited dimensional representations. The arguments that we consider have a flavour quite distinct from those generated from cases represented as factors, which lack notions of more and less favourable, and the original argumentation with dimensions from HYPO and CABARET, which always has the complete set of active dimensions in view. We would argue that the arguments produced have a good intuitive appeal and correspond to some kinds of argument commonly found when reasoning with legal cases. We consider in this paper arguments based on one and two dimensions, leaving additional dimensions largely for future work. Illustrations are provided using a set of cases well known and frequently represented in AI and Law. Although admittedly preliminary, we believe that arguing with dimensions in this way may enable a useful addition to the repertoire of legal arguments based on cases which can be used in AI and Law.

CCS CONCEPTS

• Computing methodologies → Reasoning about belief and knowledge;

INTRODUCTION

One of the pioneering systems modelling argument from precedent cases, HYPO [ 5 ], modelled cases using dimensions: relevant aspects of the case which ranged from an extreme pro-plainti↵ point to an extreme pro-defendant point. Following the successful use of factors in CATO [ 3 ], however, factors became the basis of most argumentation in AI and Law. Factors have the advantages of simplicity in that they are boolean, either present or absent, that they always favour the same party to the dispute and that they can be organised in a hierarchy of increasingly abstract factors, whereas dimensions have degree, may favour either side, and may or may not be applicable to a case. For a detailed discussion of the di↵ erences, see [ 21 ]. CATO itself gave a set of nine argumentation moves to be used to deploy factors in argumentation, and further examples of argumentation based on factors can be found in [ 19 ] and [ 24 ]. A similar form of argumentation is found in systems based on Reason Based Logic [ 14 ] and [ 23 ]. Although dimensions were used for the argumentation of CABARET [ 22 ] and never disappeared entirely [ 10 ], there has recently been a growing recognition that the boolean nature of factors is insu ciently expressive for all purposes [ 1 ], and a consequent revival of the use of dimensions [ 2 ].

Dimensions have been used for in argumentation in [ 20 ], but there the schemes are very similar to schemes based on factors, albeit the factors are replaced by points or ranges of dimensions. In [ 20 ], although dimensions are seen as the source of factors, and a link between facts and factors, the structure of dimensions is not exploited in the argumentation. In this paper we want to explore distinctive ways of arguing enabled by dimensions. We will illustrate our work with a representation of the wild animals cases introduced by Berman and Hafner [ 11 ] and subsequently much discussed in AI and Law [ 6 ]. The cases are Pierson v Post, Keeble v Hickersgill, Young v Hitchens, Ghen v Rich and Popov v Hayashi. Briefly the cases concern plainti↵ s chasing wild animals (in the first four cases) when their pursuit was interrupted by the defendant. Post was chasing a fox for sport with horse and hound, when Pierson intervened, killing the fox with a fence pole. Keeble was hunting ducks on his own land to sell them, but Hickersgill maliciously scared them away by firing guns, Young was a commercial fisherman and was in the process of closing his nets when Hitchens, another commercial fisherman, entered the nets and made o↵ with the fish Young was collecting. Ghen was a commercial whaler, who had harpooned a whale with a bomb lance, but the line broke and Rich found the whale when it washed ashore and sold it, in ignorance of the universally accepted convention that the iron holds the whale. Popov v Hayashi, which cited these wild animals cases, concerned disputed ownership of a baseball (valuable because it had been hit by Barry Bonds to break a home run record). Popov was attempting to complete his catch when he was assaulted by a mob of fellow spectators and Hayashi (who had not taken part in the assault) ended up with the baseball when it rolled free. The wild animals cases were cited when considering whether Popov’s e↵ orts had given him possession of the ball. These cases have been represented many times, with variations, in AI and Law, but we will use dimensions and case representations based on [ 8 ] in which dimensions were used to drive narratives based on cases. The dimensions and their points are shown in Table 1 and the cases we will use are represented as sets of dimension points in Table 2.

The next section will consider arguments based on one dimension. 2

ARGUING

WITH ONE DIMENSION

Let us begin by considering a single dimension.

A dimension runs from an extreme plainti↵ point to an extreme defendant point. Between these are two points of interest: the most favorable point for the plainti↵ at which cases have been found for the defendant, and the least favorable point for the plainti↵ at which cases have been found for the plainti↵ . It is of course possible that the first of these will coincide with the extreme pro-plainti↵ point and/or the the second will coincide with the extreme pro-defendant point. The two points may also themselves coincide. These points are derived from precedents and so will be liable to shift as more cases are decided, or they may derive from hypothetical cases used in commentaries or obiter dicta. The dimension can thus potentially be divided into three zones, A, B and C, although A and/or C may not exist, in which caset B will the whole range of the dimension, and the dimension is never seen as decisive for either party. Alternatively Zone B may be reduced to a single point, in which case the dimension will often play an especially crucial role in deciding the case.

