sg-crest A Singapore Government Agency Website
Official website links end with
Secure websites use HTTPS
Look for a lock () or https:// as an added precaution. Share sensitive information only on official, secure websites.
Note: To search for hearing details for a specific case, visit the hearing list page.

Chief Justice Sundaresh Menon: Keynote address delivered at the Sentencing Conference 2022


Sentencing Discretion: The Past, Present and Future

Key Messages 

1. Sentencing is a core judicial function in criminal justice. In discharging that function, the sentencing judge must strive to strike the right balance between imposing a condign sentence in the case at hand, while maintaining a broad level of consistency with other cases so that sentencing at a systemic level is fair. Sentencing guidelines and frameworks are best understood as means to enable the sentencing court to strike that sometimes elusive balance.

2. The Sentencing Advisory Panel, which was established in 2022 to issue sentencing guidelines, will come to play a very significant role in Singapore’s criminal justice system, by promoting consistency and transparency in sentencing while enhancing public awareness of sentencing.

3.     It is unlikely that Singapore will adopt artificial intelligence (AI) in criminal cases within the foreseeable future. However, the use of AI in criminal cases remains an intensely important subject that should be followed closely.


Distinguished guests,
Ladies and gentlemen,

I. Introduction

1. Good morning. Let me begin by thanking the organisers for putting this event together and inviting me to speak to you today. Since the last edition of this Conference in 2017, the sentencing landscape in Singapore has continued to develop rapidly and extensively. The courts have handed down many more sentencing frameworks to guide the exercise of sentencing discretion in a broadly consistent way. And in June this year, we saw the formation of the Sentencing Advisory Panel (or “SAP”) to formulate and publish sentencing guidelines. This Conference provides a timely opportunity for us to reflect on recent developments, so that we can look to what might lie ahead of us in the context of sentencing in Singapore.

2. My principal thesis today can be summarised in this way. Sentencing is now widely accepted as a core judicial function in criminal justice. In discharging that function, the sentencing judge must strive to strike the right balance between meting out a condign sentence in the case at hand, having adequately accounted for the particular facts and circumstances, while also maintaining a broad level of consistency with other cases so that sentencing at a systemic level is seen to be fair. The bespoke nature of sentencing is important because it aids in the ultimate goal of securing the reform of the particular offender before the court. But this must also be seen to be fair to offenders as a whole. The various tools we have such as sentencing guidelines and frameworks, and now the SAP, are all best understood as means to get us to the desired end of striking that sometimes elusive balance. This, in a nutshell, is the thesis that I will develop today and to that end, I will divide my remarks into three main parts.

(a) First, I will discuss the judicial role in sentencing. I will trace the evolution of the courts’ sentencing role and juxtapose this against the roles played by other bodies involved in the sentencing process.
(b) Next, I will turn to the exercise of sentencing discretion today. I will touch on the sentencing objectives and the end goal of sentencing, before focusing on sentencing frameworks. In this context, I will examine the importance of these frameworks in achieving some of our key objectives.
(c) Finally, I will briefly consider the future of sentencing, by touching on the SAP and the possible use of artificial intelligence (or “AI”) in sentencing.

II. Part 1: The judicial role in sentencing

    A. The rise of sentencing discretion

3. Let me begin with sentencing discretion. Although the judicial role in sentencing is very widely accepted today, it might be forgotten that this is a relatively recent phenomenon in the history of the common law. Traditionally, sentencing was not typically regarded as a facet of judicial power.[1] In most common law jurisdictions, the judicial function in criminal law was confined to finding the facts and applying the law.[2] There was no need to separately consider sentencing because historically, the courts generally had no sentencing discretion to speak of: all felonies were punishable with death.[3]

4. In the early 18th century, legislatures began to confer sentencing discretion on the courts. In England for example, Parliament first empowered judges to sentence offenders to transportation (generally for a fixed period) instead of to death.[4] However, this only applied to a small class of cases and so, the notion of judicial discretion in sentencing initially remained very limited.[5] Subsequently, in the period from about 1820, the UK Parliament started to pass legislation prescribing maximum punishments for various offences, with the court empowered to determine the length of individual sentences.[6] This has been described as “the beginning of the modern system of judicial sentencing discretion”,[7] and this was adopted in the Indian Penal Code 1860, which was later incorporated into Singapore law with the passage of the Penal Code 1871 (or “Penal Code”).

5. Over the years, the sentencing discretion of our courts has continued to grow. Two broad themes have underlain this trend: the rise of alternative sentences and the reduction of offences carrying mandatory sentences.

