Official website links end with .gov.sg
A Singapore Government Agency Website
Secure websites use HTTPS
Justice Kwek Mean Luck: Keynote speech at the Judges’ Colloquium
JUDGES’ COLLOQUIUM
Pillars of a Modern Judiciary
Mombasa, Kenya, 24 August 2026
The Honourable Justice Kwek Mean Luck
Judge of the High Court
Supreme Court of Singapore
I. Introduction
The Honourable the Chief Justice and President of the Supreme Court of Kenya, Chief Justice Koome; Honourable Justices of the Supreme Court and the Court of Appeal; Honourable Judges of the High Court and Courts of Equal Status; distinguished guests, ladies and gentlemen.
1. Good morning. I am truly honored by this invitation. I have been asked to speak about judicial independence and accountability, about effective case management, and about how the Singapore courts are organised, including our approach to artificial intelligence. But I come to learn from you, as much as to share. Before I do so, I want to first set the limits of what I can usefully offer, because a speech of this kind is only worth hearing if the speaker is honest about where his experience runs out.
II. Two jurisdictions that are not alike
2. Singapore is a city-state of about six million people on some 730 square kilometres, most of it dense urban area. We have a single unified court system in one jurisdiction, under one body of law, with English as our working language. No litigant travels more than an hour and a half to reach a courthouse. When we change a judicial practice, the entire judiciary of Singapore can easily be briefed in one afternoon.
3. Kenya faces a different order of undertaking. Your judiciary serves a population approaching sixty million across a territory some eight hundred times the size of mine, through more than a hundred courts, in a country of forty-seven counties and many languages.
4. Your docket also differs in kind, not merely in degree. Land disputes, devolution conflicts, electoral petitions and constitutional public-interest litigation appear to be features of the litigation load in Kenya. Our courts developed, adjudicating mainly commercial and criminal matters. What I share this morning arises out of Singapore’s particular context.
5. There is a second point, and it matters. I am conscious that I am speaking to a judiciary that has already achieved a high level of efficiency and output. In the last year, based on the public data, the Kenyan courts resolved around 647,000 matters against around 621,000 filed — a clearance rate of 104 per cent — and backlog fell by 27 per cent, across a hundred and sixteen courts. Given this high level of efficiency, I hope that what I share will still be of some value to you.
6. This morning, I will speak about four pillars supporting the effectiveness of our judiciary. First, independence and accountability. Second, active case management. Third, technology. Fourth, people.
III. The first pillar: independence and accountability
7. Independence and accountability are conventionally presented as being in tension, as though every increment of one was purchased at the cost of the other. In my view, that framing is mistaken. Judicial independence is not a privilege
conferred on judges. It is a guarantee given to litigants — that a case will be decided by a judge, free to decide it against the government, the powerful, popular sentiment, and even against the judge’s own interest. It exists for the litigant’s benefit.
8. Understood that way, accountability does not exist as a diametrical opposite to judicial independence. It is an underlying and necessary condition. A judiciary that cannot be held to account for how it decides, how long it takes and how it conducts itself, will not retain the public confidence, on which its independence ultimately rests. Independence unaccompanied by accountability, is not going to be independence for very long.
A. Two roles of a judge
9. Such accountability is underscored by the fact that in our view, judges need to be accountable in two ways. This is a point made by our Chief Justice Sundaresh Menon.
10. The first is in the discharge of their adjudicative role, with which we are all familiar. The second is the discharge of their systemic justice role: that is, the task of developing and operating a system for the administration of justice, that is accessible to all and meets the needs of all. It is by discharging both, that the courts secure public trust in the justice system and, more broadly, in the rule of law.
11. It is in undertaking this second role, that of systemic justice, that we as judges embrace the work that ensures the accountability of our judicial system as a whole.
12. The systemic justice role is easy to overlook, because in most large service organisations, the senior people are not involved in delivering the service. Judiciaries are unusual in this respect. As judges, we see the individual case and the system at the same time, and we work directly with the Bar, the prosecution and self-represented persons. That puts us in a better position than anyone, to see where the system can be improved and to forge the common understanding needed to change it. It is also in our professional interest: the effectiveness of the system determines our capacity to discharge the adjudicative function well.
