All the Right Notes: Artificial Intelligence and the Future of the Common Law

The Sixteenth Sir Harry Gibbs Memorial Oration

Download RTF - 132 KB

It is a singular privilege to have been invited to deliver the Sir Harry Gibbs Memorial Oration.

The list of earlier orators includes many of the most distinguished lawyers this country has produced. Their subjects have ranged widely. Some have reflected upon constitutional doctrine, others upon freedom of speech, the nature of judicial power or the evolution of legal principle.¹ What has struck me, however, in re-reading those orations over recent months is their remarkable unanimity in one respect. Whatever subject they ultimately chose to address, each paused to reflect upon Sir Harry himself and, in doing so, each described essentially the same man.

One finds repeated references to Sir Harry’s judgment, courtesy, restraint, fidelity to legal principle and an instinctive appreciation of the proper limits of judicial power. Even if one has regard to the customary courtesies extended to former judges in orations named after them, these qualities recur in the speeches with such uniformity that one is left with the distinct impression that these encomia have the benefit of actually being true.

I cannot add any personal reminiscences to those already recorded. By the time I ended my legal studies, Sir Harry had already retired from the High Court and my understanding of him came through his judgments.² However, what was immediately evident, even to an inexperienced reader, was an easy confidence that disputes were able to be made digestible and then solved by disciplined reasoning according to law.

The recurring emphasis upon the personal discernment of Sir Harry has occupied my thoughts for some time because it reflects the reality that the legitimacy of judicial power ultimately depends upon something more than intellectual ability: it depends upon the disciplined exercise of judgment.

Some years ago, Paul Keating famously remarked that if you walked into any pet shop in Australia, the “resident galah” would be “talking about microeconomic policy”.³It was a wonderfully graphic way of capturing the moment when a subject had suddenly become part of the national conversation.

I have begun to think that artificial intelligence has reached much the same stage. There are some genuinely thoughtful contributions (some of which I will mention) but one scarcely attends a legal conference without finding an earnest panel of speakers, each sporting an emblazoned conference lanyard, explaining either that AI is about to transform legal practice beyond recognition or that it poses an existential threat to the administration of justice. There is often, heaven help us, a PowerPoint presentation. There are invariably references to exponential change and the pace of technological development. Before very long someone solemnly informs us that lawyers who fail to embrace artificial intelligence will soon be replaced by those who do.

I do not intend to add another voice to that increasingly crowded aviary.

I readily accept that artificial intelligence will prove to be the most significant development affecting legal practice in my lifetime. But after listening to many presentations upon the subject (of markedly variable quality), I have increasingly found myself wondering whether many have become obsessed by the wrong question.

The discussion generally begins with the capabilities of the technology. Will it replace the new generation of junior solicitors? Will it draft submissions? Will it alter the economics of legal practice? These are all legitimate (albeit obvious) questions, but they are not the questions that most interest me.

The genesis of this lecture lay elsewhere.

Earlier this year, while preparing reasons for judgment in ASIC v Bekier,⁴ I found myself reflecting upon a phenomenon that has become so commonplace in modern commercial life that we seldom pause to examine it.

The directors whose conduct I was considering did not suffer from a shortage of information.⁵

I observed that it is now notorious that the board pack of a public company no longer arrives as a modest bundle of papers designed to assist judgment; instead, board packs often represent an electronic publishing project. No longer posted or delivered by a bike courier, the uploaded document is often heroically vast.⁶ I remarked that after seeing many such documents, spanning many proceedings as a barrister and judge, a common pattern can be seen: often there are executive summaries, followed by longer summaries of the summaries, followed by detailed papers, followed by appendices, annexures, and technical attachments whose purpose appears to be not so much enlightenment as completeness. Each section is defensible, and, in isolation, apparently sensible, but taken together, they are oppressive.

The problem confronting those directors was being focussed (and then keeping focussed) on what truly mattered. It struck me that this was not simply a problem confronting company directors; it had also become a characteristic of modern legal practice.

