◆ Common Question ◆

[A] The focus on the War on Drugs and the use of severe sentencing as a tool in that War appears to have substantially abated. I believe this is because drugs are no longer one of the most feared reasons for serious crime and an acknowledgement by many that severe punishments do not operate well in reducing demand considering the neurophysiological impacts of drugs that reduce the cognitive ability to weigh benefits and consequences and thus to modify one’s actions.

The use of data sciences is on the rise in terms of making available statistics on sentences (e.g., average length of prison terms, proportional use of probation or fine-only sentences) and in the use of algorithmic risk assessment. Both are meant to draw on big datasets and to make sentencing appear more objective, consistent, and transparent.

Somewhat related to #2 is the increased reliance upon academic disciplines to inform sentencing practices (e.g., criminology, psychology, neuroscience). For example, policymakers in several countries (e.g., U.S., England & Wales, Scotland, Ireland) have been funding literature reviews from external academics on selected topics to better understand and thus to rule on relevant issues, such as in developing new or amended guidelines.

 
Many systems heavily rely on criminal history in establishing penalties for current offenses. But they do so without properly acknowledging and then ameliorating the fact that criminal history is not as objective as many presume. Criminal history records generally rely upon arrests by police, actions that are not immune to biases and discrimination. Minorities may have greater criminal history, not just because they may have committed more crimes, but they simply may have been more likely to be arrested than non-minorities. Or, simply living in a heavily policed neighborhood renders one more likely to have a record simply by virtue of the fact of where they live. Policymakers could delve into such issues with more care. The impacts are not equal with all crimes. For example, a criminal history for murder is likely less biased than for drug usage. Another issue with criminal history is that if the guidelines and/or the judges considers criminal history and then use a risk assessment tool (almost all of which have criminal history as predictors in some form and heavily weight them) then the same prior criminal convictions might be overcounted and thus be overweighted, with the individual subjected to a more severe penalty than otherwise deserved. In other words, care should be taken that risk assessment does not inflate penalties due to this multiplicative effect.


2. What do you think of Korea's sentencing standards? What are the problems with the sentencing standards in Korea?

[A] Overall, Korea’s creation of its sentencing system appears to be a thoughtful attempt to learn from the successes and mistakes of other jurisdictions who have adopted a guidelines-based system while maintaining its own cultural niceties that are deemed important to its public’s identity. No system is perfect and there are always unintended consequences to even the most well-thought plan.

 
I find the rejection of a sentencing grid or the placement of specific time periods on offenses or offense characteristics to be a reasonable choice. The scientific literature confirms the existence of an anchor effect whereby humans when given numbers tend to orient to those numbers, which means that their judgement is impacted thereby. The anchoring effect is largely unconscious so the person may deny the weight of the numbers yet the effect may be real nonetheless.

An issue with the Korean Sentencing Guidelines is evident in Part B in which a primary consideration is the risk of recidivism, which includes specified adverse factors as if they represent risk factors and then affirmative factors which appear to represent protective factors. The Guidelines do not indicate any scientific support for those factors as actually being correlated with the risk of recidivism (which itself is given no definition). This means the use of risk assessment is not compliant with evidence-based practices which is meant to rely upon scientifically-validated research. I understand the Ministry of Justice developed an actuarial tool called KORAS-G that shows promise if it were subject to a recent validation study and showed sufficiently viable performance statistics.

There appears to be a short time frame between the adjudication of guilt and passing a sentence. If so, that might be at odds with an informed sentencing that properly can consider the relevant guidelines and have an understanding of the individual aspects of the case regarding the circumstances of the crime, the offender, and any victim. If these factors and the guidelines are as a result being considered prior to the finding of guilt, then the decision on guilt may itself have been biased by sentencing considerations as the human brain is then already almost assuming guilt in terms of considering the requisite penalty.