13. Two things follow.
14. The first is that if judges do not take up this work, someone less well placed will attempt it. There is another common law jurisdiction, where justice reforms were for a period driven by policy officers rather than judges. I was told by judges there and also observed it as an outsider, that because the policy officers did not sufficiently understand the system, their reform proposals gained no traction with the judiciary. This meant that the problems faced by the judiciary remained unsolved.
15. The second is that the systemic role is itself a form of accountability. A judiciary that takes responsibility for the state of its own system — for how long matters take, for whether the ordinary citizen have access to it — has accepted an obligation that is fundamental to the delivery of justice.
B. Some observations about accountability
16. At this point, I make three observations about accountability.
17. First, the accountability of the judicial institution is different from the accountability of the individual judge. As a judicial institution, we publish caseload statistics, monitor disposal timelines against benchmarks, and report internally each month. These establish our courts’ measurement of itself.
18. The second is that efficiency and trust are related: a more efficient judicial system enhances trust in the judicial system. However, they are different properties and can move in opposite directions. A judicial system can improve on essentially every efficiency metric while public trust declines, because clearance rates measure the efficiency of the system and its processes, while trust measures encompass a wider concept, namely whether members of the public believe they will be treated fairly.
19. Third, a judiciary that holds itself to account must build the capability to deliver judicial outcomes within a defined period. A process that can be suspended indefinitely does not do the work that accountability requires. This in turn requires active case management, and that is the subject to which I now turn.
IV. The second pillar: active case management
20. Let me begin on this, by sharing with you what our courts looked like in the 1990s. Trial dates were available only three years out, often longer. We carried more than ten thousand inactive cases, some filed over a decade earlier. Criminal cases took up to four years, appeals about two. Almost half of all matters took between five and ten years.
21. There were two broad reasons for this: first, a lack of discipline among the lawyers, and second, a lack of intervention by the courts. I dwell on the second, because it was within our own control, and because all things considered, the first cannot really be said to be a cause at all. Lawyers conduct themselves as the judicial system before them permits. If the court will invariably grant the adjournment, the adjournment will be sought. The consequences, were access to justice impeded and public confidence lowered — here, the trust problem followed the access and efficiency problem.
22. It took two to three years before objective indicia indicated that the tide had changed. By the late 1990s, some 95 per cent of civil cases were resolved within a year.
A. How the backlog was addressed
23. The strategy that addressed the backlog was this: control over the pace of litigation moved from the lawyers and the parties to the court. A case is not the private property of the parties, to be advanced or stalled at their convenience. It occupies a public resource — a judge, a courtroom, a place in a queue ahead of somebody else’s dispute. The court’s interest in its progress is a public interest.
24. This strategy found expression in various case management measures that we have adopted. These measures can be considered in four broad categories. Let me take each in turn.
B. Diversionary measures
25. First, diversionary measures. These take matters out of the litigation stream altogether. In Singapore, this has become considerably more than a docket-clearing technique. We began institutionally in 1994, with a Primary Dispute Resolution Centre, bringing judge-led settlement conferences into the ordinary life of a case. In 1997, the Singapore Mediation Centre was established under the auspices of the Singapore Academy of Law, offering commercial mediation
outside the courts. Community Mediation Centres followed in 1998 for neighbourhood disputes. Alongside these, the Singapore courts developed pre-action protocols in areas generating high volumes of standard disputes — non-injury motor accidents and medical negligence — requiring parties to exchange positions before proceedings begin, and resolving many matters before they ever become a court file.
26. These steps built up towards our Rules of Court 2021, which impose an express duty on every party to consider an amicable resolution of the dispute, before commencing proceedings. The court may order parties to attempt amicable resolution even where one is unwilling, and an unreasonable refusal may be reflected in costs.
27. I draw this out because of what it represents. Mediation in Singapore began as something offered to parties who might want it. It is now a duty owed to the court, on the footing that a party who insists on adjudicating what could have been settled, is consuming a public resource without justification. It applies the same core concept — that the pace and manner of litigation are the court’s business — one step earlier, before the case begins.