Suppose now we wish the argue for the plainti↵ . We begin by locating the case on the dimension. There are three possibilities, according to whether our case falls in Zone A, B or C.

• In Zone C the dimension provides no argument for the plainti↵ . • In Zone A there is a strong argument for the plainti↵ • in Zone B there is a an argument for the plainti↵ , but it is not decisive, 3

ARGUMENTS IN TWO DIMENSIONS

in those zones dimension 1 is insu ciently pro-P to form the basis of an argument. Thus the plainti↵ will have advanced an argument of the form of either A1 or A2. Suppose A1 has been used, so that we are in Zone F or Zone I. The defendant now needs to find a dimension that will put the case in Zone I, and argue that the new dimension is more important. Consider an hypothetical case (KeebleHypo) where the animal had been mortally wounded but where the defendant owned the land. Now the plainti↵ puts forward the argument A5 A5 In KeebleHypo, closeness of possession is decisive. The quarry was mortally wounded, and certain capture can be enough for the plainti↵ , as Keeble shows. and the defendant can respond to A5 with A6, relying on the defendant’s ownership of the land:

A6 In Keeble,the plainti↵ owned the land, but in KeebleHypo the defendant owned the land, which is decisive, since Land Ownership takes precedence over closeness of possession.

If the priority claim is accepted, and it seems to be plausible, the ball is returned to the plainti↵ ’s court. The plainti↵ must now find a dimension which, when considered with Land Ownership, will put the case in Zone I and take priority over Land Ownership. The plainti↵ ’s motive would put the case in the right Zone, but it remains likely that the preference will stay with land ownership. There is, however another way of using dimensions which may save the plainti↵ ’s case. Suppose, for example that the animal in KeebleHypo just happened, on this one occasion, to be on defendant’s land. Now the plainti↵ can appeal to Quarry-Land-Connection, and counter A6 with A7:

A7 In KeebleHypo the defendant owned the land, but the quarry had no connection with the land, and so land ownership is not relevant.

This illustrates the possible need for one dimension to reach a certain threshold before another comes into play. Or suppose the quarry was a whale and the defendant owned the beach. Now, if we consider the convention dimension, we have a conflict between two dimensions, similar to that of A6 and if the convention takes precedence over land ownership, the plainti↵ can still win.

Now suppose that the plainti↵ put forward A2, so that we are in one of the three central zones, B, E, H. The defendant needs to find a dimension which will put the case in Zones E or H, since a case in Zone B cannot defeat A2. A case in Zone H will defeat A2: certain capture is not enough to overcome defendant’s ownership of the land. Since the use of A2 suggests that some cases on this part of the dimension have been found for the plainti↵ and others for the defendant, it may well be that it is a third dimension, such as land ownership, which discriminates between them.

The more interesting cases are where we are in Zone E: here we have reasons for and against, and we must strike a balance between them ([ 15 ], [ 13 ]). The idea is that Zone E can be divided by a line (not necessarily a straight line, any curve is possible) separating it into a defendant area (to the north-west of the line) and a plainti↵ area (to the south-east). It may also be that a third dimension provides a threshold which requires consideration in this zone. For a discussion of the use balance and thresholds in statutory interpretation see [ 4 ].

Figure 4 shows how precedents constrain Zone E for closeness of pursuit and defendant motive (we put all four of decided cases in Zone E for the purposes of this example). Each of the precedents will claim an area of the Zone for its side: to the north-west for a case decided for the defendant and to the south-east for a precedent decided for the plainti↵ . Given enough cases we might try to apply statistical line fitting techniques, but it is unlikely that su cient cases are available, and the use of statistical techniques puts argumentation and explanation into a black box. As can be seen from Figure 4 it is very possible that the case under consideration will occupy an area of Zone E yet to be resolved. At this point it is possible to argue using hypothetical variations which bring the case under a precedent. Thus, the plainti↵ may try to align Popov with Ghen: 4

A8 Popov should be decided as Ghen: If Popov had caught the ball it would have been a clear case since Rich had the same motive as Hayashi. But in both cases possession was lost through no fault of the plainti↵ .

In contrast the defendant can try to bring the case under Young:

A9 Popov should be decided as Young: If Hayashi had been professionally engaged, it would have been a clear case, since Young was nearer to possessing the quarry than Popov. But in both cases defendant was attempting to earn money. but better is Pierson: A10 Popov should be decided as Pierson: In both cases, the plainti↵ were in hot pursuit. But Hayashi had a better motive than Pierson.

It is now a question of which argument the judge finds more satisfying. The hypothetical in A8 is probably correct, but I personally would go with A9 because there is a big gap between what Popov did and completing the catch. Moreover Pierson seems to o↵ er an even better precedent. Given the strength of A10, plainti↵ in Popov would be well advised to avoid arguing on the basis of closeness of pursuit (once he had failed to show that a catch had been made). The nature of the interference and the plainti↵ ’s own motive would take Pierson out of play. 4

MORE DIMENSIONS

In the previous section we saw two reasons why a third dimension might be required by one of the parties: • To provide a threshold for a dimension to be e↵ ective, as in A7 • To provide a stronger argument to counter a currently winning argument, as in the response to A6, when the planiti↵ switches focus from the closeness of KeebleHypo’s pursuit to the nature of the act which interrupted his pursuit.

In both cases, however, we can continue to consider the dimensions pairwise, using the diagram of Figure 3. Two questions now arise: (1) Can we always consider dimensions pairwise, or are there cases where we need to consider three or more together? (2) Are there other ways in which a third dimension needs to be brought into consideration?

The first of these questions relates to other issues which have arisen before when considering argumentation with cases in AI and Law. For example, when we are resolving arguments based on values, as in [ 7 ], can the value conflicts be resolved pairwise, or is in necessary to compare sets of values as in [ 17 ] and [ 9 ]? Also related is whether arguments need to be allowed to accrue as in [ 18 ] or [ 16 ], or whether they can be be held to remain separate but complimentary.

We will not go into these issues in the short paper, the purpose of which is no more than to introduce the notion 5 of reasoning with cases represented as sets of dimensions, as currently understood. These are, however, issues that will need to be confronted if this style of argumentation is pursued. 5

CONCLUDING REMARKS In this short, preliminary. paper we have draw attention to a current trend to represent legal cases as dimensions and considered the opportunities for generating novel styles of argument based on this representation. We argue that the arguments have a natural flavour and enable to use of arguing with hypotheticals which was an intention of [ 5 ] but which was not fully realised.

By looking at arguments with one and two dimensions we have seen the kind of arguments that can be generated, with one side proposing an argument based on one dimension and then the other countering by introducing a second dimension, and two ways in which these can be countered, by switching to a di↵ erent dimension, or showing that a third dimension supplies a threshold for the applicability of one of the dimensions.

A number of questions require resolution for this motivating sketch to begin to attain maturity, including: • We need to make more formal the dimension based arguments sketched in section 2 and 3; • We need to consider whether dimension based arguments can always be reduced to pairs, or whether three, or more, dimensions may need to be consider simultaneously in some cases; • We need to determine whether dimensions interact in ways additional to one supplying a threshold for the applicability of another; • We need to link the arguments into a dialectical exchange capable of implementation: for example, a dialogue game in the manner of [ 12 ].

Although this does leave a considerable amount of work to be done, this paper can be seen as opening a door to reveal new directions for exploration. We believe that just as dimensions are an invaluable tool in the representation of cases, they may be able to provide the means of extending the repertoire of arguments available to AI and Law.

References

[1] L. Al-Abdulkarim , K. Atkinson , and T. Bench-Capon. Evaluating the use of abstract dialectical frameworks to represent case law . In Proceedings of the 15th International Conference on Artificial Intelligence and Law , pages 156 - 160 . ACM, 2015 .

[2] L. Al-Abdulkarim , K. Atkinson , and T. Bench-Capon. Angelic secrets: bridging from factors to facts in us trade secrets . In Legal Knowledge and Information Systems - JURIX 2016: The Twenty-Ninth Annual Conference , 2016 .

[3] V. Aleven . Teaching case-based argumentation through a model and examples . PhD thesis , University of Pittsburgh, 1997 .

[4] M. Araszkiewicz . Argument structures in legal interpretation: Balancing and thresholds . In C. Dahlman and T. Bustamante, editors, Argument Types and Fallacies in Legal Argumentation , pages 129 - 150 . Springer, 2015 .

[5] K. D. Ashley . Modeling legal arguments: Reasoning with cases and hypotheticals . MIT press, Cambridge, Mass., 1990 .

[6] K. Atkinson . Introduction to special issue on modelling Popov v . Hayashi. Artificial Intelligence and Law , 20 ( 1 ): 1 - 14 , 2012 .

[7] K. Atkinson and T. Bench-Capon. Legal case-based reasoning as practical reasoning . Artificial Intelligence and Law , 13 ( 1 ): 93 - 131 , 2005 .

[8] T. Bench-Capon and F. Bex . Cases and stories, dimensions and scripts . In Legal Knowledge and Information Systems: JURIX 2015 : The Twenty-Eighth Annual Conference , pages 11 - 20 . IOS Press, 2015 .

[9] T. Bench-Capon , H. Prakken , and W. Visser . Argument schemes for two-phase democratic deliberation . In Proceedings of the 13th International Conference on Artificial Intelligence and Law , pages 21 - 30 . ACM, 2011 .