6. The rise of alternative sentences began in the aftermath of the Second World War, when probation was introduced for youths in 1949, and this was to enable young offenders to be dealt with in a more progressive way so that they might better be able to contribute to the post-war economy.[8] Shortly thereafter, in 1951, probation was extended to adults. Corrective training and preventive detention followed as sentencing options for habitual offenders in 1954. Then came reformative training in 1956, which was designed to equip adolescent offenders with useful skills to enable them to break out of the cycle of offending behaviour.[9]

7. The range of sentencing options remained largely unchanged for more than 50 years after that, until the passage of the Criminal Procedure Code 2010. This was a landmark piece of legislation, which among other things, introduced the range of community sentences with which we are all now familiar, such as mandatory treatment orders (or “MTOs”), day reporting orders, community service orders and short detention orders. More recently in 2018, the scope of the community sentencing regime was expanded. Among other changes, the regime now applies to persons who have previously been sentenced to reformative training or even to imprisonment terms not exceeding three months, while MTOs may now be made in respect of certain offences punishable with up to seven years’ imprisonment (up from three years’ imprisonment previously).[10]

8. The second broad theme underlying the expansion of sentencing discretion is the removal of certain mandatory sentences. Historically, mandatory sentences were somewhat rare in our criminal law. This changed in the 1970s, when Parliament introduced mandatory minimum sentences for various drug and arms offences with specific deterrence objectives.[11] Subsequently, in 1984, mandatory minimum sentences were introduced for a range of other offences such as housebreaking, robbery, snatch theft, and aggravated forms of outrage of modesty and rape.[12]

9. But a shift away from mandatory sentences began in 2007, when the Penal Code was amended, with a “key objective” of the amendments being to confer greater sentencing discretion on the courts.[13] Mandatory minimum sentences for several offences were removed, and five years later, in 2012, Parliament abolished the mandatory death penalty for certain classes of drug importation and trafficking offences and of murder. In explaining these changes, the Minister for Law stated that where possible, Parliament should “move towards giving greater discretion to the courts” and that mandatory sentences “are and should be the exception”.[14] This philosophy was again reflected in the introduction of the presumptive minimum sentence for certain immigration and regulatory offences in 2019, which empowers the courts to impose a sentence lower than the presumptive minimum sentence if certain conditions are met.[15]

10. In sum, the rise of alternative sentences and the removal of mandatory sentences have expanded both the range of sentencing options available to the courts, and the degree of flexibility afforded to the courts in calibrating the appropriate sentence for each offender. The sentencing discretion of the courts has thus increased considerably, and its exercise has now become central to the judicial role in criminal cases.

    B. The courts' role in sentencing

11. But what exactly is the judicial role in sentencing, and how does it interact with the roles of the other arms of the State in the sentencing process? The broad position is well-established. Parliament prescribes the sentencing scheme for offences in legislation. This reflects the sentencing policy and the public interest that Parliament seeks to achieve. The courts determine the appropriate sentence for the individual offender based on the facts of each case, in the light of the legislative scheme. And the Executive then carries out the sentences imposed by the courts.

12. Some nuances in this basic picture have been clarified in the cases, which establish two principles. First, it is the exclusive preserve of the judiciary to determine and impose the appropriate sentence for the individual offender. While the point has not arisen here, persuasive foreign authorities have struck down legislation that sought to allow the Executive to select, directly or indirectly, the sentence for a particular offender, on the basis that this impermissibly usurped the courts’ sentencing power.[16] By comparison, provisions of the Misuse of Drugs Act 1973 (or “MDA”) under which decisions of the Attorney-General or the Executive may have a bearing upon an offender’s sentence, but which do not empower those bodies to choose that sentence, have been held not to encroach on the judicial role and have thus been upheld.[17]

13. The second principle is that although Parliament may determine the scope of the courts’ sentencing power through the legislative scheme that it prescribes, including by specifying mandatory minimum or maximum sentences, the sentencing function of the courts cannot be enlisted to disguise what in substance is a decision of the Executive. On this basis, the High Court of Australia has struck down legislation that required the court to impose a control order on a person, without a judicial determination of guilt, upon a finding by the Executive that the person belonged to an organisation declared to be a risk to public safety. It was held that such legislation undermined judicial independence and integrity, by giving the colour of a judicial decision to what, in essence, was a decision of the Executive.[18] Such an objection has been held not to apply to the alternative sentencing regime under the MDA, which only applies after the courts have independently determined the guilt of the offender.[19]

14. The judicial role in sentencing has further evolved in other countries such as England and the United States, with the formation of sentencing councils and panels that generate guidelines to assist the courts in sentencing offenders. In these jurisdictions, the core of the courts’ sentencing function – which is to determine the appropriate sentence for individual offenders – remains unchanged. The courts are not bound to apply the guidelines in all cases,[20] and in any event, the guidelines do not dictate the sentence for any individual offender. But they play a significant role in structuring the courts’ exercise of their sentencing discretion, thus promoting broad consistency in sentences. This, too, is the central function of the sentencing frameworks that have been laid down in Singapore, though these have thus far exclusively been developed and set out in judicial decisions.