28. A dedicated forum can be diversionary in the same way. Our Small Claims Tribunals have operated since 1985, designed to be used without a lawyer, and they keep a large volume of low-value disputes out of the general civil list.
C. Dispositive measures
29. Second, dispositive measures. These deal with matters that have stalled. There are three types of dispositive measures that are common in Singapore:
a. Auto-discontinuance clears dormant cases after a defined period without requiring a judge or the other party to act on each one.
b. Peremptory or “unless” orders attach a defined consequence to a defined default, so that the next failure to comply disposes of the point rather than generating a further application.
c. Costs orders, used consistently rather than occasionally, change behaviour because they change the economics of delay.
30. What made these measures work, was that they were applied consistently and predictably, so that the legal profession could see that a default would in fact have the stated consequence.
D. Facilitative measures
31. Third, facilitative measures. These help a live matter move. This can come through the expansion of hearing capacity as well as through improved processes.
32. To address the backlog, in April 1992 the Singapore judiciary increased its capacity by opening Night Courts, with sittings in the evening to deal with the very large body of regulatory and traffic matters. They served two purposes at once: they took high-volume, low-complexity work out of the ordinary lists, freeing daytime capacity for matters that needed it more, and they spared a person charged with a minor offence the loss of a day’s wages to attend court. Thirty years on, our Night Courts are still sitting.
33. There have also been several facilitative measures introduced through our court procedures.
34. One example is the use of pre-trial conferences as the default, rather than the exception. Some pre-trial conferences are conducted by the Judges who ultimately hear the trials. I find such Judge Case Conferences to be very helpful. It is held after affidavits of evidence-in-chief have been exchanged, sometimes from six weeks to a few months before trial. It allows the judge to crystallize issues, narrow down the scope of actual dispute, or indicate preliminary concerns, and force the parties to take a systematic and disciplined approach to trial preparation.
35. We also introduced trial date certainty coupled with hearing fees, so that a date fixed is a date kept, and the cost of a vacated date falls where it belongs. Another example, is our shift towards written advocacy, to reduce time taken up in court unnecessarily, and to ensure that time that is spent in court is time spent well.
E. Monitoring and control measures
36. The fourth category, monitoring and control measures, tells us whether the other three categories of measures are working. We set three main benchmarks: the clearance rate; disposal timelines on a cohort basis, at 85 per cent within eighteen months; and trial date availability, targeted at eight weeks between set-down and trial. These targets are not just internal. They have been publicly committed to by our Judiciary. This in turn helps to ensure accountability. The data from these benchmarks provide historical and institutional-level reporting. But they do not obviate the need for proactive monitoring at the case level, which remains the responsibility of individual judges and judicial officers.
F. Two lessons from active case management
37. There are two lessons arising from our experiences with case management, that I thought to share. The first is the importance of institutionalisation. The measures introduced from 1992 were not a clearance campaign mounted only in response to the demands of that time; instead, they brought about a permanent change to how litigation is conducted. The backlog did not return, because the conditions that generated it — party-driven timetables, routine adjournment, procedural steps serving no function — were removed from the system rather than temporarily suspended.
38. The second lesson is that a system that over-relies on active case management and process improvements, may achieve efficiency at the expense of its own sustainability in the longer term. Comparative studies find Singapore posting high clearance rates while ranking among the jurisdictions with the fewest judges per capita. The conclusion usually drawn is that procedure outperforms headcount. There is some element of truth, in relation to the value of procedure, but that can be overstated. It does not mean that vacancies do not matter: a judiciary sitting at a fraction of its needed complement is a serious constraint. No procedural ingenuity can substitute for judges who are not there. In addition, a demonstration that a judiciary achieves efficiency on a constrained budget can be misconstrued by government finance colleagues as evidence that the budget is thus adequate. It would be important to continue engagement with the budget side of the house, to ensure that momentum built up is not stalled, and instead continues to be both sustained and sustainable.