Not long after I had delivered that judgment and settled upon this topic, Chief Justice Bell delivered the Harold Ford Memorial Lecture dealing with AI and directors’ duties.⁷ It is an important contribution. His lecture coincided with my own view that the real significance of artificial intelligence lies partly in what machines will be able to do, but more importantly in what their increasing capabilities force us to ask about what human beings do when they exercise judgment.

That thought, which I pondered when considering directors’ duties, has continued to occupy me.

Every judge who has had responsibility for complex commercial litigation has witnessed the extraordinary growth in the volume of information that now accompanies the resolution of legal disputes. Electronic discovery has transformed litigation (and generally not for the better). Millions of documents can now be identified, collected, searched and produced in circumstances where, not very many years ago, the relevant material would have occupied only a handful of lever-arch folders. Indeed, when Sir Harry was a barrister in Brisbane, it might have been a relatively thin manilla folder.

Written submissions have often become oppressive. Electronic Lists of Authorities regularly reproduce cases that stand for propositions that are blindingly obvious (and a miscellany of obscure cases that would never have been reported in an age of greater discrimination). Expert reports commonly extend to hundreds of pages.

The extravagant nature of much modern litigation has occurred because each procedural innovation, viewed in isolation, has appeared sensible, each development in technology has made additional information available, and each refinement has promised a more complete understanding by a party of the controversy before the Court. The cumulative consequence, however, has been profound and, in my view, increasingly unmanageable.

An important and interesting question is how artificial intelligence can now be harnessed to address these unwelcome developments. I will return to this later in these remarks.

But a more profound question is what the increasing sophistication of these technologies reveals about those aspects of the management and resolution of litigation that genuinely require the exercise of human judgment. Put another way, what will be left for the judge?

I think we need to think differently about what it is that lawyers and judges contribute to the administration of justice. Much of what has traditionally occupied our time has involved acquiring and organising information, and it is hardly surprising that a profession which has attached considerable value to those skills (and derived profit from their application) should regard with apprehension a technology which can perform many of them with extraordinary speed. But they are not, I think, what lies at the heart of what good lawyers and good judges do. What matters is the capacity to exercise judgment: to understand which facts really matter, to distinguish the important from the merely available, to appreciate the consequences of competing choices, and to decide what should be done in circumstances in which knowledge of all the relevant information does not itself supply the answer.

That prompts examination of the quality we call “judgment” (for which Sir Harry was rightly praised) and reflection upon what is it about its exercise that cannot simply be reproduced by ever more sophisticated processing of information?

These questions did not arise with the development of artificial intelligence or even with the Industrial Revolution. They have occupied philosophers and theologians for more than two millennia. Indeed, one of the striking features of returning to those earlier writers is the realisation that they were grappling with problems that are recognisably our own. They asked what distinguishes wisdom from mere knowledge and what qualities are required of those entrusted with responsibility for others.

Let me refer to a very recent exploration of an aspect of this topic. One of the unexpected pleasures of preparing this Oration was reading Pope Leo XIV’s recent encyclical Magnifica Humanitas.⁸ I suspect that a number of those present may be wondering what place a papal encyclical has in a lecture concerned with the future of the common law and artificial intelligence. The dinner speaker tonight is far more qualified to speak of things theological.⁹ But the answer, I hope, will soon become apparent.

Put shortly, it is because I found myself increasingly persuaded that his Holiness had identified the question that almost everybody else discussing artificial intelligence in the legal context had either overlooked or had paid insufficient attention.¹⁰

The Pope begins from a different starting point from the ubiquitous conference speaker discussing artificial intelligence. He asks what it means to be human.

That is an unusual place to begin a discussion about technology. It is also, I think, the right place.

Pope Leo, of course, expressly acknowledges that technology has always accompanied human civilisation and has often greatly improved the conditions under which people live. The issue, as he sees it, is not technology itself, but rather whether technology remains the servant of humanity or gradually becomes its master. The issue facing the world is therefore whether we have retained a sufficiently coherent understanding of mankind to know the purpose for which that power ought to be used.¹¹

That struck me because, although expressed theologically, it seemed to describe a question with which the common law has quietly wrestled for centuries.

Lawyers in this room would probably find the expression “legal anthropology” faintly alarming or pretentious (or both). At first glance, it does seem the sort of concept which might be thrown around by undergraduates after they have been introduced to critical legal studies.

The hesitation of some in reflecting on abstract philosophical concepts is unsurprising as many are sceptical of grand theory, noting that the common law proceeds incrementally.

But, at bottom, anthropology is the study of humanity and of what it means to be human. As lawyers, we must recognise that every doctrine of law or equity proceeds, whether explicitly or implicitly, from assumptions about the nature of the human person: that individuals possess reason and are capable of deliberation; that human choice matters (which is why we distinguish accident from intention); that motive and purpose matter (which is why negligence is different from fraud); and that justice sometimes requires strict legal rights to be tempered by equitable principles informed by conscience.

These aspects of our legal system might fairly be described as anthropological. On any view, they reflect a particular conception of what it means to be human.

That is why I found the Pope’s discussion so illuminating. He counsels that before we ask what artificial intelligence can do, we should first ask what sort of beings we are asking it to assist and what values we are asking it to give effect to or facilitate.

The Pope develops this theme through two striking images which possess a significance extending well beyond their religious origins. The first is the Tower of Babel; the second is the rebuilding of Jerusalem. They represent, in his account, two fundamentally different ways of understanding progress.¹² Babel is the image of a civilisation captivated by technical capacity, convinced that mastery, uniformity and power alone can secure human flourishing. Jerusalem, by contrast, is rebuilt through shared responsibility, dialogue, humility and recognition that genuine community cannot be engineered simply by increasing technical capability. The distinction is not between technology and its absence but between a technology ordered towards domination and one ordered towards a shared perception of the common good.

It is here that the Pope’s encyclical intersects with one of the oldest questions in Western philosophy.

Long before the common law emerged, and centuries before Christianity itself, Aristotle confronted the distinction between knowledge and judgment. In the Nicomachean Ethics, particularly in Book VI, he distinguished scientific knowledge, technical skill and practical wisdom.¹³ The distinction reflects something fundamental about human decision-making. Knowledge and technical competence will take us a very long way, but there remain decisions about what ought to be done which admit of no mechanical answer. It is in those circumstances that wisdom and practical judgment matter.

It seems to me that the common law has always been, whether consciously or unconsciously, an expression of that insight. Judges should be appointed because, through experience, learning, temperament and character, they have demonstrated an ability to exercise judgment in circumstances where competing considerations must be weighed and where certainty can be elusive.

If that is right, the arrival of artificial intelligence should cause us less to fear for the future of judging than to think much more carefully about what judging has always been.

There is a less elevated way of making the same point. Like many Australians of my generation (and I suspect a good number of people in this room), I grew up on a steady diet of British comedy in the 1970s courtesy of the ABC. Among the programmes that were compulsory viewing in our home was The Morecambe & Wise Show.¹⁴ One sketch has endured particularly well and is readily available on YouTube.¹⁵

André Previn, appearing as himself, is attempting to conduct Grieg’s Piano Concerto in A minor while Eric Morecambe is doing unspeakable things to the piano part. Eventually Previn can bear it no longer and protests that Morecambe is playing all the wrong notes, at which point Eric rises from the piano, grabs the distinguished conductor by his lapels with magnificent mock anger and gives the explanation upon which the whole sketch turns. He says: “I’m playing all the right notes, but not necessarily in the right order!”¹⁶

The conceit works because the defence is simultaneously absurd and, at a sufficiently literal level, plausible. The constituent pieces may all be present, but that is plainly not enough.

There is something in that joke which captures the distinction I have been trying to draw. Information supplies the notes; judgment makes the music. A machine may become extraordinarily proficient at locating the notes, arranging them according to a specified criterion or criteria and identifying relationships between them which no unaided human mind could hope to discover. Those capacities are of immense potential value to the law. But the judicial function is to understand what matters, what does not, how competing considerations are to be weighed and, ultimately, how the law is to be applied to human circumstances.