III. Part 2: The exercise of sentencing discretion - the present position

15. I turn to the second part of my address, which will focus on sentencing frameworks. I want to begin by recalling the objectives and the end goal of sentencing. These are central to the sentencing process, for a fundamental issue for every sentencing court is what goal it seeks to achieve by punishing the offender.[21]

    A. The objectives and end goal of sentencing

16. The four sentencing objectives are well-known: deterrence, retribution, rehabilitation and prevention. Deterrence, rehabilitation and prevention are outcome-driven sentencing considerations, which focus on achieving beneficial consequences for society usually while bringing about the reformation of the offender. By contrast, retribution requires that the punishment imposed on an offender be proportionate in some way to the harm caused and to the offender’s culpability. In this way, retribution often acts as a counterweight against the more goal-driven sentencing considerations.[22] The importance and interplay of these sentencing objectives vary from case to case, depending on the nature of the offence and offender-specific factors.[23]

17. Notably, there is an important link between the sentencing objectives and the sentencing frameworks. Our courts have laid down two main types of sentencing frameworks: offence-specific frameworks for individual offences, and offender-specific frameworks such as those relating to young offenders[24] and repeat offenders for whom corrective training may be appropriate.[25] Generally, the court should first assess the dominant sentencing objective(s) before deciding what type of framework to apply. An offence-specific framework will commonly be calibrated by reference to the prescribed punishment for the relevant offence for a typical offender, and this will generally be an imprisonment term or a fine or, in some circumstances, both. But the dominant sentencing objective(s) may render it more suitable for the court to impose a different type of sentence instead. For example, where rehabilitation is the dominant sentencing objective because the offender is young, the appropriate sentence may be probation, reformative training or a community sentence. Alternatively, where prevention is a key sentencing goal because the offender is a repeat offender, corrective training or preventive detention may be suitable. Offence-specific sentencing frameworks generally do not provide for the imposition of such alternative sentences because, as I have just said, they are typically based on the usual sentences prescribed for the offence; hence, my observation that the sentencing court should first establish the dominant sentencing objective(s), before it assesses whether to apply an offence-specific or an offender-specific framework.

18. But apart from the four sentencing objectives, it is vital that the sentencing court not lose sight of the end goal of sentencing. In most cases, this will ultimately be to “secure the rehabilitation, reform and reintegration into society of all offenders, without undermining broader societal goals of preserving law and order”.[26] As I have explained on previous occasions, the rehabilitation and reintegration of offenders is critical for at least two principal reasons.

(a) First, rehabilitation and reintegration are crucial to reducing recidivism and thus preventing crime, which is a core goal of the criminal justice system. Rehabilitation tackles the root causes of criminality, which if not addressed can give rise to repeated cycles of offending. And reintegration ensures that upon completion of their sentences, discharged offenders have a chance to build a new life and to break away from undesirable associations and habits that can lead them to reoffend.[27]
(b) Second, justice and human dignity demand that once an offender has repaid his debt to society by serving his sentence, he should be allowed to re-join the community, and not be kept outside of it. Ex-offenders must be encouraged to resume the rights of full citizenship, and to chart a life free from the shadow of crime.[28]

19. In short, both the sentencing objectives and the end goal of advancing the rehabilitation and reintegration of offenders will guide the court when it sentences the offender before it. That said, this can leave too much room for variance and an important principle that underlies the legitimacy of the criminal justice system as a whole is that it must be fair across classes of offenders. This concept explains why, for instance, the doctrine of prospective overruling, which is rarely applied outside the domain of sentencing, has sometimes been invoked by our courts in establishing or clarifying sentencing frameworks. For example, in the recent case of ABC, the court held that the Pram Nair[29] sentencing framework applies to all cases of sexual assault by penetration of a minor, regardless of whether the victim had consented, but gave only prospective effect to this ruling.[30] This was because offenders in the appellant’s position had not been sentenced under the Pram Nair framework in earlier cases, hence receiving far lower sentences than would have been imposed under that approach, and to apply that framework to the appellant would therefore have been grossly unfair. More generally, fairness across offenders requires, in the context of criminal sentencing, that as far as possible, and subject to some principled exceptions, like cases should broadly be treated alike. This search for broad consistency in sentencing is aided by more concrete guidance on how a court should determine the appropriate sentence in a particular case. Such guidance can be found in sentencing frameworks, which not only provide signposts to guide a court through the sentencing process, but also help secure broad consistency.

    B. The role, forms, and limits of sentencing frameworks

20. This brings me to sentencing frameworks and their place in the sentencing process. This has become such a significant part of our sentencing jurisprudence that it has inspired my colleague District Judge Kow Keng Siong to write a book on the subject, giving invaluable guidance, and I do commend it to you.[31] In my remarks today, I will focus on offence-specific frameworks, which form the bulk of our sentencing frameworks, but some of what I will have to say also applies to offender-specific frameworks. As I have noted, frameworks serve to secure broad consistency in sentencing outcomes, by establishing patterns of reasoning that can guide the court to a sentence that is both appropriate on the facts of the case and broadly in line with the sentences imposed on similarly placed offenders.[32]

21. To that end, our sentencing jurisprudence has developed three broad principles regarding sentencing frameworks: first, that there are different types of frameworks that each will be appropriate for different types of offences; second, that an arithmetical methodology should not be used because sentencing is generally better understood as an art, rather than as a science; and third, that it may not be appropriate to lay down any sentencing framework in some instances. Let me develop each of these points.