G. Balancing risks from efficiency
39. I would not be giving an honest account if I described these reforms that Singapore adopted, without mentioning the risks they involved. Robust case management measures may transfer risk onto practitioners. For example, in criminal matters, there is the risk of an accused person facing trial in court, whose counsel is not ready to mount a considered defence on his behalf. Judges need to be alive to such risks and manage each case accordingly, bearing in mind that ultimately, case management exists to facilitate access to justice, and our ultimate imperative is to ensure that justice – in its various facets – is done in each case.
Therapeutic justice in the Family Justice Court
40. This in turn raises a further question: are there scenarios where traditional adjudication may not be the best way to achieve justice in the dispute at hand? The common law trial is a remarkable machine, but it has a specific design. It is adversarial: each side builds the strongest case it can and tests the other’s. It is retrospective: it determines rights and liabilities arising out of past events. And it is binary: somebody wins. These features suit a commercial dispute between parties who will part ways whatever the outcome. But they suit poorly a dispute between people with a relationship that they cannot totally leave behind.
41. Nowhere is this clearer than in family proceedings. We may have experienced family cases that are oozing with venom. A system that pushes parties in family disputes to lawyer up in the aftermath of a marriage breakdown is counter-intuitive and counter-productive. The adversarial process asks each parent to assemble the worst available account of the other, then returns them, having done so, to the task of raising a child together. The process itself damages what the family will most need afterwards – that is, the ability to heal, to try to repair what has broken between them, and to move on from the troubles of their past to a brighter future.
42. Singapore’s response has been to change the judicial approach. Over the last decade, our Family Justice Courts have moved from a traditionally adversarial system to one that is more judge-led and semi-inquisitorial. In October 2024, the Family Justice Courts launched the Therapeutic Justice Model. Its premise is that the object of family proceedings, is not to determine a winner but to help a family reach a workable future, with the welfare of the children as the organising concern. It brings judges, court family specialists, counsel and social agencies into a single multi-disciplinary approach, and asks the parties to take ownership of the outcome. In July 2025, it was extended to our Youth Courts, where the court works with schools and agencies on a young person’s reintegration into society, rather than confining itself to disposing of the charge.
43. I raise this in a discussion about court efficiency quite deliberately. A family matter resolved therapeutically usually consumes less court time than one fought to judgment, so there is an efficiency dividend. But that is not the reason for doing it. The core reason is that for certain classes of disputes, the adversarial trial can produce a formally correct answer, but at a cost to the parties that the formal answer does not justify. It is worth considering, in assessing our judicial docket, whether adjudication in its traditional or existing form, is what each category of case actually needs, in the wider interests of justice.
V. The third pillar: technology
44. Let me turn now to a third pillar in the efficiency of Singapore’s judicial system, that of technology. Our technology journey has run close to thirty years across three Chief Justices, in four phases.
45. The first, from 1997, was electronic filing: Chief Justice Yong, having cleared the backlog, took the view that to stop it rebuilding, it was also important to digitise.
46. The second phase, built case management portals on that foundation.
47. The third phase extended those systems to people without legal representation.
48. The fourth, from 2020 onwards, began with video hearings and now runs to the consideration of how artificial intelligence may be used in our court systems.
49. Let me share three broad lessons that we imbibed from this period of technology driven reform.
50. First, the importance of judicial leadership. I served as an Assistant Registrar during the first phase, and one memory has stayed with me vividly. I was demonstrating our electronic filing system to a very senior judge of another country. He commented that the concept of electronic filing, was not revolutionary — they had tried the same thing at home. However, it had not materialised despite many discussions. The difference, he said, based on his observations, was the political will of Chief Justice Yong, and his willingness to proceed despite reluctance from the Bar. I share this not to make a point about ignoring the sentiments of the Bar, whom we regard as important stakeholders and partners in the justice system, but about the critical role that judicial leadership plays in bringing about reform – even, and indeed especially, when the immediate effects of such reform may be unpopular.
51. Second, it is important to re-engineer the business process before we digitise it, to put the operational and court users in the same room from the start, alongside the technical people. This ensures that we develop solutions that work for the people that will use them.
52. Third, continuity. Our digitisation journey took thirty years across three Chief Justices. Successful reform of this kind must necessarily outlive the reformers, which means embedding it in rules, budgets, training and the culture and expectations of the judiciary and legal profession.
A. Technology for litigants
53. Through the different phases, we came to appreciate that systems facing the judicial institution, and systems facing the litigant, are very different things. Electronic filing digitises what the court registry is already doing on paper. It is valuable. It is a prerequisite. But it does not bring into the digital system, a person who is not currently using it. There is also a need to design a system that helps litigants.
54. Our Community Justice and Tribunals System is built on that premise. It covers our small claims and other disputes on one platform. It assumes that the user has no lawyer to represent her. This digital system offers a pre-filing assessment, so a person can find out whether they have a claim – and, if so, whether they have what they need to pursue that claim – before embarking on court proceedings. It allows filing at any hour and lets parties settle online without attending court. We built it on national digital infrastructure, linking court notices to the digital applications Singaporeans already use and trust, so that it is more convenient for them.
55. One caution: extending technological platforms to litigants is important, but digital-first design also risks excluding the digitally illiterate. A digital program, for all its ambitions, invariably works well only if paired with assisted filing at the courts and the retention of some non-digital pathways. Our efforts to design around and for litigants must bear this in mind.
B. Artificial intelligence
56. Let me turn next to artificial intelligence, on which I want to be plain about the limits, the promise and our approach.
57. Large language models operate by matching semantics. They do not, at present, have logical sequencing capabilities. They do not reason through a problem in ordered steps in the way a lawyer or judge would recognise. That limits how far AI can assist us in decision-making. It is why the existing AI legal systems are strong in certain areas — summarisation, review, classification — which are well suited for semantic matching capabilities, and less strong in prediction and judgement, which require some level of logical reasoning.
58. But that may change in time. I had a conversation late last year with a senior AI engineer at one of the leading AI companies in the world. I was talking about the pursuit of the holy grail: combining semantic capability with logical sequencing. This engineer told me that they are feeding their AI model terabytes of data every day, and that the answers coming back appear closer and closer to simulating human thinking, even on a semantic matching basis.
59. This made me consider, that the day may come, where it appears that generative AI is engaged in logical thinking, even when its underlying model is that of semantic matching. This led me to look further into the area of cognitive science.
60. It appears that our thinking process is to some extent at least, a product of semantic matching. We learnt the alphabet by pattern — A, B, C. We learnt that one plus one equals two in much the same way, long before understanding why it is so. If our own reasoning capacity is built on a semantic foundation, an AI machine could conceivably exercise what appears to us as logical sequencing capabilities, using the semantic matching model it now uses, even in complex areas of logical analysis. Add on to this, the wider realms of possibilities that agentic tools can introduce, through the coding of specific work methodologies, and the possibilities are there for imagination.
61. As an illustration of the sheer pace of technological change in the area of AI, take hallucination. Two or three years ago it was a key objection to the use of AI tools in legal work. However, today, for large language models that are developed for and targeted at legal research, hallucination has largely become yesterday’s problem.
62. In Singapore, I chair LawNet Technology Services, the company that runs LawNet, the online legal research tool for Singapore law, which is used by our lawyers and judges. By restricting the source materials to a curated set, engaging in an iterative training process of the AI outputs with the help of legal professionals, and incorporating citations in the generative AI replies, we have narrowed down the risk of inaccuracy to a very low percentage in LawNet.
63. All of this raises the question of where the line should be drawn. How far should judges rely on AI tools? At what point have we devolved our judicial responsibility to that of the AI machine? Fundamentally, judging is our responsibility. It cannot be devolved to artificial intelligence. Judicial reasoning is not merely a means to an outcome; it is a substantial part of what makes the outcome legitimate — it is how the losing party is shown that they were heard and understood, and how the public is shown that the decision was the carefully considered product of a process rather than a preference. There is a risk that judicial decision-making becomes a black box, opaque and therefore unaccountable. A judge cannot be accountable for a decision he cannot fully explain, and he cannot fully explain a decision he did not fully make. But as these systems become more capable and more prevalent, the line between reasoning devolved to AI and reasoning assisted by AI will become progressively harder to draw.
64. In the Singapore judiciary, we take three tracks in managing this.
65. First, cognizant of these risks, we are adopting a textured approach towards AI adoption in Singapore. We have started with summaries of court documents in our Small Claims Tribunals, where users are overwhelmingly self-represented. We are also experimenting with and studying AI developments, either by engaging vendors in pilot projects and building certain capabilities internally.
66. Second, we have introduced guidelines to clarify the boundaries of acceptable conduct viz AI. Externally we have published a Registrar’s Circular, neutral on whether court users may use generative tools but explicit that they remain responsible for what they file. Internally we have published guidelines naming, and emphasizing the need to guard against four risks: human reason risk, confidentiality risk, hallucination risk and accountability risk.
67. The third track, is in my view, the most important. Guidelines can only do so much. Ultimately, this comes back to the values and the integrity of the judges we appoint, and whom we continue to develop. A judge who understands that the reasoning is an integral part of his judgment, and who feels a sense of responsibility in exercising that judgment, will not surrender it, whatever the capabilities of the AI tool the desk is offering. A judge who does not, will find the tool very obliging. Hence, the selection and appointment process for judges, and ongoing judicial training, becomes central to how a judiciary manages artificial intelligence well.
68. Let me end this brief survey of the technology pillar, with an eye on the future. We cannot avoid the tide of AI. It is here and now. It will change behaviour, whether or not we plan for it. There are two risks I would particularly watch for.
69. The first is undetected algorithmic bias in the AI models providing legal assistance. If an AI system carries a tilt we cannot see, and we come to rely on it heavily over time, it could gradually modify our understanding of legal principles or societal values, without our realising that this has happened. That risk is very modest today, but it can grow as reliance grows. And the unseen incremental nature of such a change over time is precisely what makes it dangerous.
70. The second is skills atrophy in the generation of lawyers and judges coming through, who for example, may never have to construct an argument from first principles because AI has always produced the first draft. Legal and judicial craft is learnt by doing it, perhaps badly at first, and improving through working on it repeatedly, and learning through reflection and intention.
71. Both risks point the same direction. They call for some degree of friction — deliberately designed into the systems we introduce. This can be a nuanced exercise. They also call for us to reshape the way we train lawyers and judges, so that our judges learn to adapt to and use AI, without losing our core skills.
VI. The fourth pillar: people
72. This leads into the fourth pillar that I wanted to speak on. Our people. In the course of traversing the other three pillars, I have also spoken on the importance of this. In the Singapore Judicial College, we invest not only in adjudicative training but in equipping judges for the systemic role, which does not come naturally, as judges are first and foremost assessed in terms of adjudicative suitability. The College also invests in developing our judges as people, by prompting reflection on judicial values and judicial philosophy.
73. People matter — and not only judges. Our court administrators are also important to the success of any judicial reform. There is a lesson I took from my time in the administrative service, when I was dealing with healthcare policy. There is a natural tendency for governments to design increasingly sophisticated financing schemes to meet different needs without falling into fiscal deficits. This makes healthcare financing more and more complex. However, what a patient actually experiences of that policy, is not the policy in its nuanced complexity. It is whatever the nurse at the counter understands and can reasonably explain.
74. Whoever delivers is the face of our system. In the judicial system, these are judges, but they are also court administrators. The last mile matters, and so do the last man and woman in our chain of judicial administration.
VII. Conclusion
75. Let me close where I began. What I have shared, grew out of Singapore’s own particular conditions.
76. At the same time, what travels across jurisdictions, is the role that judges can play. Judges are unusual in being able to see both the individual case and the system. We sit in the courtroom and we can examine and drive judicial reform. We know where processes can be improved because we not only know about these processes in the abstract, from a distance. That is why the systemic justice role is not an optional addition to judicial office, but part of what judicial office requires, if our judicial system is to be accountable.
77. There is an African proverb that I am told is well known in this part of the world: alone, a youth runs fast; with an elder, slow; but together, they go far. For us in Singapore, it is going together, as a judicial community of judges and court administrators, that has allowed the judicial system to go as far as it has.
78. This morning, I would also value hearing from you and learning from your judiciary, so that all of us together can go further, than where we are now.
79. Chief Justice, Honourable Judges, thank you for your kindness in hearing me. Asante sana.
2026/08/28
Share this page:
