AustLII Home | Databases | WorldLII | Search | Feedback

Journal of Law, Information and Science

Journal of Law, Information and Science (JLIS)
You are here:  AustLII >> Databases >> Journal of Law, Information and Science >> 1994 >> [1994] JlLawInfoSci 14

Database Search | Name Search | Recent Articles | Noteup | LawCite | Help

Scott, Brendan --- "Legal Expert Systems: A Practioner's Perspective" [1994] JlLawInfoSci 14; (1994) 5(2) Journal of Law, Information and Science 227

Legal Expert Systems
A Practitioner's Perspective

BRENDAN SCOTT[*]

Abstract

The purpose of this article is to explain what role a legal practitioner sees for expert systems in a legal environment, having regard to "pragmatic" considerations. That is, what things are considered important given the perspective of day to day practice in a legal office? What things would a Legal Expert System (LES) have to consider in order to be of use to a lawyer advising a client?

1. Introduction

The practice of law is primarily a pragmatic pursuit. Current discussion of expert systems in law is of a theoretical nature. In general, analyses of the operation of legal expert systems have been undertaken in the context of a university computing environment1. This article highlights those functions desirable for an expert system to be of use in private practice, primarily by an exposition of those functions required for the conduct of such a practice. In particular, use will be made of two case studies, one a litigious action, the other a commercial advice.

As this discussion progresses, it will become apparent that certain aspects of giving advice cannot be incorporated in traditional expert systems reliant on rule- or case-based reasoning, because the reasons behind a conclusion, the "why", are subordinated by the answer. One of the reasons that the "why" of an answer is overlooked by these systems is related to the paradigm which serves as their basis. That paradigm, logical inference, promotes a highly dynamic view of the law in which the conclusion, rather than its context, is elevated to a position of alternative, but ultimately equivalent, paradigm of equivalence. The equivalence paradigm has the advantage of making the context of a conclusion or statement explicit, therefore less likely to be lost as the formalism is processed.

The basis of this inquiry is comprised of two hypothetical situations. In the first, action must be taken in a litigious context. In the second, a commercial advice must be given. These two have been chosen as, for a commercial law firm, they together comprise the majority of the work that comes in.

The practice considered in this paper is that of a firm in metropolitan Sydney.

2. Litigation

Litigation is comprised of various levels of escalation, and may be regarded as primarily an "active" branch of the law. The term is usually associated with action in court, but covers a broader range of activities. Some of the things involved in litigation include writing letters of demand, arranging trap purchases, taking witness statements, briefing counsel, negotiating settlements and conducting mediations. We will, however, refer to litigation in the context of a case to be run before a court.

In the course of conducting litigation, legal advice is given to the client. This advice is more of a short term (ie course of the proceedings), tactical nature. This can be contrasted with the advice given in a commercial context which is usually of a long term, strategic nature.

Because advice and decisions made in litigation are tactical in character, they are liable to change or may become invalid in response to emerging appraisals of the opponent's positions or tactical considerations arising from the manoeuvring of the opponent. Litigation advice is volatile in relation to a few well defined but poorly determined facts2. It is important that this is kept in mind during the conduct of the litigation. An initial decision not to investigate a particular avenue may have to be revised in light of the opponent's later use of that information.

The nature of the litigation process, whereby interaction is largely written, and the parties are bound by the issues they have pleaded, lends a high degree of definition to the representation of the relevant legal "atoms". In effect, when pleading in response, the legal issues are often manifest and the alleged facts are certainly so (the purpose of pleadings is to expressly state the material facts on which a party will rely in the conduct of its case). Although it is improper, some parties go so far as to expressly plead what they believe to be the relevant legal issues.

The conduct of litigation depends upon the relative weight each party can have assigned to its version of the facts. Some facts may be legally established, but inconsequential. Others, the court may express doubt about, yet be crucial. The (apparent) relative weight3 of the facts may swing erratically during the conduct of the proceedings.

3. Commercial Advice

Commercial advice is advice asked of a lawyer in a commercial context. That advice may be, for example, what the implications of a given contract are, how a party can or should structure its business affairs, what the tax implications of certain action are, or whether certain action is legal and, if not, why not.