22. First, our courts have laid down different types of sentencing frameworks. We have identified in the cases, at least five main approaches: the “single starting point” approach, the “multiple starting points” approach, the “benchmark” approach, the “sentencing matrix” approach and the “sentencing bands” approach.[33]

23. The choice between these approaches turns on the features of the offence in question. Relevant considerations include the type of factors that affect the gravity of the offence, the typical way in which the offence manifests, and the prescribed punishment regime for the offence. Taking each of these in turn:

(a) The type of factors affecting the gravity of the offence will often be critical to the choice of approach. Where the gravity of the offence is largely a function of a single variable that is measurable using a quantitative metric, the “multiple starting points” approach is often adopted. This involves setting different starting point sentences, that correspond to different classes or degrees of seriousness of the offence.[34] For example, our courts have laid down sentencing frameworks for the offence of drug trafficking, under which the weight of the drugs trafficked is used to differentiate between different classes of the offence each attracting different starting point sentences.[35] This approach is appropriate for drug trafficking offences because the gravity of the offence will usually be linked to one factor: the amount of the drugs in question.[36] By contrast, where the gravity of an offence turns on a set of principal factual elements that are common to most presentations of a given offence, the “sentencing matrix” approach is usually favoured. Under this approach, the criminal conduct is assessed in terms of those key elements.[37] And where the offence occurs in a wide variety of ways, such that there are no key factual elements common to cases of the offence, the “sentencing bands” approach may be more viable.
(b) As to the way the offence typically manifests, where this often happens in a certain way, or where a particular form of offending is common and merits special attention, the “benchmark” approach may be viable.[38] This involves identifying the typical sentence for an archetypal case of the offence. One example is the benchmark sentence of four weeks’ imprisonment for a simple assault of a public transport worker, where the offence is prosecuted under section 323 of the Penal Code, and where the offender pleads guilty and is untraced.[39] Such an approach is unlikely to be helpful where the offence tends to manifest in many ways, none of which call for special attention. Thus, in crafting a sentencing framework for the offence of rape, the Court of Appeal noted that there is no typical case of rape and accordingly eschewed the “benchmark” approach.[40]
(c) Similarly, the prescribed punishment regime may also be relevant to the choice of approach for a sentencing framework. For instance, in a recent case in which the court laid down a sentencing framework for a species of the offence of reckless or dangerous driving (namely, for such offences which do not lead to personal injury),[41] the court observed that Parliament had prescribed different sentencing ranges for different types of harm that could arise (such as death and grievous hurt). It followed that a “sentencing matrix” based on the two factors of harm and culpability would not be apt in cases where the range of outcomes on the “harm” axis was non-existent or unhelpfully narrow as would be the case where the harm was death. In that case, the court adopted the “sentencing bands” approach instead.[42]

24. The second key point about sentencing frameworks is that an arithmetical methodology will not be appropriate. This follows from the role of sentencing frameworks.

25. Sentencing frameworks strive to secure broad consistency among sentences. They are not meant to achieve outcomes that would yield a mathematically perfect graph.[43] Such precision is impossible, because although we might reasonably be able to distinguish more serious from less serious offences, it is simply not practicable, nor even always possible, to neatly arrange every instance of a given offence along a scale showing precisely calibrated degrees of severity. The circumstances of offences and offenders vary much too widely[44] and thus, some cases will simply not lend themselves to such a precise comparison. For instance, it might be possible to agree that one broken finger is a less serious injury than two. It may not be as clear if the comparison was between a dominant thumb and two little fingers.

26. More fundamentally, a formulaic approach to sentencing frameworks is inimical to the very exercise of sentencing. Sentencing calls for an “evaluative ethical judgment[45] that accounts for all of the relevant facts. The aim is to arrive at the right sentence for the case at hand and this must depend on a judgment of a number of factors. Yet an arithmetical method encourages the assessment of the severity of an offence to be based on just one or two factors. This achieves the veneer of consistency between cases, but only at the price of distorting and downplaying other factors material to sentencing.[46] This danger was noted in a 2018 case regarding the offence of cheating at play in a casino.[47] The court rejected a proposed sentencing framework that centred on the amount of money cheated, explaining that this carried the risk of eclipsing or diluting the significance of other relevant sentencing considerations.[48]

27. In addition, sentencing guidelines that aim to achieve mathematical precision are often very complex. They are thus prone to cause confusion and uncertainty, which undercuts their fundamental purpose of guiding the sentencing court to an appropriate sentence, and may also unduly fetter the court’s discretion to determine the appropriate sentence in the light of all the facts and circumstances. These concerns underlay the recent decision of a 3-judge panel of the General Division of the High Court to eschew a sentencing framework for private sector corruption offences that had been laid down in an earlier case.[49]

28. The third important theme of our jurisprudence on sentencing frameworks is that it may be inappropriate to lay down such frameworks in some cases.