There is, however, another aspect of this which I think requires attention because judgment does not take place in a vacuum. It operates within a tradition and, more fundamentally, within an inherited understanding of the values by reference to which choices are made. It is here that the natural law tradition, which has been lurking somewhat unobtrusively behind what I have said thus far, assumes importance.

I do not propose to embark upon an excursion through the controversies that have surrounded natural law theory over the centuries, still less to suggest that the common law can be explained as some uncomplicated application of Thomist philosophy. The history is far richer and more complicated than that.¹⁷ The more modest point I wish to make is that the common law developed within a civilisation in which there existed, notwithstanding profound disagreements about many things, a broadly shared conviction that law was not simply an exercise of organised power, that reason was capable of saying something about how human beings ought to live together, and that certain values were not created merely by the fact that the sovereign had chosen to recognise them.

The intellectual ancestry of those ideas is long. Aristotle’s account of practical wisdom, to which I have already referred, was taken up and transformed within the Christian natural law tradition. Aquinas’s celebrated account of law as an ordinance of reason directed to the common good, to which I have referred elsewhere,¹⁸ matters for present purposes because it expresses a conception of law which profoundly influenced Western legal thought: law is rational, it has purposes external to the mere exercise of power, and those purposes are connected with the good of persons living together in community.¹⁹ The common law grew within a shared understanding of the role the law played in an ordered and cohesive society with shared values.

To say that conduct is dishonest, an action is reasonable, or that a response is proportionate, is necessarily to make an evaluative judgment. The law can formulate rules which guide that judgment, sometimes with considerable precision, but it cannot eliminate the need for judgment by reference to shared values.

For much of the common law’s history this did not present any difficulty because the legal system operated within a society possessing a coherent moral vocabulary. There existed a shared intellectual inheritance, derived from classical thought, Christianity, the natural law tradition and the historical experience of English institutions, which allowed lawyers to argue about justice, conscience, reasonableness, duty and responsibility while understanding the language in which the argument was being conducted.²⁰

The genius of the common law was that it did not ordinarily need to state that inheritance as some sort of philosophical creed. Until relatively recent times, it could take much of it for granted. Its characteristic method was testing general propositions against particular circumstances and developing principle. But that practicality of method should not be mistaken for an absence of underlying (but often unarticulated) values.

This brings us back, by a rather long route, to Pope Leo. What I take to be one of the central insights of Magnifica Humanitas is that technology cannot answer the question of the ends to which technology should be put. The encyclical describes artificial intelligence as a valuable tool, but insists that its use be responsible and transparent and directed towards human dignity and the common good. ²¹

The warning is important because the more capable the instrument becomes, the easier it is to mistake capacity for purpose. A machine may tell us with astonishing sophistication how an objective can be achieved; it cannot, merely by becoming more sophisticated, establish why that objective ought to be preferred.

There is a particular danger for lawyers in all this. It is tempting to imagine that the threat posed by artificial intelligence is that machines may gradually come to think like human beings. I am more concerned about the converse: that lawyers may gradually come to think like machines, partly because we have come to mistake the accumulation and processing of information for judgment, but also because the erosion of the shared values which once supplied the common law with an assumed moral vocabulary may make it increasingly difficult to identify the ends by reference to which judgment is to be exercised.

The grotesquely enlarged board pack I described in Bekier, the millions of documents captured by electronic discovery, the hundreds of pages of expert opinion and the ever-expanding written submissions are not products of artificial intelligence.²² We managed to create all those things ourselves and the technology arrived after the habits were formed.

And this is where the practical consequences become important. Decisions whose legitimacy depends upon the exercise of human judgment within the traditions I have described should not be delegated to AI; but there is equally little sense in insisting that human beings continue to perform, at enormous expense, tasks involving no such judgment which machines may now be able to perform incomparably more efficiently.

At this point I can imagine some of you wondering whether we have travelled rather a long way from the practical administration of justice. I want to suggest that we have not. Indeed, unless the preceding discussion has some practical consequences for the way in which courts conduct litigation, it is of limited interest to me.

The Federal Court has already taken a first step. Earlier this week, the Court’s Use of Generative Artificial Intelligence Practice Note was launched, which recognises the potential of artificial intelligence to increase efficiency, while seeking to guard against the obvious risks to the integrity of the judicial process, including errors such as hallucinated authorities and the misuse of confidential material.²³ Similar guidance has now emerged elsewhere in Australian courts and the profession.²⁴ All of this is sensible and necessary, but thus far has been directed principally to regulating the use of artificial intelligence within the processes we already have.

What I have in mind as the next stage is more ambitious: courts themselves identifying aspects of litigation in which artificial intelligence can be harnessed and used to reduce cost and complexity. Let me identify three possibilities.

I Expert Evidence

The first arises from an aspect of commercial litigation with which I have become increasingly frustrated: the use of expert evidence. Although there are many aspects of opinion evidence I could pick, I will focus on one that was featured in a judgment I delivered a couple of weeks ago: event studies in securities class actions.²⁵

As I explained in that judgment, an event study is a technique of financial economics. Broadly speaking, it seeks to identify whether the return on a security around a particular event differs from the return which would otherwise have been expected. But anyone who has observed securities class actions over the last generation will be familiar with a curious regularity.

The applicant commonly produces an expert whose event study identifies a significant abnormal return attributable to the relevant corporate disclosure; the respondent commonly produces another expert whose analysis reaches the opposite conclusion. I remarked recently that a cynic might regard this as the econometric equivalent of the medical reports once tendered respectively by plaintiffs and insurers in District Court personal injury litigation.

Event studies are models. Their results depend upon assumptions concerning such matters as the estimation window, benchmark index, factor structure and event window, while the treatment of confounding information may itself depend upon contestable judgments about what information was conveyed to the market at a particular time.²⁶ The apparent mathematical precision of coefficients and statistical significance should not obscure the fact that the counterfactual return is inferred through a model constructed upon assumptions.

Anyone experienced in litigation will recognise that adversarial bias of experts remains a real problem, despite efforts to address partiality.²⁷ Additionally, the costs are enormous. Those costs extend well beyond the fees of the experts themselves, substantial though they may be and, particularly where overseas experts are retained, sometimes extraordinary, because around each report grows an elaborate apparatus: the formulation of assumptions, preparation of briefs, conferences with solicitors and counsel, drafts, supplementary instructions, responsive reports, conclaves, joint reports and cross-examination.

Considerable professional skill is devoted to the process. But experienced judges and barristers learn that more often than not the dispute ultimately returns to the factual assumptions upon which the opinions depend.²⁸

There will remain many cases in which specialised knowledge can only sensibly be supplied by a suitably qualified human expert, but I am far from convinced that the present process is the only conceivable one.

One can at least imagine a different model in some cases. Where the real controversy concerns the factual premises upon which specialised knowledge is to operate, the Court might first determine those premises, or identify a constrained range of permissible factual assumptions, and then use an appropriately designed artificial intelligence system to assemble and synthesise the relevant body of specialised knowledge. The parties would retain the ability to identify error, challenge the material upon which the synthesis was based and call expert evidence where a genuine controversy remained. There would plainly be difficulties to work through, including ensuring that the parties and the Court could identify the sources upon which the synthesis was based and test its reliability, but I see no reason why those difficulties should be insuperable. Courts have long used a form of special jury, a referee, to assist with specialised questions without surrendering the judicial function.²⁹ The essential point is that the Court remains responsible for the judgment; the machinery by which the relevant specialised knowledge is assembled need not remain frozen in its present form.

The object would not be to replace the expert with a machine merely because a machine is cheaper. It would be to identify more precisely what it is that we genuinely require supposedly independent experts to do.