Commercial advice is of a strategic nature, in that it seeks to establish legal defences against future attack4. Commercial advice in relation to a contract has the effect of clarifying the obligations of each party and, in a sense, setting the rules for litigation if, at some time in the future, something goes wrong. It gives the inquirer a level of comfort in relation to legal assaults that might be made against him or her.

In giving commercial advice, facts are often assumed, but are mutable. The problem is in determining whether the set facts establish a legal conclusion and the extent to which they can be modified without altering the conclusion. The facts are poorly defined but well determined, in that their relevance will remain largely unchanged, but can themselves be subject to variation which is not insignificant.

The relevant facts can be changed explicitly by the client or, as some times happens, implicitly as the lawyer becomes more acquainted with the problem. For example, the lawyer may give advice based on prototypes of equipment, which is later invalidated because it is not technically feasible to mass produce the equipment to the specifications of the prototype. Sometimes the client itself is unsure of the facts or may inadvertently fail to refer to a pertinent detail.

Because the facts can be changed the client is interested to know how and to what extent they can be, or must be changed. Thus, in advising a client, it is insufficient to respond to their query by "Dear Madam, further to your query of 27/8/94: 'yes'. Yours sincerely ..." Rather, legal conclusions must be substantiated by reasoning. The client sees the "why" to be just as important as the "what" because it may suggest that things which they have control over are relevant. If they can change them, they alter the legal conclusion. Similarly, if there is a legal conclusion they are concerned to guard against, they need to know what (or what not) to do to avoid it.

4. Case Studies

Given the general background above, it is appropriate to present two practical examples of the processes a lawyer goes through when asked to give advice. They show that, even though the lawyer may know the relevant aspects of the law, it is still appropriate that automation be applied to help him or her structure the advice or argument. Of course, if the lawyer is not familiar with the law, an expert system might direct him or her as appropriate.

5. Case Study (Litigation)

5.1 Facts

Your client retains you to commence a court action against a debtor company for US$50,000 for goods sold and delivered. The client wants to claim interest at the rate specified in the contract - 20% per annum.

5.2 Method

In order to commence action, one must first draft and file a statement of claim. The starting point to that is knowing in which court the action must be brought, as different courts deal with claims for different monetary amounts.

Decide on the relevant court

Draft pleadings

capacity

contract

payment in $US

interest

order

shipment

non-payment

damage

- principal

- interest

claim

- damages

- costs

File and serve statement of claim

Diarise service date plus 28 days for entry of default judgment.

Advise client of status of action.

With this in mind, we would expect the computer to:

(a) accept information we know about the claim.

(b) determine what Court is relevant.

(c) produce a check list of action, perhaps similar to the following:

Note that this action is for US$50,000, and should therefore be brought in the District Court: s44(1) District Court Act.

(i) Draft pleadings (forms 1,9)

company search on defendant (for registered office address)

capacity

contract

breach

damage

(ii) File (fee = $135, [2050] O'Grady (take 4 copies))

(iii) Serve

(iv) diarise service date plus 28 days for default judgment

(v) advise client of progress

Once this level is reached, details need to be flushed out, both of the structure of the action, and of the means by which each element will be proved.

This will require the checklist to be modified as follows:

contract

existence

payment in US$ (cl 5(a))

payment within 30 days (cl 5(b))

interest rate (cl 5(c))

orders made (invoices)

breach

non payment,

continuing refusal to pay

damage

principal

interest

claim

principal

interest

contractual

under rules (s83A)

costs

In the above scheme we have implicitly included further relevant legal and non-legal information. We have, for instance, included:

• information about why the action is taken in the District Court (s44(1) of the District Court Act sets the jurisdiction of that court at $250,000);

• which court forms to use

• what the filing fee is, and where to find it (at paragraph [2050] of O'Grady's District Court Practice - the looseleaf service most often used in the office);

• that 4 copies should be taken to the registry;

• how certain contractual terms will be evidenced (clause references, and invoices, these will become the particulars of the claim); and

• where the power to claim interest comes from.

All of this information may be relevant as the matter progresses. Some of it, eg to take 4 copies, would be inappropriate for inclusion in a traditional LES, but is of great importance practically. If you only take the original to the registry, it will be stamped and filed leaving nothing to serve on the defendant! (in fact the registry will not accept a single original without a service copy and a copy for the affidavit of advice).

6. Case Study (Commercial)

6.1 Facts

Your client comes to you seeking trade practices advice in relation to the validity of disclaimers in its computer software licences, which are shrinkwrapped in its products. The products are sold for $100 each.

6.2 Method

In your opinion, there may be an issue of consumer protection. Action is essentially in the form of a checklist, but is closer to the traditional form of a rule based system (here section numbers are from the Trade Practices Act (Cth) 1974):

supply of goods to a consumer?

implied conditions under:

s69 (title, no encumbrance, quiet possession)

s70 (correspondence with description)

s71 (quality or fitness)

s72 (correspondence with sample)

supply of services to a consumer?

implied conditions under:

s74(1) due care and skill

s74(2) fitness for purpose

is the s68A limitation relevant?

do disclaimers go beyond s68A limitation?

give standard warning re s52, s53,

suggest alternative disclaimer.

While this looks relatively straight forward, as practitioners we realise that the position of computer software as "goods" is unclear5. We can be sure that they are either goods or services, but not which of these in particular is applicable. In writing our advice therefore both possibilities must be taken into account, possibly stating which is favoured.

A checklist of the letter of advice would be something like:

Introduction

Summary

Explanation of consumer protection provisions.

applies to supply of goods or services to consumer

explain supply to a "consumer" (s4B) (definition)

explain supply to a "consumer" (this instance)

explain goods or services

consequent warranties if goods

consequent warranties if services

explain s68A liability limitation

restriction to "not of a kind ordinarily acquired for personal domestic or household use or consumption"; applicability in this case

explain possible sanctions

s52 misleading or deceptive (civil)

s53(g) misleading or deceptive (criminal)

Conclusion

In this case the applicable law is not clear, making traditional rule based systems unhelpful. However the indeterminacy of the law is no bar to the lawyer giving advice and should not be a bar to the use of automated reasoning, provided that reasoning is flexible enough to allow for both alternatives simultaneously. In this case the LES would function as an intelligent outliner system6.

7. Contrasting

It is demonstrated that a computer system that "merely" knows and answers the law is not of practical use to a practitioner in advising his or her client. In a litigious setting it is important that the lawyer can integrate the legal conclusions with practical implementation information. In a commercial setting the client not only wants to know whether or not a disclaimer is unlawful, but why it is unlawful and/or how it can be modified to make it lawful.

In neither situation would one expect the computer to anticipate the exact details of the two actions or advice. One would expect it however to provide a base template that could be adapted for the case in point and which, in turn, could later serve as a base template for similar action and later serve as an expert (sub-)system in relation to the matter in point.

8. Customisability

The previous section shows the need for the computer system to admit some degree of customisability in relation to the facts of a given case. The computer system, however, also needs to be customisable for what might be termed "local" legal knowledge. For example, that one should take 4 copies when filing a document, or that one should claim s83A interest as an alternative to contractual interest7. One could not expect either of these or the myriad of other local legal idiosyncrasies to be anticipated by an expert system. Nor would one expect it to contain other useful, non-legal knowledge8. We should expect, however, that the system could be customisable through normal usage to take account of them as a user may choose (or perhaps even learn these local aspects heuristically).

9. Mnemonic

Arising partly out of customisability and partly out of system transparency, the system functions as a mnemonic. The litigation system, for example, records what evidence will be used in the course of the trial to establish a given legal proposition. It can also be used to highlight what issues yet remain to be proved, or addressed. It could conceivably be integrated with litigation support systems to provide an overview of the entire case strategy.

10. Localised Inference

Expert systems are premised on the assumption that legal characterisations are either true or false, but not both. In the commercial advice case study that question did not admit a definitive answer. This does not necessarily militate against an expert system coping with the "problem", as long as it has a design which is suitably fluid to allow it to accept both propositions as true and investigate mutual consequences (even if the propositions are, logically, mutually exclusive). This localised, or ad hoc inferencing, would be driven by the operator, who would guide the LES along promising routes.

11. Intuitive Response

The information required in each case should be able to be entered in the form in which it is set out above. In practice a lawyer already has a good idea of what will be relevant in deciding an issue. Instead of being forced through a tedium of questions answered as "don't know" (perhaps, more appropriately "don't care"), the lawyer should be allowed to specify those legal points that he or she thinks are relevant. The computer should fit together whatever of those pieces it can.

Alternatively, there could be an option for the lawyer to be further interrogated, perhaps as a memory jog, at his or her request.

12. A New Paradigm?

At present LES work with a paradigm of logical implication - if A, B and/or C then D9. This paradigm is essentially dynamic in character, and promotes the assumption that it is the end, D, which is important, rather than the collection of facts and their relative inter-relatedness10. This paper shows that, as a practical matter, this is not the case. What is required of an expert system is a means to identify the answer to the problem, and to help structure an explanation of why, because it is the "why" which is the most important part of the advice.

For example, if the answer is what the client would like to hear, it will need to know whether it will remain valid if, for example, its marketing structure or some other part of its business changes. Otherwise the client will want to know what it can, or must do, to reach a positive outcome.

To this end therefore, a different, though equivalent formalism is posited - that of set theoretics and equivalence.

13. Assumptions

Underlying our approach are two assumptions, which we state explicitly here.

1. The law is capable of formal representation.

This statement says that there exist formal methods by which the law can be represented. Some examples may be case law and statute law.

2. No representation can be complete.

By this we mean much more than what is contemplated by, for example, Goedel's Theorem. This statement should be read to mean that no formal system can ever completely describe that which it is intended to represent11.

Thus "the law" is more than the sum of all its representations. This implies that no LES can be complete, and that therefore an adequate LES must incorporate some means of "ad hoc" reasoning.

14. The Equivalence Paradigm

14.1 Statement of form

The main form of the equivalence paradigm is

a ~ b; or a \subset b

where a and b are arbitrary representations.

14.2 Meaning

The symbols

a ~ b

read "the meaning of a is equivalent to the meaning of b". For example

"A has a cause of action " B owes A a duty of care;

in tort against B" ~ B breached that duty; and A has suffered damage because of it."

The point being that the right hand side of the equation may be thought of as one (legal) atom if that is appropriate, not three.

The symbols

a \subset b

read "the meaning of a is contained within the meaning of b."

The symbol "" means "implies".

15. Why "Equivalence"?

The adoption of an equivalence paradigm implies stability and stasis. Where before the user of an implicative system was forced by its dynamism to a single conclusion in isolation, now he or she finds an equilibrium which expresses the conclusion in the context of the discussion as a whole, and which shows the structure of the discussion on its face. The aim under an equivalence regime is to demonstrate how given facts are reflected in the "pre-existing and eternal" system of law12.

16. Formal Difference

There is no formal difference between traditional syllogistic methods, and an equivalence paradigm. For example technically:

(a ~ b) ~ ((a b) and (b a))

Similarly

(a \subset b) ~ (a b)

The difference is a practical one, and arises in the way we conceptualise the law based on these formalisms. An implicative formalism suggests a dynamic conceptual structure, whereas one based on equivalence suggests a static one. An equivalence based structure is more amenable to the proper contextualisation of a given case. It is less likely to inhibit localised or informal reasoning as it is seen as being less rigid than a method of implication13. Finally, it will support "sideways chaining" more readily than an implicative system. Where, for example we have a collection of equivalences:

a ~ d;

b & c ~ d;

i & e ~ f;

g & h ~ i

we can more easily move sideways through equivalences

a ~ f

d ~ i & e

d ~ g & h & e

or other combinations as appropriate because the inferencing does not preconceive any "consequent" as being the object of the chaining. Thus it is possible to express the relationship between, say, a legal relation d and the underlying facts which establish d, or, alternatively, to search for possible relationships between two atoms. The system does not constrain us to search for those things the LES designer considered important (ie the conclusion, or the initial premises).

17. Practical Difference

The practical difference which arises from the use of an equivalence-based paradigm is that we may more readily manipulate one formal representation to arrive at legally equivalent representations. That is, sideways chaining can be put to practical use with the computer keeping track of how the argument has developed, including the assumptions that have been made.

In effect, the computer implements an initially pre-defined but dynamic ad hoc logic to a given representation of the legal "initial conditions" with a view to arriving at a "final solution" (ie a formally equivalent representation, given the transformation rules of the logic) which is more legally comprehensible. Thus, through a sequence of moves it might track through from:

A told B that C would work properly

and C worked at 5Hz

and C should have worked at 10Hz

and A had no reasonable grounds for believing C would work at 10Hz

To:

A told B that C would work properly

and C didn't work properly

and A had no reasonable grounds for believing C would work properly.

At this point the lawyer would posit the equivalences:

"C would work properly" ~ D; and

"C didn't work properly" ~ "D was incorrect".

Then the LES could deduce:

A told B D

and D was incorrect

and A had no reasonable grounds for believing D

The LES, recognising this as a pre-existing equivalence would then conclude:

A has engaged in misleading and deceptive conduct under s 52 of the Trade Practices Act 1974.

Further, the equivalence paradigm is not restricted to first order statements about legal atoms. It is conceivable that certain forms of legal reasoning, including "ad hoc" legal reasoning, may be equivalent to others in the same fashion that the contrapositive of an implication is logically equivalent to the statement alone.

Under an equivalence paradigm information corresponding to a user's initial query can be analogised to mercury poured over an uneven surface - it glides along the channels until it finds the place in which it best fits. A logical paradigm purports to predict its final rest before it falls and to place it there without any human interaction.

18. Conclusion

The role of a computer therefore is both as teacher and pupil. It presents a practitioner with its "best guess" guide as to how to proceed in the form of a conceptual skeleton. Where, in the practitioner's opinion, that skeleton is incomplete or incorrect, the computer must accept and learn from the practitioner's suggestions ("tell me to take 4 copies for filing next time"). Where appropriate, the computer can suggest information that might be helpful. The practitioner could either supply that information or posit equivalences. Those equivalences may imply the information or render it irrelevant. The LES will record and store those equivalences for later use and produce, or suggest, appropriate implications immediately. Where necessary the practitioner will use the computer as a form of "matter notebook", recording salient facts relevant only to that matter. To a large extent the LES is not recording the "what" of the answer so much as the "how".

A conceptual skeleton can also form the basis of an accounting system for concepts. With properly integrated word processing software, it will be able to provide a summary of what points remain to be addressed in the advice. When a document undergoes a number of substantial revisions, there is the danger that a given concept will be lost from one draft to the next. Such a system would be of invaluable use in that context to give a continual update of what concepts remain to be addressed.

This paper has attempted to show that in a commercial practice, the most important part of a lawyer's job is explaining why, or how a conclusion is reached. A mere answer of "yes" or "no" is of little worth without further material which substantiates and qualifies the conclusion14.

This substantiation should be left to the lawyer, being the one most closely acquainted with the facts of the case, the sophistication of the client15, and the import of the answer to the inquiry.

Perhaps the most important point made by this paper is that law is inherently a slippery subject. Its very fluidity almost has the standing of an independent justification for its existence. If a computer is to be successfully applied in this area it must be able to cope with that fluidity. Finally, the law must present its answers in the context of its questions. If a question must be reshaped too much, the answer becomes irrelevant. A LES must be prepared to accept formulations as given to it and to present not only an answer, but a contextualisation of that answer. An equivalence paradigm is one means of achieving these goals.

End Notes

1 See, for example, the notes cited in the bibliography.

2 That is, the statement of the facts is, to a large extent, static (well defined), but whether or not they have been established and given sufficient weight will not be known until judgment has been handed down (poorly determined).

3 Where, during the discussion, reference is made to "weight" accorded to the "facts", in neither case are measurable qualities referred to. In each case they should be interpreted as subjective evaluations by the relevant party. Thus, if the weight of a fact shifts, that should be interpreted as "the relevant party reassesses the importance of a given fact, and accords it more, or less, significance".

4 That is not to say all commercial advice is defensive in character, only that that is what we have adopted for discussion. These arguments are valid mutatis mutandis, when applied to other advice.

5 See, for example: ASX Operations Ltd v Port Data Australia Pty Ltd (1990) 27 FCR 460, Toby Constructions Products Pty Ltd v Computa Bar (Sales) Pty Ltd [1983] 2 NSWLR 48.

6 LMCS, xxii.

7 Some judges are known for not allowing s83A interest in an award if it was not claimed in the originating process. Thus, if contractual interest cannot be made out, the plaintiff is left with no interest on the debt.

8 Such as the name and telephone extension of the firm's filing clerk.

9 See, eg., Susskind, 133ff; Moles 91, 139; Zeleznikow 91, 174ff. While not prima facie evident, this is also true of case based systems in that the operation of a case based system can be viewed in the nature of: If the values of the attributes are a1, a2, a3, then the answer is D. This structure is implicit in the case based system's distance function.

10 This has been realised to some extent already within the literature, Gray 92, Conover.

11 That is, for example, that there are true statements that cannot be represented in the system.

12 Whether or not the law is "in fact" "pre-existing and eternal" is not a question considered by this paper. It suffices to say that the law may be (and has been) idealised as such.

13 Indeed, the use of "meaning" in the definitions of the symbols preserves the malleability of the system.

14 It should be noted in passing also that irrelevant detail should not be included.

15 Legally trained clients require a higher degree of substantiation through the use of citation than a layman, who, in contrast, can find citation a hindrance.

Bibliography

[Allen 82] Layman Allen, Towards a Normalized Language to Clarify the Structure of Legal Discourse; in Deontic Logic, Compututional Linguistics and Legal Information Systems, Volume II, A.A. Martino (ed), North-Holland Publishing, 1982, at page 349

[Allen 86] Layman Allen and Charles Saxon, Analysis of the Logical Structure of Legal Rules by a Modernized and Formalised Version of Hohfeld Fundamental Legal Conceptions, in Logic, Informatics, Law, Martino and Socci Natali (eds), Elsevier Science Publications, North Holland, 1986, at page 385.

[Conover] Matt Conover, Applying Three-Dimensional Thinking and Systems Technologies to Jurisprudence and Legal Management; Paper presented at the First Annual Conference Systems Theory and Information Technology in Jurisprudence and Legal Education, Yale Law School, June 26-27, 1986.

[Gray 92] Pamela Gray, A three Dimensional Model of Legal Language, Invited Paper for the Second International Conference on Substitute Technology in the Law School, Chicago, USA 30 July - 1 August, 1992.

[Hunter] Daniel Hunter, Alan Tyree, John Zeleznikow, There is Less to this Argument than Meets the Eye, Journal of Law and Information Science, Vol 4, No 1, 1993 at 46.

[Kowalski] Andrej Kowalski, Case-based Reasoning and the Deep Structure Approach to Knowledge Representation, in Proceedings of the Third International Conference on Artificial Intelligence and Law, 1991 at page 21.

[LMCS] Legal Management Consulting Services Pty Limited, 1992 Australasian Legal Software Directory, Washington Press, Sydney, 1992.

[Miller] Russell Miller, Annotated Trade Practices Act, 15th Edition, The Law Book Company, Australia, 1994.

[Moles 91] Robert Moles, Logic Programming - An Assessment of its Potential for Artificial Intelligence Applications in Law, Journal of Law and Information Science, Volume 2, No. 2, at page 137.

[Moles 92] Robert Moles and Surend Dayal, There is More to Life than Logic, Journal of Law and Information Science, Volume?, No?, at page?

[Popple] James Popple, Shyster: A Pragmatic Legal Export System; PhD Thesis, Australia National University, 1993.

[Skalak] David Skalak and Edwina Rissland, Argument Moves in a Rule - Guided Domain, in Proceedings of the Third International Conference on Artificial Intelligence and Law, 1991 at page 21.

[Stamper] Ronald Stamper, The Parallel Development of Laws and Their Related Information Systems, Onderzoek Wetgeving en Informatievoorziening, draft as at 31 January 1992.

[Susskind] Richard Susskind, Expert Systems in Law, Clarendon Press, Oxford, 1989.

[Zeleznikow 91] John Zeleznikow, Building Intelligent Legal Tools - The IKBALS Project, Journal of Law and Information Science, Volume 2, No 2, at page 165.

[Zeleznikow 92] John Zeleznikow and Daniel Hunter, Rationales for the Continued Development of Legal Expert Systems, Journal of Law and Information Science, Volume 3, No 1, at page 96.


[*] Brendan Scott BSc(Hons) LLB is a solicitor in the Sydney office of the firm Mallesons Stephen Jaques, working in their Trade Practices, Intellectual Property and High Technology group.


AustLII: Copyright Policy | Disclaimers | Privacy Policy | Feedback
URL: http://www.austlii.edu.au/au/journals/JlLawInfoSci/1994/14.html