(a) To begin with, an offence may arise in so many different ways that a single framework covering all its instances may simply not be viable. And so, in 2016, the High Court refrained from laying down a sentencing framework for the offence of giving false information to a public servant, noting that this offence captured such a wide range of conduct that it was “exceptionally difficult to identify a single set of principal factors which could form the basis of a sentencing framework”.[50] In other cases, our courts have addressed this problem by developing a sentencing framework only for a particular variant of the offence, for which the principal sentencing factors are more confined. For example, the High Court recently laid down a sentencing framework for cheating to procure sex. While the generic offence of cheating did not lend itself to a sentencing framework given the many ways in which it can arise, a framework for this species of the offence was found to be viable in view of “the narrow and more constrained forms” it might take.[51]
(b) It may also be inappropriate to lay down a sentencing framework where there is a dire paucity of case law. In such cases, it may sometimes be difficult to identify relevant sentencing factors or principled sentencing ranges.[52] Laying down a sentencing framework in such situations risks “prematurely ossifying the law”.[53] Nevertheless, this concern must be balanced against the need to provide guidance, which might be compelling where, for example, there are many pending prosecutions for the offence, or where conflicting sentencing approaches have already emerged in the case law.[54] It may also sometimes be necessary to craft a sentencing framework from first principles, with the caveat that it may be refined with the accretion of case law.[55]

29. Ultimately, what lies behind this nuanced picture of the limits of sentencing frameworks and their development is the role that they are meant to play. And I emphasise: they seek to provide principled guidance to sentencing courts, but in view of the nature and incidence of the offence, such guidance may have to be limited or provisional, or eschewed altogether in favour of other methods for securing broad consistency in sentences.

    C. The application of sentencing frameworks

30. Let me turn to the application of sentencing frameworks. This raises two notable points.

31. First, sentencing frameworks should not be applied in a mechanical way. Although such frameworks provide a vital frame of reference for the sentencing court, sentencing should never be relegated to a formulaic exercise, and the sentencing court must always be mindful of its responsibility to ensure that the sentence fits the crime.[56]

32. This point was stressed in a 2016 case that concerned the offence of possessing counterfeit goods.[57] In an earlier case, the High Court had laid down a sentencing framework which identified the number of infringing articles as a relevant sentencing factor. In the 2016 decision, the court cautioned against focusing on the number of infringing articles alone, noting that this might not accurately reflect the culpability of the offender since, for example, a pair of earrings could count as two items whereas an item of far higher value might only count as one.[58] Similarly, in the context of property offences, our courts have emphasised that although the value of the property involved is a key sentencing factor, sentencing in such cases does not involve “the application of a mathematical formula” and thus, the sentences will not bear a relationship of linear proportionality with the amounts involved.[59]

33. The second key point is that the application of sentencing frameworks and sentencing more broadly will rarely be regarded as raising questions of law that may be referred to the Court of Appeal under the criminal reference procedure. In a sense, this follows from the fact that a sentencing court is concerned with determining and imposing the condign sentence for the case that is before it. This was explained in the Bernard Lim[60] case, where the offender was convicted on two charges of providing false information to a public servant under section 182 of the Penal Code and was sentenced to the maximum fine. This was affirmed on appeal. The Prosecution then filed a criminal reference to refer a question to the Court of Appeal. This was framed in terms of whether the default starting sentence for an offence under section 182 of the Penal Code should be a custodial sentence, where the offender was a public servant or the employee of a statutory board and gave false information during investigations into improprieties relating to procurement and/or allegations of abuse of office or power.[61]

34. The Court of Appeal declined to answer the question. The court observed that the Prosecution was in essence seeking to establish a sentencing benchmark, and held that it was inappropriate to do so by way of a criminal reference for two main reasons.[62] First, sentencing is an intensely fact-sensitive inquiry, and the determination of the appropriate sentence in a given case will generally not present a question of law, let alone a question of law of public interest.[63] Second, an issue of sentencing can only be couched as a question of law if it is framed by reference to a broad set of facts. This widens the range of scenarios that may fall within the scope of the question, and if a sentencing norm is established on this basis, it might apply to scenarios of varying gravity, causing unfairness.[64] On the other hand, confining the question to the facts presented would render it impossible to view it as having any sort of wider interest.

IV. The future of sentencing

35. Let me turn finally to the future of sentencing in Singapore. As I explained at the start of this address, sentencing courts have various tools such as sentencing guidelines to aid us in striking a balance between meting out a condign sentence in the case at hand and ensuring fairness across offenders. In the coming years, more such tools may become available. I will briefly address two developments on the horizon: (a) the SAP, and (b) the possible use of AI in sentencing.

    A. The SAP

36. Let me start with the SAP. It may be useful to recall the background to this development. In March 2013, the Judiciary formed a Sentencing Council consisting of a panel of Supreme Court judges and a Senior District Judge. A key goal of the Sentencing Council was to achieve greater consistency in sentencing, by securing guidance on sentencing methodology.[65] To this end, the Sentencing Council created a mechanism for appropriate cases to be placed before a 3-judge panel of the High Court with a view to the issuance of guideline judgments that set down sentencing frameworks. Since March 2013, at least 16 guideline judgments have been issued by such benches. In addition, the Court of Appeal has issued at least six such judgments. And together with judgments issued by a single Judge of the High Court, the Supreme Court has issued around 100 guideline judgments since the formation of the Sentencing Council.