II Litigants in person

My second example concerns a very different problem. Litigants in person rarely have nothing to say. Sometimes submissions are made carefully and economically, but what is vexing are those cases in which a potentially significant point lies buried within a mass of material that is emotional, legally misconceived or simply very difficult to follow. The judge cannot responsibly assume that because ninety-nine propositions are hopeless, the hundredth is also hopeless. There may be a nugget among the dross, with the result that considerable judicial and opposing party resources can be consumed identifying the case actually being advanced before anyone can begin to determine whether it has any merit.

Significant use of AI in all types of civil litigation is not something that is coming; it has already come. Self-represented litigants are already using AI and, as the technology becomes ubiquitous, they will do so in vastly greater numbers. We might as well move a step beyond merely seeking to control its use by the parties.

Curiously enough, after I had drafted this Oration, the Australian Financial Review reported last Monday what seemed to me a rather striking illustration of the possibilities.³⁰ An academic, apparently ignorant of industrial law when his dispute began, successfully represented himself before the Fair Work Commission using artificial intelligence in what appears to have been a transparent and directed way: to identify a potential claim, analyse his material, locate and verify authorities, anticipate opposing arguments and assist him in presenting his case. The same report referred to the Commission’s very different experience with large volumes of undisciplined AI-generated material containing hallucinated authorities and meritless argument. That contrast in use by self-represented litigants is important, but even now there is probably nothing particularly novel about what happened in the academic’s case (other than it received publicity).

One possibility would be a court-controlled artificial intelligence tool through which a self-represented litigant might be required to put their material. The system could require the litigant to identify the orders sought, the material facts said to justify those orders, the legal propositions relied upon, and the evidence said to support them, and then produce a structured synthesis which remains linked to the litigant’s original material. Rather than allowing artificial intelligence simply to generate ever more legal verbiage, the Court would be using it to impose some order and discipline upon material which might otherwise obscure the case the litigant is trying to advance.

That would not confer an illegitimate forensic advantage. The opposing party benefits because it can understand the case it must meet; the judge benefits because the real propositions can be identified without wading through the brume of irrelevant material; and, most importantly, the litigant benefits because an arguable point is less likely to disappear beneath the manner in which it has been expressed.

There would plainly need to be safeguards. The original material must remain available, the system must not silently abandon an argument because it considers it weak rather than unarguable, the litigant must be able to correct the synthesis, and the judge must remain alert to the possibility that the machine has misunderstood something. There may be difficulties in designing such a system, but they are unlikely to be insurmountable ones.

None of this would require the courts becoming software developers. In practical terms, a court could procure access to existing commercial models through a secure, court-controlled portal, specify the universe of material upon which the model may operate and the tasks it may perform, and insist that its outputs remain traceable to their sources and subject to human supervision. That approach would also be broadly consistent with the direction of the Commonwealth’s GovAI project, which is developing secure infrastructure through which government can make controlled use of commercial AI models without becoming permanently tied to a particular provider.³¹

Sometimes when I hear the expression “access to justice” used unthinkingly by persons who stand to make money out of providing such access, I confess I feel like reaching for my revolver. But this proposal could provide real access to justice. If artificial intelligence is going to place something approaching sophisticated legal assistance within the reach of people who could never afford to purchase it, that ought to be regarded as an extraordinary opportunity. The challenge for courts is to harness that opportunity without allowing the corresponding flood of poorly directed machine-generated material to make the administration of justice slower and less intelligible. Used in this way, artificial intelligence would not replace the judgment required to decide the case, but might allow a person who cannot afford sophisticated legal assistance to organise and present a case in a form in which that judgment can properly be exercised.

I suppose it might be going too far to extend this suggestion to the legally represented. I must confess, however, that having read some written submissions over the years, there have been occasions when the temptation would have been considerable.

In any event, this seems to me a use of artificial intelligence entirely consistent with the human-centred approach I have been describing because the machine does not decide the case or determine what justice requires, but brings order to the information and assists in making the real controversy intelligible, leaving the judgment where it belongs.

III Discovery

The third example is discovery, an area in which the contrast between the legal world of Sir Harry Gibbs and our own could scarcely be more pronounced.

The sheer number of documents is only part of the change. Commercial life itself has become documentary in a way that earlier generations could not have imagined. Every email, message, presentation, spreadsheet and successive electronic version of a document may be retained somewhere. The result is that the evidentiary universe in a major commercial case can consist of millions of documents before anyone has formed a sensible view about the few dozen that actually matter.

Our response to this abundance has largely been to improve the machinery by which documents are searched and reviewed. Search terms became more sophisticated and technology-assisted review has developed. Teams of solicitors working like drones in light-deprived offices became expert in managing electronic databases. Some of that, perhaps, represented progress, but we have still tended to conceive of discovery as an exercise in finding individual documents responsive to specified criteria rather than understanding the evidentiary corpus as a whole.

Artificial intelligence potentially changes this. One can envisage a system which goes far beyond identifying documents containing particular words by constructing objectively generated chronologies from millions of records, identifying communications between relevant actors, tracing the development of particular decisions, recognising documents which contradict an emerging chronology, identifying apparent gaps in a sequence and allowing lawyers and judges to interrogate the evidentiary corpus by reference to propositions rather than merely keywords.

The point can be illustrated by the limitations of conventional search methodology itself. In one recent discovery controversy in which I was involved, more than 132 combinations of searches had apparently been run against particular terms, while it was contended that much simpler searches would have identified directly relevant documents more effectively.³² More searching does not necessarily produce more understanding.

I think we can be considerably more ambitious than using artificial intelligence merely to perform the old task more quickly. In a document-heavy case, the universe of discovered documents could itself be analysed by an appropriately designed system which, by reference to the issues actually in dispute, produces a chronology for the use of the Court and identifies those documents which, on an objective analysis of the documentary record, appear material to the determination of particular disputed questions of fact.

The significance of such a process would be that the Court would begin not with competing chronologies reflecting the inevitable forensic choices of the parties, but with a common and independently generated account of the documentary record. None of this need encroach upon party autonomy. A party would remain free to adduce any admissible evidence upon which it wished to rely, to challenge the system’s analysis, to identify documents it considered had been wrongly omitted, or to contend that the chronology gave insufficient weight to particular material. Safeguards could be put in place to recognise and isolate privileged material.

But the objectively generated issues-based chronology could provide the starting point, with a party wishing to depart from it being required to explain why the Court should do so, or why additional documentary material was necessary to resolve the issue in question. Such a process would not determine what the evidence proves, which remains a matter for the judge, or prevent the parties from advancing the cases they wish to advance. It would instead provide the Court with a neutral map of the documentary terrain before the parties begin arguing about the route through it.

There is, of course, nothing novel in saying that discovery, or litigation generally, should be conducted efficiently and at a cost proportionate to what is really in dispute. We have embedded the overarching purpose in our procedural law, and we invoke it repeatedly in case management.³³ The difficulty is that the rhetoric has not matched the reality.

Saying things is considerably easier than changing the entrenched practices by which complex litigation is conducted. Anyone who has spent much time in the judicial system knows how resistant established practices can be to fundamental change. Altering the assumed procedural direction of modern litigation seems to require the turning circle of HMAS Melbourne.

I think we ought to be far more ambitious. A civil justice system in which an ordinary middle-class person, with a good claim and something important at stake, cannot sensibly afford to vindicate a legal right is, to that extent, a system which is failing. That was not the ordinary experience of civil justice in Sir Harry Gibbs’ professional lifetime, and we should not become so accustomed to present conditions that we regard them as inevitable.

Moreover, the problem is not confined to the private cost borne by litigants. Courts administer a finite public resource, and the allocation of judicial time has consequences. Every unnecessary day absorbed by mega-litigation, every week spent dealing with mountains of documents or elaborate processes which could have been avoided, is time unavailable to somebody else: perhaps a small business proprietor whose livelihood depends upon the prompt resolution of a dispute, perhaps parties whose modest resources are being consumed while they wait, or, more gravely, a person whose liberty is at stake.

I commenced my remarks by referring to the frequency of discussions concerning AI. Indeed, since completing what I thought was the final draft of this speech, there have been two very recent and useful contributions which suggest that thinking is moving in the direction I have advocated.

Last Monday evening, Sir Geoffrey Vos, the Master of the Rolls, delivered the John Lehane Memorial Lecture and urged lawyers and judges to embrace artificial intelligence “wholeheartedly”, including for document review, research, drafting, summaries and chronologies and assistance to litigants in person.³⁴ His approach is characteristically ambitious, and I agree with much of it. Where I am less persuaded is his suggestion that fully autonomous machine-made judicial decisions might be acceptable with the informed consent of the parties. Consent does not, to my mind, quite reach the heart of the problem. If judgment is the human and evaluative activity I have been describing, there are aspects of adjudication by an arm of government which must remain human whether the parties are content to surrender them or not.

The second was the Federal Court’s seminar upon the launch of the AI Practice Note last Wednesday. What I found encouraging about that discussion was the recognition that the present regulatory response is only a beginning, and particularly the emergence of quite practical thinking about implementation. One development discussed was that the Victorian Bar is exploring a centrally procured portfolio of enterprise-grade artificial intelligence platforms which would give barristers access to commercial models within arrangements addressing security, confidentiality and privacy.³⁵

One sees unmistakable signs of an increasing realisation that AI gives us a fundamental opportunity to rethink processes which have made litigation prohibitively expensive and unnecessarily protracted. I hope this realisation is accompanied by the imagination, and the will, to take it.

And there is an important connexion here with Sir Harry Gibbs himself. As foundation Chairman of the Australian Institute of Judicial Administration, Gibbs was involved in the discussions from which its first research project emerged, an empirical examination of delay and efficiency in civil litigation.³⁶ The 1985 report found that an average personal injury case could take about four years to reach disposition, notwithstanding that the great majority of matters consumed no hearing day at all.³⁷ Around the same period, the High Court moved from almost exclusive reliance upon oral argument to requiring outlines of principal points, a development Justice Kirby later recalled Gibbs regarded as an efficient means of identifying what counsel were really saying.³⁸

There is a temptation to regard the forms of procedure with which we are familiar as though they possessed some intrinsic virtue, when their virtue lies only in whether they assist the just resolution of disputes. Skeleton arguments, electronic discovery, referees, case management and electronic court books were all innovations once. The relevant question is never whether the procedure resembles the procedure with which we grew up. It is whether it enables the Court to identify the real controversy and quell it according to law at a cost which bears some rational relationship to what is at stake.

AI therefore presents an opportunity which is considerably more interesting than whether a machine can draft a competent submission or find a case more quickly. It allows us to reconsider now entrenched processes which developed in a world constrained by the capacities of human beings to collate and summarise information. Those constraints are disappearing fast, and we should not assume that the procedures built around them must remain.

But this takes me back to where I started. I want machines to do more of the work which does not require judgment precisely so that human beings can spend more time doing the work which does. The danger arises if, having discovered how extraordinarily capable the machine is at the former, we gradually allow it to assume the latter. Judgment is something more than a sophisticated form of information processing because it is embedded in human experience and informed by values which the machine did not create and cannot ultimately justify.

It is in that sense that Pope Leo’s insistence upon beginning with the human person rather than the technology seems to me of direct importance to lawyers. The more powerful the instrument becomes, and the more effectively we use it to relieve judges and lawyers of work which does not require judgment, the more important it becomes to understand what judgment is, why it must remain human, and the purposes and values it exists to serve.

In short, our lodestar should be using AI to implement reforms allowing lawyers to spend more time exercising professional judgment, and judges to spend more time doing what Sir Harry did with such distinction: identifying what really matters and deciding it according to law.

Perhaps that is also the answer to Eric Morecambe. The future of the common law will not depend upon our ability to accumulate still more notes. We have more than enough of them already. It will depend upon whether we retain the human judgment necessary to hear the music.

That, I think, would be a use of artificial intelligence entirely worthy of the tradition exemplified by Sir Harry Gibbs.