[10] T. Bench-Capon and E. Rissland . Back to the future: dimensions revisited . In Legal Knowledge and Information Systems - JURIX 2013 : The Fourteenth Annual Conference . IOS Press, 2001 .

[11] D. H. Berman and C. D. Hafner . Representing teleological structure in case-based legal reasoning: The missing link . In Proceedings of the Fourth International Conference on Artificial intelligence and Law , pages 50 - 59 , 1993 .

[12] T. F. Gordon . The pleadings game . Artificial Intelligence and Law , 2 ( 4 ): 239 - 292 , 1993 .

[13] T. F. Gordon and D. Walton . Formalizing balancing arguments . In Computational Models of Argument - Proceedings of COMMA 2016 , pages 327 - 338 , 2016 .

[14] J. Hage . A theory of legal reasoning and a logic to match . Artificial Intelligence and Law , 4 ( 3 -4): 199 - 273 , 1996 .

[15] M. Lauritsen . On balance . Artificial Intelligence and Law , 23 ( 1 ): 23 - 42 , 2015 .

[16] S. Modgil and T. Bench-Capon. Integrating dialectical and accrual modes of argumentation . In Computational Models of Argument: Proceedings of COMMA 2010 , pages 335 - 346 , 2010 .

[17] H. Prakken . An exercise in formalising teleological case-based reasoning . Artificial Intelligence and Law , 10 ( 1-3 ): 113 - 133 , 2002 .

[18] H. Prakken . A study of accrual of arguments, with applications to evidential reasoning . In Proceedings of the 10th International Conference on Artificial Intelligence and Law , pages 85 - 94 . ACM, 2005 .

[19] H. Prakken and G. Sartor . Modelling reasoning with precedents in a formal dialogue game . Artificial Intelligence and Law , 6 ( 3 - 4): 231 - 87 , 1998 .

[20] H. Prakken , A. Wyner , T. Bench-Capon , and K. Atkinson . A formalization of argumentation schemes for legal case-based reasoning in aspic+ . Journal of Logic and Computation , 25 ( 5 ): 1141 - 1166 , 2015 .

[21] E. L. Rissland and K. D. Ashley . A note on dimensions and factors . Artificial Intelligence and Law , 10 ( 1-3 ): 65 - 77 , 2002 .

[22] D. B. Skalak and E. L. Rissland . Arguments and cases: An inevitable intertwining . Artificial intelligence and Law , 1 ( 1 ): 3 - 44 , 1992 .

[23] B. Verheij , J. Hage , and A. R. Lodder . Logical tools for legal argument: a practical assessment in the domain of tort . In Proceedings of the 6th International Conference on Artificial Intelligence and Law , pages 243 - 249 . ACM, 1997 .

[24] A. Wyner and T. Bench-Capon. Argument schemes for legal casebased reasoning . In Legal Knowledge and Information Systems - JURIX 2013: The Twentieth Annual Conference , pages 139 - 149 , 2007 .

[14] H. Prakken . Argumentation Logics: Games for abstract argumentation . http://www.sta↵ .science.uu.nl/⇠ prakk101/al/ chongqing10.html, 2010 . [Online; accessed 01-April-2017].

[15] R. S. Sutton and A. G. Barto . Reinforcement learning: An introduction , volume 1 . MIT press Cambridge, 1998 .

[16] G. A. Vreeswik and H. Prakken . Credulous and sceptical argument games for preferred semantics . In European Workshop on Logics in Artificial Intelligence , pages 239 - 253 . Springer, 2000 .

[17] E. Wiewiora . E cient Exploration for Reinforcement Learning . PhD thesis , Citeseer, 2004 .

[18] M. Wooldridge . An introduction to multiagent systems . John Wiley & Sons, 2002 .

[19] M. Wooldridge . An introduction to multiagent systems . John Wiley & Sons, 2009 .

[20] T. Yuan , D. Moore , and A. Grierson . Computational agents as a test-bed to study the philosophical dialogue model” de”: A development of mackenzie's dc . Informal Logic , 23 ( 3 ), 2003 .

[21] T. Yuan , D. Moore , and A. Grierson . A human-computer debating system prototype and its dialogue strategies . International Journal of Intelligent Systems , 22 ( 1 ): 133 - 156 , 2007 .

[22] T. Yuan , D. Moore , and A. Grierson . A human-computer dialogue system for educational debate: A computational dialectics approach . International Journal of Artificial Intelligence in Education , 18 ( 1 ): 3 - 26 , 2008 .

[23] T. Yuan , J. Schulze , J. Devereux , and C. Reed . Towards an arguing agents competition: Building on argumento . In Proceedings of IJCAI2008 Workshop on Computational Models of Natural Argument , 2008 .

[24] T. Yuan , V. Svansson , D. Moore , and A. Grierson . A computer game for abstract argumentation . In Proceedings of the 7th Workshop on Computational Models of Natural Argument (CMNA07) , 2007 .