37. The current system has been commended for leading to “a greater degree of coherence … rationality and strength in the positions taken in our sentencing regime”.[66] However, there have been some limitations. Presently, an authoritative sentencing guideline can only be established when an appropriate case reaches the Supreme Court. Further, in some cases, information that might be relevant to the formulation of a sentencing framework, such as data relating to crime trends, might not be placed before the sentencing court.

38. In March 2021, the Minister for Law announced that the Government, after discussions with the Attorney-General’s Chambers (or “AGC”) and the Judiciary, had decided to form the SAP.[67] The Minister explained that the key function of the SAP would be to issue sentencing guidelines, which would be persuasive but not binding on the courts. These guidelines would be published and accessible to the public, thus promoting public knowledge about the likely sentence for an offence and the relevant sentencing factors. The formation of the SAP would also enable the proactive creation of guidelines to address areas that might otherwise take some time to come before the appellate courts, and allow all stakeholders to contribute to their development.

39. The SAP was duly established earlier this year. It comprises members of the Judiciary, the Ministries of Law and Home Affairs, the Police, the AGC and the Criminal Bar. A dedicated website for the SAP has been set up,[68] thus allowing the public easily to access relevant materials including the SAP’s sentencing guidelines. In June this year, it was announced that the SAP had held its inaugural meeting, where the members had agreed to study areas including guidelines on general sentencing principles.[69]

40. The work of the SAP has just begun. But I have no doubt that in the years to come, it will advance our overarching goals of promoting consistency and transparency in sentencing while enhancing public awareness of sentencing, and thus come to play a very significant role in our criminal justice system.

    B. AI in sentencing

41. Finally, I touch on the possible use of AI in sentencing. In the context of criminal justice, this a vexed subject. It is helpful to consider how other jurisdictions have used AI in sentencing, and the issues this might give rise to.

42. First, AI has been used to assess an offender’s risk of recidivism for the purpose of sentencing by several states in the United States including Kentucky, Pennsylvania and Wisconsin. These states have adopted AI risk assessment tools, which use demographic data (such as data relating to the age and gender of offenders) and offence-related data to assess an offender’s risk of recidivism.[70] In the 2016 case of Loomis, the offender argued that the court’s use of the COMPAS tool, an AI risk assessment tool that is widely used in the US, breached his due process rights.[71] The Wisconsin Supreme Court dismissed this challenge, but stressed that although a sentencing court could consider such a risk assessment, it was just one of the factors to be considered, and could not be used to determine such things as whether an offender should be incarcerated or the severity of the sentence.[72]

43. Perhaps more radically, AI has also been used to generate sentencing recommendations for sentencing courts in Malaysia. In 2020, a pilot programme for the use of AI in sentencing began in Sabah and Sarawak, covering the offences of drug possession and rape respectively.[73] Details such as the weight of the drugs and the age of the accused are entered into the system, to generate a non-binding sentencing recommendation.[74] Judges have followed the AI system’s recommendation in about a third of cases. In other cases, they have reduced the recommended sentences to account for mitigating factors or increased them to enhance the deterrent effect of the sentences.[75]

44. As I said earlier, the use of AI tools in sentencing has attracted controversy and it remains contentious. For example, both the Malaysian and the US AI tools have been criticised by some on the basis that the underlying algorithms are opaque, and the offender does not have an opportunity to test the data and assumptions that underlie them, which could reflect biases.[76] Indeed, in Loomis, the court cited a study that found that the COMPAS tool was far more likely to incorrectly assess black offenders as having a high risk of recidivism than it was to make the same error regarding white offenders.[77]

45. It remains to be seen if these and other concerns can be satisfactorily addressed. The use of AI in criminal cases requires careful consideration, and it is therefore fitting that this Conference will discuss this topic in detail tomorrow. I personally do not at this time see us moving in this direction in the criminal justice field within the foreseeable future and I think that AI may, if at all, play a role initially in enhancing access to justice in simpler civil cases. Nonetheless, it remains an intensely important subject that we should all follow closely.

V. Conclusion 

46. In 1810, Sir Samuel Romilly, in one of the earliest discussions of judicial discretion in sentencing, noted wide disparities in sentencing outcomes which he decried as a “lottery of justice”.[78] And, in a study of sentencing in Singapore published around 25 years ago, Justice M Karthigesu observed that for a long period of our history, there had been no consistent sentencing policy and sentences had therefore been inconsistent.[79]

47. Sentencing has long ceased to be a “lottery of justice”. Sentencing frameworks have enhanced consistency, coherence and transparency in sentencing decisions. They have therefore advanced the perennial quest of the sentencing court which, as I have emphasised, is to impose a condign sentence on the individual offender while ensuring fairness across offenders. Sir Samuel Romilly expressed the hope that “there might be, if not a perfect uniformity in [sentencing], yet the same spirit always prevailing, and the same maxims always kept in view”.[80] As I have tried to illustrate, the story of sentencing discretion in Singapore has been a journey towards that end, and our journey continues. Thank you very much for your attention this morning, and I wish you a wonderful Conference.