 
Many jurisdictions offer some form of second-look sentencing, which here means that at some point in the future of a long prison sentence, an appropriate person or body reconsiders if that lengthy penalty still appears to be reasonable. Things may have changed in the meantime, such as society’s judgement on the severity of the crime committed. Over time, the public’s view of an act may have shifted along with cultural changes, broader acceptance of acts previously considered deviant, or just an evolution in what is considered top priority. In some societies, for example, certain acts that were at one time considered heinous criminal offenses no longer hold the same meaning so that the same person sentenced in a prior time may be judged far less harshly today (e.g., marijuana, abortion, homosexual acts, adultery). A second-look sentencing regime would permit a reduced sentence if current sentiment no longer finds the continuing penalty to be just. Of course, public sentiment may shift in the other direction, but rules of criminal generally would not permit an increased sentence.

The guidelines could provide greater assistance in how to assign supervisory or rehabilitative conditions to a probation sentence.

 
More special attention to family abuse and domestic violence could be a significant contribution as these cases have their own unique and difficult contexts.

3. Amongst Korean citizens, there is a large gap between the perception of the law, and the court's sentencing. Is this a global trend, and not just in Korea? How should we proceed to close this gap?

[A] The gap between the public’s perception of sentencing and actual sentencing practices and statistics is an issue far beyond Korea. We cannot say it is worldwide as methodologically appropriate studies have tended to focus on specific countries. In the countries or locales that these studies have surveyed, a relatively consistent result is that respondents typically have little accurate information about their jurisdiction in terms of criminal punishments. Experts believe there are several reasons for this gap: (1) the public can be swayed by news reports of high profile cases, such as a widely covered story of a rapist who received a light sentence, and thus the public might then believe that judges are too lenient for rape cases generally rather than understand that there may have been something unique about the case or a particularly sympathetic individual judge; (2) criminal law and procedure are complicated and the public may not be cognizant of its intricacies; (3) the lack of understanding the difference between a sentence imposed and the sentence actually served (e.g., blame sentencing judges for what is perceived as light sentences when it might have been parole authorities who released early).

It is not, therefore, unexpected that the public will not have a great understanding of actual sentencing practices. Even experts in sentencing in a particular jurisdiction can be surprised when confronted with actual statistics in that same jurisdiction. I myself have for years worked with the actual sentencing datasets for the federal criminal jurisdiction in the U.S. I have to run data analyses because it would be inappropriate to make assumptions despite a vast knowledge of the area. At times I am surprised by the results of certain analyses where the data say something unexpected. For example, I might run a query on the percentage of individuals who received a probation sentence (i.e., no prison time) for a particular crime and find a much higher (or lower) proportion than expected.

 
To close the gap, the Commission might (a) make available a user-friendly way for the public to obtain statistics on sentencing, such as using a web interface that allows the user to select ways to parse the data (e.g., by crime, by year, by type of sentence, median sentence length), (b) engaging with stakeholder groups (neighborhood associations, victim support groups, advocacy groups), having discussions, giving presentations with information on sentencing practices, and offering a forum for concerns, (c) advocate for instruction on criminal law and procedure in secondary and university education, (d) hold webinars for the interested public, (e) proactively work with the media to educate them about sentencing practices and also potentially reduce how much they might otherwise hype singular and unique cases that do not represent the norm in sentencing outcomes.

 
One of the reasons for the perception of leniency could be tackled from another perspective, which is when judges are reducing sentences based on myths. A prime example of this is in sentencing for rape and sexual assault. Judges who hold rape myths (e.g., that the victim who drinks or dresses “provocatively” are responsible or that the offender who is drunk is less culpable) are more likely to reduce sentences. If judges were educated with scientific research about why those myths were incorrect, perhaps they would be less likely to accept the offender’s mitigation arguments.

4. Should sentencing be carried out only by humans? Can fair sentencing by artificial intelligence, so-called AI judges, be possible?

[A] As a general rule, jurisdictions are not seriously considering entirely AI sentencing. Any form of AI in the future is envisioned as another information point for humans to consider when making sentencing decisions. An important reason is that there still is considered a symbolic and cultural value in a human imposing punishment on another human. Many jurisdictions also continue to embrace some form of mercy and AI cannot be merciful in the human sense. Another way to consider this is through the lens of justice. No matter now detailed the laws or guidelines might be, there will always be cases that do not fit the mold in terms of the sentence actually feeling as if justice was done. There are just too many crimes, ways that crimes are committed, contextualizing circumstances, and individual factors that no even highly structured system can possibly cover.

◆ Individual Questions ◆

1. What do you think about the use of artificial intelligence (AI) algorithms as a means of assessing the risk of recidivism in the courts? (#U.S. Supreme Court of Wisconsin, in 2017, ruled that it would be illegal if the court relied entirely on artificial intelligence algorithms while evaluating the risk of recidivism of the accused, but legal when used as an auxiliary means)

[A] I believe the Wisconsin case got some aspects right while other conclusions were ill-informed. I see a legitimate role for algorithmic risk assessment in sentencing as one information source but not as the sole trigger for the penalty. One reason is that the perceived likelihood of the individual reoffending has played a traditional role in sentencing, but typically based on the gut instinct or intuition of the human decisionmaker. Commonly, judges who have defendants who appear before them are likely to perceive that others who share those common sociodemographic attributes are also at high risk of recidivism. Yet humans have conscious and unconscious biases in predicting what another person will or will not do. And humans are active creatures with free will and this makes predictions of behavior always imperfect. This means that the human-derived judgements will be biased and likely often inaccurate. Then studies show that humans (judges included) tend to overpredict the likelihood of recidivism for known offenders. Thus, risk assessment tools can help temper those high predictions with statistics indicating likely much lower rates than expected.

One legal issue I have problems with concerning the Wisconsin decision is insufficient respect for the due process rights of defendants to be able to challenge the algorithm for its accuracy, its biases, and its applicability to the defendant. The Canadian Supreme Court case of Ewert was better in this regard. Ewert determined that the algorithm there had not been tested on indigenous populations and thus could not be used for individuals in those groups (such as Ewert). Risk assessment tools should not be assumed to be universal as there are risk-relevant differences in sub-populations. For example, the risk factors and their salience are known to vary for men versus women and for adults versus children. There are other permutations where risk factors may vary based on race, ethnicity, nationality, religion, health status, etc.

The Wisconsin court in Loomis also did not address the issue of whether the risk tool was fit for purpose. It did not seem to be aware of exactly what the tool was predicting. In fact, the tool predicts any type of reoffending, even very minor technical infractions (e.g., violating a supervisory condition that without supervision would not be a crime (think drinking alcohol, or violating an imposed curfew)) years into the future. This is probably not what sentencing judges are interested in when deciding to impose a more serious prison term. They are more likely to want to know the probability of serious reoffending in the near future.

 
Another issue that should be addressed is that there is no agreement in the field as to what “high” or “low” risk actually means in terms of the probability of reoffending (however reoffense is defined). These terms are merely relative to the population originally studied as to the groups that have a higher recidivism rate than others. But those rates can vary by population, subpopulation, or jurisdiction. A specific tool may, for instance, show a high risk rate of reoffending of 30% in one site but a 60% rate in another site. Or different tools can vary whereby a tool predicting violent crime may have a high risk associated with a 10% recidivism rate and a more general tool will indicate high risk at a 75% recidivism rate. The same is true for “low risk” buckets that may vary by place and different tools categorize low risk differently. Without care in informing the users and stakeholders what these terms mean, they may be making incorrect assumptions. A common misconception I see by officials in real-world settings is assuming a “high risk” label means more likely than not or greater than 50% chance when that is actually not supported by the tool’s actual statistics.

The discussion so far has assumed algorithmic risk assessment. There is a difference in those that are truly AI in the sense of machine-learning or other development methods that are weeding out human involvement. We are not yet at a stage to ascertain what other issues may arise when the human developer is further removed from the developmental process.

2. Recently, serious crimes such as murder due to stalking have occurred one after another in Korean society. Regarding this, there is a discussion going on in Korean courts to consider the 'recidivism risk' more actively than before when deciding whether to detain stalking or sex offenders. What is your opinion as an expert in the risk of recidivism? For reference, under the Criminal Procedure Act of Korea, the reasons for detention of criminals are limited to 3 reasons: 1) denial of residence, 2) destruction of evidence, and 3) fear of fleeing. For example, I wonder if you think that the reason for arresting criminals should explicitly include 'risk of recidivism'.

[A] Research that is focused on the recidivism risk of stalkers is in its infancy. In other words, we don’t know yet a lot about the likelihood of stalkers committing a new crime of stalking, particularly the stalking of new people (versus stalking the same victim repeatedly). Thus, there could be no real role for risk assessment of stalking per se. Then there are different types of stalkers that present different profiles: domestic stalkers (those who stalk their current or former intimate partners) vary from stalkers of acquaintances, strangers, or public figures.

In contrast, a healthy amount of research is available about the recidivism of sex offenders. An important point is that most legitimate research shows that sex offenders have a much lower rate of committing new sex crimes than the public assumes. Criminal justice officials and the public who are not aware of the evidence tend to assume that sex offenders are highly likely (some estimates over 70%) to commit a new sex crime if released. Studies tend to show very low new rates, typically under 10%. But, caution is still suggested. The rates can vary depending on how one defines “sex offense.” What is of most concern is hands-on sexually offending. But many of the studies lump together hands-on and those not hands-on (e.g., voyeurism, public nudity, downloading child sexual assault images, lewd phone calls). The rates would be even lower single digits if those other types were excluded, with the focus on hands-on crimes. Even all hands-on offenses are not necessarily of the greatest concern and thus may not justified detention in the minds of some officials (e.g., statutory rape cases in which the sexual contact was consensual, but the law prohibits the age difference such as a committed couple where one is 18 years and the other is 15 years).

Hence, if there was a change to include the risk of recidivism, the law should be clear on the risk of exactly what types of recidivism and how high that probability should be (in percentage terms) to justify detention. Still, as suggested above, there is simply not enough evidence to know about the risk of stalking itself.


3. Recently, the Ministry of Justice of Korea announced a legislative notice of amendments to the Juvenile Act and the Criminal Act, including lowering the age of criminal minors by one year from the current under 14 to under 13 while announcing comprehensive measures for juvenile crime. In the UK, the age of criminal minors is 10 years old. Do you think that strong punishment for juvenile crimes would be effective in preventing juvenile crimes and recidivism? What do you think is an effective policy tool for this?

[A] Using criminal punishments on those who are age 10 to 14 is more about retribution than it is about deterrence. Deterrence theory requires three items: likelihood of being caught, the swiftness in being brought to justice, and the severity of the punishment. It also generally requires the individual be aware of these three items. Those age 10 to 14 are unlikely to have much awareness about any of those things, much less have an accurate awareness. In terms of the likelihood of being caught, that is more a matter of law enforcement and thus this Act is likely not overly relevant to that aspect. The swiftness in punishment is always a problem in a system, such as Korea, that values due process as this takes a lot of time to ensure that the process is fair and gives defendants sufficient opportunity in which to defend themselves. The last attribute is the one most relevant to the potential new law that would lower the age of criminal culpability. The problem with that is juveniles have no real idea of what punishment they might possibly face in the end, which undermines deterrence theory. In sum, the justification for the new law would be more about retribution and the public’s desire to hold younger people accountable than it would be about preventing or deterrence crime by those younger people.

Brandon Garrett

[A] There is a greater demand for more consistency in sentencing and for more evidence-based approaches that impose punishments that can effectively deter and rehabilitate persons convicted of crimes.In general, sentencing provisions alone cannot accomplish those goals.Crime is most effectively deterred by better apprehension and detection of crime than by sentencing.Alternative sentencing that can effectively rehabilitate persons requires resources for treatment, and is a greater focus in many systems.

2. What do you think of Korea's sentencing standards? What are the problems with the sentencing standards in Korea?

[A] The sentencing approach in Korea does not follow a grid model, but instead describes sentencing considerations that relate to the category of crime, what might make particular acts more aggravated or less aggravated, and what might make a particular offender less blameworthy. It is useful to set out a common and consistent framework for considering criminal sentences. However, it will be important to collect sound data on how consistently these guidelines are used in practice. Where judges retain a great deal of discretion, it is important to study how that discretion is used.

3. Amongst Korean citizens, there is a large gap between the perception of the law, and the court's sentencing. Is this a global trend, and not just in Korea? How should we proceed to close this gap?

[A] It may be more important to ensure that judges are independent and that review of sentencing decisions on appeal can take into account their compliance with guidelines and consistency in imposing sentences.Low-court judicial discretion is important, but there should be mechanisms to address unfairness and arbitrariness in individual or systematic sentencing decisions.One mechanism is appellate review.Other mechanisms include legislative sentencing rules, a greater role for lay decisionmakers, appellate review, and a greater role for probation authorities to determine sentence length after a conviction.

4. Should sentencing be carried out only by humans? Can fair sentencing by artificial intelligence, so-called AI judges, be possible?

[A] Sentencing schemes that rely on algorithms or quantitative information are designed by humans and when they are used, they provide information that human judges consider in their discretion.It is important to use careful analysis of data to study sentencing and to potentially improve it.If quantitative tools are adopted, they must be carefully assessed and validated, as well as interpretable by the judges and lawyers that use them.In general, the types of information that are predictive in criminal cases tends to be very simple information.Complicated artificial intelligence is not needed.Repeat offending is often predicted, as best as it can be, by very simple models that focus on a persons age and prior criminal history.There is no need to use complicated “black box” systems, which can be unfair and misleading.

◆ Individual Questions ◆

1. What is the justification for criminal punishment other than civil and administrative sanctions against corporations?

[A] One important justification is to secure the cooperation of the corporation in criminal investigations of individual employees and officers.The threat of a criminal prosecution can help to ensure that a full investigation is conducted internally.However, in some countries, criminal procedure rules can make that type of cooperation more difficult.

Larger and more punitive fines are not necessarily a justification for corporate criminal punishment.While criminal penalties can be more punitive and may be justified, punitive civil and administrative fines can also be imposed.

2. In Korea, Serious Accident Punishment Act is in force to strengthen criminal penalties for business owners in the event of a serious accident such as a death or accident at a company. However, I know that the United States does not have a law similar to Korea's Serious Accident Punishment Act that criminalizes industrial accidents. Why is that?

[A] There are a range of more specific laws in the United States that apply to different types of workplace accidents and deaths. There are a range of workplace health and safety laws.Criminal prosecutions are possible under federal worker safety laws, including the Occupational Safety and Health Cat of 1970, as well as environmental crimes, and other more specific laws, such as a mine safety statute.

3. There is a growing public opinion that the criminal punishment of corporations should be strengthened. However, there are concerns that if the criminal punishment of corporations is strengthened, they may face economic problems. What do you think of this?

[A] There is a more serious concern that if corporations that commit crimes are not adequately punished, then they will have an unfair competitive advantage over companies that follow the law. A company should fully internalize the social costs of criminal behavior in order to create optimal incentives.Indeed, larger penalties may be required if it is likely that not all violations are disclosed, while conversely, rewarding self-reporting and full compliance can incentivize detention and prevention of corporate crime.

Jorg Kinzig

◆ Common Question ◆

[A] This is a difficult question to answer. Obviously, one way of looking at the global sentencing trend is to look at incarceration rates. According to the latest version of the “World Prison Population List” published in December 2021, Europe is the only continent that has seen a decrease in its total prison population since the year 2000. It is interesting to note that both Korea and my home country Germany experienced a decrease in the prisoner rate during this period. Since 2000, the prisoner rate, the number of prisoners per 100,000 of the national population, dropped in Korea from 136 to 105 and in Germany from 85 to only 70. That said, even though Korea shows a moderate prisoner rate when compared to other East Asian countries, it still incarcerates many more people than Germany.

2. What do you think of Korea's sentencing standards? What are the problems with the sentencing standards in Korea?

[A] To be honest, I was very impressed when I saw Korea’s Sentencing Guidelines for the first time because we do not have a comparable framework in Germany. From my point of view, German judges do have more discretion than their Korean counterparts in terms of determining a sentence. I am fairly certain that criminal courts in Germany would reject the idea of being bound by a strict top-down scheme.

3. Amongst Korean citizens, there is a large gap between the perception of the law, and the court's sentencing. Is this a global trend, and not just in Korea? How should we proceed to close this gap?

[A] Criminal Courts in Germany are not only composed of professional judges but also to a certain extent include lay judges. The reasoning behind this special setup of our courts is the idea to involve the public in the jurisdiction. In fact, in our research institute we recently conducted a survey where we asked a couple of thousand lay judges about their perception of our criminal justice system. Moreover, in Germany it is not only the so-called yellow press that criticizes criminal courts for their lenient decisions from time to time. In my opinion, it is a permanent task for legal scholars, like myself, to explain to the public how our criminal justice system works and why it is not always preferable to pass harsh sentences.

4. Should sentencing be carried out only by humans? Can fair sentencing by artificial intelligence, so-called AI judges, be possible?

[A] This, again, is a rather difficult question which I will try to answer from my German perspective. In my view, it is highly unlikely that the German legislator will introduce sentencing guidelines or even AI judges within the next years. This may have to do with the fact that until now German practitioners have strong reservations about all systems that aim to prepare, propose or even replace human judicial decisions.
Instead, some German scholars promote the idea to establish a sentencing database. That way judges could find out more about how similar cases were evaluated and tried by their colleagues which in turn could lead to more consistent and comparable decisions. That would be a big step forward because right now we see remarkable disparities in sentencing similar cases between different regions in Germany.

◆ Individual Questions ◆

1. What is the biggest difference between the civil law system and common law system in determining the risk of recidivism?

[A] This is a difficult question to answer. As far as I know professional guidelines for violence risk assessment like HCR-20, VRAG, PCLR are used in many countries regardless of whether they have a civil or a common law system.

That said, countries based on the common law system treat the question of dangerousness in different ways. In the German civil law system, for instance, all sentences imposed by criminal courts have to be strictly proportional to the amount of guilt that the perpetrator brought upon himself through the offence in question. Therefore, German judges are only allowed to a very limited extent to impose a harsher sentence if they consider an accused person to be dangerous. To react to the dangerousness of somebody is primarily the responsibility of so-called measures of reform and prevention.

As the European Court of Human Rights has pointed out in M. v. Germany, there is no system of preventive detention in many other European countries. Instead the dangerousness of the offenders in question is taken into account both in the determination and in the execution of their sentence. On the one hand, this results in increased prison sentences in light of offenders’ dangerousness, especially in cases of recidivism. On the other hand, offenders’ dangerousness generally has an influence both on their conditions of detention and on their chances of benefiting from a reduction of their sentence and/or release on probation.

2. What are the criteria for determining the risk of recidivism in Europe and what are your views on the bail system?

[A] The legal systems - especially in terms of criminal law - differ quite strongly across Europe which is why it is difficult to provide an overall answer to this question. In the case of Germany, it is always the duty of German criminal courts to determine a sentence and to examine whether all the prerequisites of the different stationary measures are fulfilled.

Still, in cases where stationary measures are being considered German criminal procedure also invokes the necessity to consult an expert witness who has to make a statement during the main hearing as to the defendant’s condition and his treatment prospects. These experts exert significant influence over the decision of the court. It is the task of the judges to supervise the work of these experts.

Furthermore, in Germany we do not have a bail system like for example in the USA. Remand detention may be ordered against the accused if he is strongly suspected of having committed the offence and if there is grounds for an arrest. That said, remand detention may not be ordered if it is disproportionate to the significance of the case or to the penalty and in cases where measures of reform and prevention are likely to be imposed.

3. In making a sentencing decision, how much do you think it is most appropriate to judge the risk of recidivism?

[A] I think it should always be the duty of criminal courts to decide on this issue. In cases where stationary measures are being considered an additional expert witness should be consulted and judges should supervise the work of these experts. Therefore, when it comes to evaluating the risk of recidivism of offenders, profound professional training of both the expert witnesses and the judges is absolutely necessary to ensure that sentencing decisions are appropriate.

Kuratanoi

◆ Common Question ◆

[A] I am not fully aware of global sentencing trends. I would like to learn more about it through this conference.

2. What do you think of Korea's sentencing standards? What are the problems with the sentencing standards in Korea?

[A] In Japan, no sentencing guidelines have been established, so I watched them with great interest. Although it may be a matter of operation rather than a matter of the Sentencing Guidelines themselves, I felt that it is important to understand the purpose of each statement in the Sentencing Guidelines, such as why a particular circumstance is a reason for aggravation or reduction, and to use the Guidelines without being inflexible in their operation.

3. Amongst Korean citizens, there is a large gap between the perception of the law, and the court's sentencing. Is this a global trend, and not just in Korea? How should we proceed to close this gap?

[A] As to the first part, whether the gap between public opinion and sentencing is a global trend or not, I am afraid I do not have the knowledge to give an appropriate opinion.

To speak of what I am taking care to do to bridge the gap between public opinion and sentencing, first of all, I believe that it is important to explain the reasons for the sentence in specific terms, both to the defendant and to the public. In addition, in Japan, in Saiban-in trials, Saiban-in selected from the public participate in sentencing. In doing so, we are able to hear the public's honest opinions about the sentence. I believe that it is important to reflect such opinions in the sentencing of the case in question and other cases as well.

4. Should sentencing be carried out only by humans? Can fair sentencing by artificial intelligence, so-called AI judges, be possible?

[A] Even if a "fair" sentence could be calculated mechanically based on previous data, neither the defendant nor the public would be convinced by the reasoning that "this is the sentence according to previous data". It is necessary to explain not only the conclusion but also the reasons for reaching the conclusion based on specific circumstances, and we believe that this requires human power.

◆ Individual Questions ◆

1. What do you think about the biggest difference in sentencing policies between Japan and Korea?

[A] I think it is a way to incorporate the public's opinion into the sentencing process. In Japan, the public directly participates in the sentencing of serious crimes through Saiban-in trials. In Korea, the public's opinion is indirectly reflected in the sentencing guidelines by incorporating their views in the establishment of sentencing guidelines.

2. The sentencing correlates with legal opinion of civil society in that country. Is there any difference between public opinion and the court's position on Japan's sentencing in criminal cases?

[A] In Japan, public opinion is taken into account in the sentencing of certain serious crimes by having Saiban-in, elected by the public, participate in the sentencing process. However, the court's sentencing decisions, including those made by Saiban-in trials, are sometimes criticized by public opinion. The court must not be swayed by public opinion, but must sentence the accused based on the law and evidence, and in consideration of fairness. For this purpose, it is important for the court to properly explain the reasons for the sentence to the public.

3. Japan, unlike Korea, strictly judges the crime of breach of trust. It is said that more cases of breach of trust are resolved in civil courts rather than punishing managers in criminal courts. Why would you think this would happen in Japan?

[A] Not only breach of trust, but in general, prosecutors in Japan prosecute only when there is a high probability of conviction based on accurate evidence, and even when there is such a probability, they are allowed to refrain from prosecuting at their discretion (the principle of convenience in prosecution).

I would like to add that In Japan, the number of convictions for breach of trust in the last five years has been around 20 per year. However, the number of cases of breach of trust, plus special breach of trust and business embezzlement, is around 400 per year.


written by Sooyoun Park, Suhyun Han, Yongkyung Lee

sypark·shhan·yklee@lawtimes.co.kr