* I am deeply grateful to my law clerk, Ian Mah, and my colleagues, Assistant Registrars Huang Jiahui and Tan Ee Kuan for all their assistance in the research for and preparation of this address.

[1]     Mohammad Faizal bin Sabtu v PP [2012] 4 SLR 947 (“Mohammad Faizal”) at [37]–[38].

[2]     Colin Munro, “Judicial Independence and Judicial Functions” in Sentencing, Judicial Discretion and Training (Colin Munro & Martin Wasik eds) (Sweet & Maxwell, 1992) (“Munro”) at p 27: sentencing was not identified as part of the judicial power in Blackstone’s Commentaries and The Federalist Papers.

[3]     David Thomas, “Sentencing in England” (1983) 42 Maryland Law Review 90 at 106.

[4]     David Thomas, “Judicial discretion in sentencing” in Exercising Discretion: Decision-making in the criminal justice system and beyond (Loraine Gelsthorpe & Nicola Padfield eds) (Willan Publishing, 2003) ch 3 (“Thomas, “Judicial discretion in sentencing”) at p 51.

[5]     Thomas, “Judicial discretion in sentencing” at p 51.

[6]     Thomas, “Judicial discretion in sentencing” at pp 51–53.

[7]     Thomas, “Judicial discretion in sentencing” at p 51.

[8]     Sundaresh Menon CJ, Keynote Address, Sentencing Conference 2017 (26 October 2017) (“Sentencing Conference 2017 Keynote Address”), para 4.

[9]     Sentencing Conference 2017 Keynote Address, paras 5–8.

[10]    See sections 337(1)(d) and 337(2)(c) of the Criminal Procedure Code 2010.

[11]    Under the Misuse of Drugs Act 1973 and the Arms Offences Act 1973.

[12]    Singapore Parliamentary Debates, Official Report (26 July 1984), vol 44 (Mr Chua Sian Chin, The Minister for Home Affairs).

[13]    Singapore Parliamentary Debates, Official Report (22 October 2007), vol 83 (Assoc Prof Ho Peng Kee, The Senior Minister of State for Home Affairs).

[14]    Singapore Parliamentary Debates, Official Report (14 November 2012), vol 89 (Mr K Shanmugam, The Minister for Law) at p 1261.

[15]    See ss 169, 173, 175, 176, 182 and 186 of the Criminal Law Reform Act 2019.

[16]    Moses Hinds v R [1977] AC 195; Mohammed Muktar Ali v R [1992] 2 AC 93.

[17]    Mohammad Faizal (regarding the sentencing scheme for repeat drug consumers, under which decisions of the Director of the Central Narcotics Bureau to order drug consumers to be admitted to rehabilitation centres triggered the imposition of mandatory minimum sentences); Prabagaran a/l Srivijayan v PP and other matters [2017] 1 SLR 173 (“Prabagaran”) and Abdul Kahar bin Othman v PP [2018] 2 SLR 1394 (“Abdul Kahar”) (concerning the alternative sentencing regime under s 33B of the MDA).

[18]    State of South Australia v Totani (2010) 242 CLR 1 at [82]–[83].

[19]    Prabagaran at [77]; Abdul Kahar at [48].

[20]    The threshold for departing from guidelines varies between jurisdictions. In England, the Sentencing Council’s sentencing guidelines are binding albeit the court may depart from them if it is contrary to the interests of justice to follow them: see s 59(1) of the Sentencing Act 2020 (UK). By contrast, the US Federal Sentencing Guidelines are non-binding, following United States v Booker (2005) 543 US 20.

[21]    Sundaresh Menon CJ, Opening Address, Sentencing Conference 2014 (9 October 2014) (“Sentencing Conference 2014 Opening Address”), para 7.

[22]    PP v ASR [2019] 1 SLR 941 (“ASR”) at [130]–[131]; PP v Low Ji Qing [2019] 5 SLR 769 at [78] and [80].

[23]    Sentencing Conference 2014 Opening Address, para 7.

[24]    PP v Koh Wen Jie Boaz [2016] 1 SLR 334; ASR at [94]–[102].

[25]    Sim Yeow Kee v PP and another appeal [2016] 5 SLR 936.

[26]    PP v Siow Kai Yuan Terence [2020] 4 SLR 1412 (“Terence Siow”) at [1]; Sentencing Conference 2014 Opening Address, paras 41–50.

[27]    Sentencing Conference 2017 Keynote Address, paras 1 and 51.

[28]    Sundaresh Menon CJ, Opening Address, Reintegration Puzzle Conference 2014 (30 July 2014), para 9.

[29]    Pram Nair v PP [2017] 2 SLR 1015.

[30]    ABC v PP [2022] SGHC 244 at [59]–[63]. 

[31]    Kow Keng Siong, Sentencing Frameworks in Singapore: Quick Reference Guide and Roadmap (Academy Publishing, 2022).

[32]    PP v Wong Chee Meng and another appeal [2020] 5 SLR 807 at [57]; M Raveendran v PP [2022] 3 SLR 1183 at [46].

[33]    Ng Kean Meng Terence v PP [2017] 2 SLR 449 (“Terence Ng”) at [26] and [39].

[34]    Terence Ng at [29].

[35]    Terence Ng at [30].

[36]    Vasentha d/o Joseph v PP [2015] 5 SLR 122 at [23].

[37]    Terence Ng at [34].

[38]    Terence Ng at [32].

[39]    Wong Hoi Len v PP [2009] 1 SLR(R) 115 at [20].

[40]    Terence Ng at [32].

[41]    Wu Zhi Yong v PP [2021] SGHC 261 (“Wu Zhi Yong”).

[42]    Wu Zhi Yong at [27]–[29].

[43]    Mohd Akebal s/o Ghulam Jilani v PP and another appeal [2020] 1 SLR 266 at [20(b)].

[44]    Lee Shing Chan v PP and another appeal [2020] 4 SLR 1174 at [34].

[45]    Terence Ng at [48]; Dinesh Singh Bhatia s/o Amarjeet Singh v PP [2005] 3 SLR(R) 1 (“Dinesh Singh Bhatia”) at [24].

[46]    R A Duff, “Guidance and Guidelines” (2005) 105(4) Columbia Law Review 1162 at 1173–1174.

[47]    Logachev Vladislav v PP [2018] 4 SLR 609 (“Logachev Vladislav”).

[48]    Logachev Vladislav at [48].

[49]    Goh Ngak Eng v PP [2022] SGHC 254 at [33], [34], [42] and [43].

[50]    Koh Yong Chiah v PP [2017] 3 SLR 447 at [46]; see also PP v Tan Kok Ming Michael and other appeals [2019] 5 SLR 926 at [104].

[51]    Wong Tian Jun De Beers v PP [2021] SGHC 273 at [38].

[52]    Cai Mei Ying v PP [2019] SGHC 24 at [6]; Lau Wan Heng v PP [2022] 3 SLR 1067 at [40].

[53]    Kwan Weiguang v PP [2022] SGHC 121 at [35].

[54]    Sue Chang v PP [2022] SGHC 176 (“Sue Chang”) at [50]–[51].

[55]    Sue Chang at [52].

[56]    Dinesh Singh Bhatia at [24].

[57]    Tan Wei v PP [2016] 3 SLR 450 (“Tan Wei”).

[58]    Tan Wei at [16].

[59]    PP v Lam Leng Hung and other appeals [2017] 4 SLR 474 at [368], citing PP v Tan Cheng Yew and another appeal [2013] 1 SLR 1095 at [184].

[60]    PP v Lim Yong Soon Bernard [2015] 3 SLR 717 (“Bernard Lim”).

[61]    Bernard Lim at [3].

[62]    Bernard Lim at [20].

[63]    Bernard Lim at [21] and [30]–[31].

[64]    Bernard Lim at [23] and [28].

[65]    Chao Hick Tin JA, “The Art of Sentencing – An Appellate Court Perspective”, Sentencing Conference 2014 (9 October 2014), para 3.

[66]    Benny Tan Zhi Peng, “Assessing the Effectiveness of Sentencing Guideline Judgments in Singapore Issued Post-March 2013 and A Guide to Constructing Frameworks” (2018) 30 SAcLJ 1004 at para 22.

[67]    Singapore Parliamentary Debates, Official Report (5 March 2021), vol 95 (The Minister for Law, Mr K Shanmugam).

[68]    See <>.

[69]    Ministry of Law, “Press Release: Establishment of the Sentencing Advisory Panel” (2 June 2022) <>.

[70]    Jennifer Walker Elrod, “Trial by Siri: AI Comes to the Courtroom” (2020) 57 Houston Law Review 1083 at 1089–1090.

[71]    State v Loomis 371 Wis 2d 235 (Wis, 2016) (“Loomis”) at [34].

[72]    Loomis at [98] and [109].

[73]    Christina Erin Ong, “The AI Dilemma: Rise of the Machines in Malaysian Criminal Sentencing”, Legalatte (26 July 2021) <>.

[74]    Mahyuddin Daud, “Artificial Intelligence in the Malaysian Legal System: Issues, Challenges and Way Forward” (2022) 39(1) INSAF 1 at 9–10.

[75]    “As Malaysia tests AI court sentencing, some lawyers fear for justice”, The Straits Times (12 April 2022) <> (“AI court sentencing”).

[76]    AI court sentencing; Noel Hillman, “The Use of Artificial Intelligence in Gauging the Risk of Recidivism” (2019) 58(1) Judges’ Journal 36 at 37.

[77]    Loomis at [62]–[63].

[78]    Samuel Romilly, Observations on the Criminal Law of England (London: Cadell and Davies, 1810) (“Romilly”) at p 21.

[79]    M Karthigesu, “Criminal Sentencing Policy” in Singapore Academy of Law, Review of Judicial and Legal Reforms in Singapore Between 1990 and 1995 (Butterworths Asia, 1996) at p 389.

[80]    Romilly at p 16, cited in Thomas, “Judicial discretion in sentencing” at p 71.

Topic: Speech, Speeches


Share this page: