Contractualism and the Death Penalty

It is a truism that there are erroneous convictions in criminal trials. Recent legal findings show that 3.3% to 5% of all convictions in capital rape-murder cases in the U.S. in the 1980s were erroneous convictions. Given this fact, what normative conclusions can be drawn? First, the article argues that a moderately revised version of Scanlon’s contractualism offers an attractive moral vision that is different from utilitarianism or other consequentialist theories, or from purely deontological theories. It then brings this version of Scanlonian contractualism to bear on the question of whether the death penalty, life imprisonment, long sentences, or shorter sentences can be justified, given that there is a non-negligible rate of erroneous conviction. Contractualism holds that a permissible act must be justifiable to everyone affected by it. Yet, given the non-negligible rate of erroneous conviction, it is unjustifiable to mete out the death penalty, because such a punishment is not justifiable to innocent murder convicts. It is further argued that life imprisonment will probably not be justified (unless lowering the sentence to a long sentence will drastically increase the murder rate). However, whether this line of argument could be further extended would depend on the impact of lowering sentences on communal security.

No matter how careful courts are, the possibility of perjured testimony, mistaken honest testimony, and human error remain all too real. We have no way of judging how many innocent persons have been executed, but we can be certain that there were some.

Thurgood Marshall 1
A guilty verdict may be mistaken in two different senses. A verdict is mistaken in the first sense if it does not logically follow from the evidence. In the second sense a verdict is mistaken, notwithstanding that it may logically be supported by the evidence, when it fails to conform with the facts … . To avoid conviction of the innocent safeguards have to be created against both kinds of mistakes. By taking great care we could, in principle, eliminate errors of reasoning. Several rules of evidence single out certain types of evidence for special treatment with a view to minimizing the risk of such errors. The corroboration and hearsay rules, for instance, are designed to prevent unwarranted reliance on unreliable testimony. However, no matter what we do we cannot completely eliminate mistakes of fact. This is an inescapable feature of inductive reasoning: inferences can only be reached as a matter of probability and not as a matter of certainty.

A. A. S. Zuckerman 2
It is a truism that there are erroneous convictions in criminal trials. Yet the extent to which they occur and the reason why they do are not well known. It is also not clear what normative conclusions we could reasonably draw from the premise that erroneous convictions happen. This article focuses on these issues. I shall argue that, based on the existence of erroneous convictions alone, the death penalty should be banned because this form of punishment cannot justifiably be meted out to innocent individuals erroneously convicted as a result of the fallibility of a criminal trial, even if we may never find out who they are.
Recent legal findings show that 3.3% to 5% of all convictions in capital rape-murder cases from 1982 through 1989 in the United States were erroneous convictions. 3 In section II (1)-(5), I shall explain why a certain number of erroneous convictions in criminal cases are inevitable. The presence of mistaken convictions supplies a strong ground for finding capital punishment morally impermissible. I shall argue that this ground can be explicated especially effectively in terms of a moderately revised formulation of Scanlon's contractualism. 4 Contractualism holds that an act is permissible if and only if it is justifiable to everyone affected by it. 5 It does not seem justifiable for the court to mete out the death penalty to individuals convicted of capital crimes by mistake. A contractualist will disagree with utilitarians who hold that the death penalty is justifiable despite the existence of erroneous convictions. The most thorough contractualist rejection of this utilitarian position is to reject utilitarianism itself. These issues will be discussed in sections III to VII.
In sections VIII to X, I consider the normative implication of the pervasiveness of erroneous convictions. In section VIII, I argue that a contractualist account would reject the death penalty as a sentence for murder convicts. In section IX, I argue that, based on contractualism, there is a reasonably good argument for rejecting even life imprisonment for the offense of murder. I end with the conclusion that in order to ascertain the permitted punishments for murder, given that there are erroneous convictions, we would need more empirical data.
Besides this argument from mistaken convictions, there are several other well-known arguments for banning the death penalty. I shall not comment on any of them here, except to note that the contractualist elaboration of the argument from mistakes is an independent argument. quietly broke into Thompson's apartment and raped her at knifepoint. During her ordeal, Thompson tried not to provoke her assailant. Instead, she concentrated on memorizing the face of the rapist, in the hope that she might one day identify him and bring him to justice. In the dim light from the street, she studied this face for half an hour, and tried to remember its every detail, even his mustache and scar. Thompson then tricked him by offering to make him a drink. She desperately ran out from the back door, naked. The rapist raped a second woman later on the same night.
At the police station, Thompson sketched out the face of the rapist. The computer then matched the sketch with the face of a man called Ronald Cotton, who had a record of a minor sexual offense, and (according to a criminal detective) had dated white girls. At the lineup, or identification parade, Thompson picked out Cotton. The detective told her that she had done well.
At the trial, Cotton denied that he had raped Thompson. He and his family gave an alibi defense. His lawyer cross-examined Thompson, and suggested that she had made a mistake. She replied: "How dare you doubt me?" Given that she had studied her assailant's face for half an hour and, above all, had no reason to lie, her evidence was very persuasive for the jury. Cotton was convicted of having raped Thompson and another woman, Mary Reynolds, who was believed to have been raped by the same man.
The real rapist, Bobby Poole, was later imprisoned for another offense. Poole told other prison inmates that he had committed the crime for which Cotton was convicted. Cotton appealed on the grounds that Poole (who looked somewhat like Cotton) was the real culprit because of what Poole had told the other inmates. There was a retrial, and a voir dire 7 over whether the evidence provided by the inmates was admissible or not. Poole denied that he had raped Thompson. Thompson also could not recognize Poole. Consequently, the purported evidence was ruled inadmissible, and Cotton was convicted of rape at the retrial. Eventually, after serving 11 years in prison, Cotton was freed. There was a fragment of sperm in the rape kit, and the sperm's DNA matched Poole's. Poole confessed when the police arrested him.

II. How Evidence Can Be Unreliable
(1) Eyewitness Evidence This is one of the many cases built on eyewitness evidence, which is often unreliable. 8 The police's computer matched Thompson's sketch to the likeliest-looking individual in its database of persons with a criminal record. Because Poole had no criminal record and was therefore not on the computer's database, he was not matched. Instead, the computer picked Cotton, who had a criminal record and most closely resembled the face sketched by Thompson.
When the detective inappropriately said "Well done!" to Thompson at the lineup, her confidence that she had picked the right man soared. Thus she fixed her attention on Cotton, and consequently she could Hon-Lam Li not recognize the real culprit, Poole, when she was asked about Poole at the retrial.
In this case, Thompson made an honest mistake. Although the legal system may do a reasonably good job in distinguishing those who tell the truth from those who lie, it does not do well with a witness who has made an honest mistake. 9 The jury had no reason to disbelieve Thompson when she pointed her finger at Cotton. 10 The case of Cotton involved rape rather than murder. But there have also been numerous murder cases that were overturned because of DNA evidence. 11 We should realize, however, that not every case has DNA evidence. Many do not. (

2) Mistakes in Criminal Convictions
The jury's verdict is based on the evidence presented in court. Whether or not it is a guilty verdict, there is a risk of its being mistaken. Such a risk stems from different stages of the trial, when the judge rules that a piece of evidence is admissible or not, 12 when the jury decides whether a piece of evidence is credible or not, and finally when the jury decides whether the defendant is guilty or not. Even if the criminal procedure is followed flawlessly, a mistaken verdict can still result because there is an epistemic gap between the evidence and the verdict.
Just about any trial is a piece of guesswork. If we are lucky-viz. if the police, lawyers, and judges are competent-we get a piece of educated guesswork, but guesswork nevertheless, and there might still be a mistaken verdict. Because of the epistemic gap between the evidence and the verdict, it is possible in almost every criminal trial that the verdict is mistaken.
Do I deny that a jury ever knows that a defendant is guilty? This would depend on what we mean by "know." If by "know" we mean that one knows that p if and only if (a) one is justified to believe that p and (b) it happens that p, then a jury sometimes knows that a defendant is guilty of an offense, but this account is dubitable. 13 If, however, by "know" we mean that one knows that p if and only if one knows p with certainty, 14 then the jury does not know that any defendant is guilty of any offense because it cannot be certain that any defendant has committed murder. 15 Even if the prosecution has proved beyond any reasonable doubt that a defendant has committed an offense, it does not follow that we know that the defendant has committed the offense. 16 There are ways in which the jury would be invited to convict if one or more of the various possibilities obtain. 17 First, the defendant might be said to be lying if his story is unlikely, but just because his story is unlikely, it is not necessarily untrue. Some events that have taken place have indeed been extremely unlikely. 18 Second, even if a defendant lies in court, it does not follow that he has committed murder. 19 Third, there are risks in a conviction if the only evidence presented by the prosecution is circumstantial evidence. 20 Even though circumstantial evidence (such as DNA) can be very strong in casting doubt on a charge, it is often quite inconclusive as evidence against the defendant (see subsection [5] below).
Fourth, recent studies show that evidence offered by direct witnesses is often unreliable, not because such witnesses are dishonest (though they Contractualism and the Death Penalty sometimes are), but because it is difficult to determine the identity of a stranger in a fleeting moment. This is especially so in cases of crossracial identification. Further, because our memory is fragile, it is easily influenced by suggestions. In addition, Gary Wells points out that if the real culprit is not present at the lineup, a witness will tend to pick the "next-best" substitute. 21 (3) Trial as Guesswork A trial is very different from a laboratory test, which is generally reliable. The jury's conviction depends on an inference from available evidence to the conclusion that the defendant is guilty. There are several types of fallibility. First, there is human fallibility as, for example, when an eyewitness says that she saw the defendant. Another type is fallibility of the defendant's mental state, which I will discuss in subsection (5). Third, there is the fallibility of the trial process. Even if criminal procedure is flawlessly followed, mistaken verdicts can still result.
(4) Possibility of Mistake in a Murder Trial A murder trial is more prone to miscarriage of justice than non-capital cases for the following reasons. 22 In most trials, the victim's evidence in court is often crucial. 23 Even though a trial is bound up with guesswork, the victim's evidence can throw light on the motive of the assailant, why it happened, and how it happened. When she testifies, her evidence upon cross-examination by the defendant's lawyer might show whether she was telling the truth, inadvertently making mistakes, or simply fabricating a story. In a murder trial, however, the victim is necessarily unable to give evidence in court, and hence the valuable evidence from the victim is necessarily unavailable to the jury. Because this is so, much greater reliance is usually placed on eyewitness evidence of other witnesses (which is often unreliable), or "circumstantial evidence" (which is often inconclusive), or both.
Because the victim necessarily cannot give evidence in a murder trial, the amount of guesswork to be done by the jury greatly increases. So do the risks of misjudgment and the risks of miscarriage of justice. In the 1984 case of State of North Carolina v. Darryl Eugene Hunt, a 19-year-old black man, Darryl Hunt, was charged and subsequently convicted of rape and murder of a white woman in North Carolina. 24 Even though there was scant physical evidence against him, several eyewitnesses for the prosecution gave evidence in court that they saw Hunt on top of the victim. After serving a prison term for 19 years, Hunt was exonerated because his lawyer was able to trace the DNA evidence to the real culprit. 25 The real culprit was arrested and confessed. This case shows that eyewitness evidence could be completely unreliable. It also shows that the prosecution can be utterly unreasonable when it is unwilling to back down, even after it has been shown that the DNA of the culprit does not match the defendant's. 26 (5) "Isn't there 'Hard Evidence' in Some Cases?" In 2011, Jared Loughner attempted to assassinate congresswoman Gabrielle

Hon-Lam Li
Giffords. Many people saw him shoot Mrs. Giffords and he killed six other people. Even in this case, however, there is a reasonable doubt whether Loughner was really guilty, because to commit a serious crime, one must not only have committed the act, but must also have the necessary mental capacity. 27 It looks as though Loughner might have been rightly judged insane. 28 Apart from the errors of physical evidence (such as arresting the wrong person), there is another type of error that can easily creep into the deliberation of the verdict. This is the error regarding the mental state of the defendant. In order for anyone to commit a serious crime, he must have the necessary mens rea (guilty intention) as well as having committed the actus reus (guilty act). As Charles Black points out, even if the police have the right person, questions arise about the mental state of the accused. Did he kill in self-defense, carelessly, recklessly, intentionally, or in premeditation? Was the killing an accident? Was he insane, provoked, or involuntarily intoxicated? Black persuasively argues that the jury has much room for discretion and for guessing the correct answer, and consequently also has much room for error: It is very different when one comes to the question, "Was the action of which the defendant was found guilty performed in such a manner as to evidence an 'abandoned and malignant heart'?" (This phrase figures importantly in homicide law.) This question has the same grammatical form as a clearcut factual question; actually, through a considerable part of its range, it is not all clear what it means. It sets up, in this range, not a standard but a pseudo-standard. One cannot, strictly speaking, be mistaken in answering it, at least within a considerable range, because to be mistaken is to be on the wrong side of a line, and there is no real line here. But that, in turn, means that the "test" may often be no test at all, but merely an invitation to arbitrariness and passion, or even to the influence of dark unconscious factors. 29 This is an even easier kind of error, because the jury is entrusted with the power to exercise discretion, which would be very difficult to exercise in a principled way. 30

III. Miscarriages of Justice: Erroneous Convictions
Some hard data is available. Of the people executed in Europe from 1843 to 1943, it was later discovered that 27 cases involved innocent people. 31 Hugo Bedau and Michael Radelet conclude that of 7,000 persons who were executed in the United States between 1900 and 1985, 23 were later found to be innocent of capital crimes. 32 These numbers -27 and 23 represent the tip of a large iceberg. The number of innocent people who were executed was likely much greater because only the "lucky ones" had their convictions overturned after their executions.
In a recent study, D. Michael Risinger shows that in order to obtain the true percentage of erroneous convictions in all serious cases, we must use not only the number of erroneous convictions as the numerator, but more importantly must also obtain the relevant denominator. 33 Risinger focuses on rapemurder cases where there was a Contractualism and the Death Penalty request for DNA evidence, and such evidence existed. He estimates that the rate of erroneous conviction in capital rape-murder cases in the U.S. between 1982 and 1989 based on these data was between 3.3% and 5%. 34 Contrary to Justice Scalia's and Joshua Marquis's view that the error rate in felony cases is extremely low -27 erroneous convictions out of 100,000 convictions 35 -Risinger shows that this complacency is totally unfounded. So what is the normative implication from the fact that the rate of erroneous conviction in capital rape-murder cases in the U.S. in the 1980s was 3.3% to 5%?
What might proponents of capital punishment say? They reply in one of the following ways: first, Assistant Attorney General Stephen Markman and his assistant Paul Cassell challenge the validity of the earlier study conducted by Bedau and Radelet on the grounds that it was too subjective and inconclusive. 36 Contrary to Markman and Cassell's view, however, Risinger's study shows that 3.3% to 5% of all 319 capital rapemurder convictions (from 1982 to 1989) were erroneous, and that there is no reason to think that the rate of error would be very different in those cases in which no DNA evidence is available.
Second, proponents of capital punishment argue that even if the Radelet-Bedau study is valid, capital punishment has social benefits that cannot be shown to be unjustified. Arguing in the same way as Ernest van den Haag, they stress that the benefits outweigh the costs. 37 By benefits, they refer especially to the "incapacitation" of the criminals and "deterrence," as well as "just desert." 38 Further, fully aware of the possibility of convictions by mistake, American lawyer and politician Rex L. Carter holds the opinion that such occurrences-like the killing of innocent people in a war-are a necessary evil. 39 Surely, a utilitarian would hold that the sentencing policy causing the loss of some people's lives can be justified on the grounds that other people may benefit from the policy, viz. from the additional deterrent effect of execution over and above life imprisonment. This view suffers from two defects. The first one is an empirical difficulty. Issac Elrich's econometric paper notwithstanding, 40 empirical studies have shown that while efficiency in catching the culprits will considerably increase the deterrent effect of a punishment, the additional deterrent effect of capital punishment over life imprisonment cannot be measured. As Roger Hood observes, "econometric analyses have not provided evidence from which it would be prudent to infer that capital punishment has any marginally greater deterrent effect than alternative penalties" and that "it is futile therefore for such states to retain capital punishment on the grounds that it is justified as a deterrent measure of unique effectiveness." 41 The second problem is a moral one. Even if the death penalty had greater deterrent effect over life imprisonment, could it be justified? Some proponents of the death penalty-such as van den Haagargue that utility can be traded across lives, as long as utility is maximized. Similarly, on this view, if the death penalty can save lives by deterring killers, the convictions Hon-Lam Li and execution of innocent people by mistake can in principle be justified. I shall show in sections IV-VII that this view is flawed.
IV. The Normative Implication of Erroneous Convictions: Utilitarianism v. Contractualism In order to arrive at a reasonable position on the relevance of erroneous convictions to the kind of punishment that a court should mete out, we need the most plausible ethical theory. To see that utilitarianism is not that theory, consider the question whether it is permissible for the court to mete out a "cruel and unusual" punishment (such as dismemberment), provided that utility will be maximized. Most of us would reject this form of punishment, even if utility could be maximized. 42 Another form of punishment is vicarious punishment. If a criminal commits a murder, the punishment would fall not only upon the murderer himself, but also upon his parents, spouse, children, teachers, and friends, who had nothing to do with the crime. A criminal might not care about his own life, it is argued, but he might care about the lives of his close relatives and friends. Suppose that vicarious punishment occasions extra deterrent effect over non-vicarious punishment, and suppose further that this extra deterrent effect outweighs the side effect of insecurity feared by everyone. Even so, most of us would reject vicarious punishment as being inherently unfair. Even if a "cruel and unusual punishment," or alternatively a vicarious punishment, or a (combined) "cruel and unusual" vicarious punishment, were to achieve greater deterrent effect and (as a result) a promotion of utility, 43 these forms of punishment are morally unacceptable. One might ask for an explanation as to why this is so. I think the best explanation is a contractualist one. 44 Put in terms of justification, contractualism holds that an act, policy, or law is morally permissible if and only if it is justifiable to everyone affected by it. 45 However, "cruel and unusual" punishment, as well as vicarious punishment, cannot be justified to those who suffer them. This is so because, first, the offenders do not deserve "cruel and unusual" punishment, since such punishment would exceed what they deserve. I have argued elsewhere but can here only state that the plausible idea of negative desert (or negative retributivism) would put a cap on what someone would deserve. 46 Second, the relatives and friends of offenders simply do not deserve vicarious punishment at all. Again, this conclusion follows from the idea of negative desert that one who has not committed any crime does not deserve to be punished. 47 Although I do not accept the idea of positive desert (or positive retributivism)that an individual who has committed a crime deserves to be punished-which is certainly too large a topic to be taken up here, I accept the idea of negative desert, because one can accept the idea of negative desert without accepting the idea of positive desert. 48

V. Contractualism: Some Preliminary Thoughts
The best way to understand contractualism (as expounded by T. M. Scanlon) is to see it as an account that seeks to replace utilitarianism. 49 Normative utilitarianism (a first-order theory) entails various moral implications that are difficult to accept. For instance, normative utilitarianism seems to entail that, faced with the choice of saving a life and alleviating people's headaches, but not both, we must choose to alleviate headaches if they are sufficiently numerous. Most people would find this conclusion repellent. 50 A contractualist believes that it is a mistake to maximize utility by aggregating relatively trivial utilities to outweigh a significant claim, such as a life.
Consider a real-life example. Any educational establishment in Hong Kong-and I take it that the same situation would hold in Western countries as well-must be accessible to students with various kinds of disabilities, unless this would cause a school "unjustifiable hardship." 51 Suppose that in a secondary school, out of 1,000 students there is one student with a disability who needs to use a wheelchair, or alternatively needs access to psychological counseling. Accommodating this student would cost the school resources that could have been used for upgrading the school library instead. Suppose further that not admitting this student would cause her not to be enrolled in any school at all. A utilitarian would say that, depending on how the utilities work out, not admitting such a student could be permissible, perhaps even mandatory, if acting according to this decision would maximize utility. A utilitarian would care only about the collective utility, and the plight of the disabled student might be outweighed by the gain of other students (viz. a better library). However, it seems that in the context of today's society, not admitting this student would violate her right to education. 52 If this is the correct conclusion, what is the justification for this view?
A contractualist will think that ruling such a student out would be unjustifiable to her, on the grounds that the brute bad luck of the disabled student is undeserved and that ethical considerations would require the community to redress the situation as much as is practicable, even if doing so does not maximize utility. This approach would provide children with disability a level playing field. It is not clear what a utilitarian account might be, other than insisting that the correct decision is sometimes counter-intuitive, and that rejecting this student could be the right thing to do.
Finally, as I argued above, various kinds of punishment, such as "cruel and unusual" punishment and vicarious punishment, even if they might maximize utility (viz. by deterring crimes), are unacceptable. Again, contractualism can account for the rejection of these punishments on the grounds that they would not be justifiable to those who would receive them. It is, however, not clear what a utilitarian could plausibly say in reply.
As should now be obvious, the differences between utilitarianism and contractualism can be Hon-Lam Li understood as a difference in attitude toward aggregation. Can relatively small utilities be aggregated to outweigh a major claim? Utilitarianism says yes, but contractualism says no.

VI. Contractualism: The Theory
Despite its many counter-intuitive implications, normative utilitarianism still has many supporters. This is so, Scanlon observes, because of the intuitive appeal of the foundation of normative utilitarianism, namely philosophical utilitarianism. Philosophical utilitarianism is a meta-ethical, or second-order, thesis about the subject matter of morality, namely that the only fundamental moral facts are facts about individual wellbeing. In order to undermine normative utilitarianism, Scanlon argues, we must sap the force of philosophical utilitarianism by offering a more plausible alternative meta-ethical theory. Scanlon's contractualism provides such an alternative. As a metaethical theory, it informs us about what moral wrongness or impermissibility consists in. 53 Yet contractualism also has normative implications. How is contractualism supposed to work? Suppose we are interested in finding out whether an act, A, if performed under the given circumstances is morally permissible or not. To begin with, we need to find out whether there exists any principle which no one could reasonably reject, and which would disallow A. If there is a non-rejectable principle that would disallow A under the given circumstances, A would be impermissible. 54 In going through various principles, how do we know whether a principle, P, is reasonably rejectable or not? According to Scanlon, a principle is the "general conclusion about the status of various kinds of reasons for action." 55 Not all kinds of reasons have weight. The reasons must, firstly, be "generic reasons," or reasons based on generally available information of what individuals would want in certain situations, characterized in general terms, rather than detailed and particular information (which we do not usually know). 56 This is so because we want to be able to settle questions of right and wrong in the abstract, since the particular identity of the individuals affected is irrelevant. 57 Secondly, the generic reasons must be "personal (generic) reasons," which "have to do with the claims and status of individuals in certain positions." 58 Impersonal reasonssuch as reasons having to do with trees, natural wonders, non-human animals as long as the well-being of individuals is not affected-do not concern the claims of individuals and fall outside the scope of contractualism, or the realm of "what we owe to each other." 59 To appeal to individuals' personal (generic) reasons, we would need to know, first, how P would affect different individuals' well-being in the broad sense. 60 Scanlon's way of understanding well-being, however, departs from a consequentialist or welfarist model. For one thing, "reasons for being concerned with how one is 'affected' by a principle are not limited to reasons having to do with one's welfare," such as one's wanting to be able to "determine by Contractualism and the Death Penalty one's choice whether some result is likely to occur." 61 For another, there are reasons arising from the harmful effect caused if a certain type of action is generally permitted. It might be that, as in a case of environmental damage, "no individual action on its own causes a harm to which anyone could object," but that such an action is still impermissible because of the consequences should it be generally permitted and performed. 62 Second, a different basis on which individuals could object to a principle is that it is unfair, unjust, or rigged. 63 Thus, principles should be set aside if they violate valid considerations of fairness, justice, or perhaps even desert, as well as other deontological moral principles.
According to contractualism, an act, policy, or law is permissible if and only if it could be justified to everyone affected by it. Contractualism offers an interpretation as to what it is to count as a justification. Treating someone as a mere means is ipso facto not to justify the act, policy, or law to her. Contractualism demands that the justification be based on grounds that no one could reasonably reject and that maximization of utility cannot (always) be counted as a justification. Contractualism affirms the importance of consequences but denies that consequences are the only relevant considerations. Considerations such as fairness, justice, and negative desert are also relevant and important. 64

VII. Scanlon v. Harsanyi
For a contractualist, justification is built upon an understanding of impartiality. John Harsanyi is a utilitarian who purports to establish average utilitarianism on the basis of contractualism. 65 One important difference between John Harsanyi, on the one hand, and T. M. Scanlon, on the other, lies in their different understandings of the idea of impartiality. In trying to obtain normative principles, Harsanyi needs to adopt some interpretation of ethical impartiality. He interprets this condition as the one in which all parties (behind the veil of ignorance where they do not know their identity) would have an equal chance of being anyone in society. 66 As Scanlon points out, the supposition that we could obtain normative principles from self-interested parties trying to maximize their self-interest under conditions of ignorance involves a covert transition from the plausible contractualist idea that we must be able to justify our acts to everyone concerned, to the problematic idea that what is justifiable to everyone concerned is what would maximize an individual's self-interest behind the "veil of ignorance." 67 Why is this supposition problematic? With reference to Harsanyi, the short answer is that we cannot justify the situation to those who lose out in society ("the losers") on the grounds that average utility is high. Scanlon rightly points out that the fact that one option promotes higher average utility than another one does not settle the matter as to which option is morally justified. 68 A more detailed answer is that the kind of impartiality required by contractualism must ensure that a course of action be justified to every individual affected, regardless of their identity.

But Harsanyi interprets this condition
Hon-Lam Li as the one under which all parties would have an equal chance of being anyone in society. This interpretation, Scanlon rightly points out, is mistaken. For the losers can reasonably reject any principle that purports to justify their much worse condition by reference to the fact that others are better off (and hence average utility is higher). Scanlon thinks that the problem lies in the possibility that a much worse condition suffered by a few can be justified by the fact that many people enjoy a somewhat higher level of well-being. He believes that his contractualism can avoid this problem of aggregation, and that the origin of the problem is the supposition discussed above. 69 I am in total agreement with Scanlon here.

VIII. The Death Penalty
Without Harsanyi's version of the "veil of ignorance," average utilitarianism would be undermined. Although both a utilitarian and a contractualist should consider consequences as morally significant, only a contractualist can appropriately value fairness, justice, and negative desert. Even if "cruel and unusual" punishment and vicarious punishment could maximize deterrent effect and (under some circumstances) utility, a contractualist would reject these forms of punishment as impermissible because they are disallowed under the circumstances by the non-rejectable principle that a criminal should not receive a sentence in excess of his desert.
In a similar way, the death penalty cannot be justified to those who are executed but are actually innocent. 70 For execution consequent upon an erroneous conviction is disallowed by the principle that an individual does not deserve to be convicted and executed unless he has committed murder. 71 I shall refine this line of argument in subsection (3) below.
(1) Why Mistakes in Convictions Might Be Tolerated When legislators enact a law to criminalize a particular type of act, they do so in the knowledge that some innocent persons will be erroneously convicted as a result of the fallibility of the trial process. Even if every legal procedure were followed flawlessly according to due process, there would be no sufficient guarantee that no innocent persons would be erroneously convicted, since erroneous convictions will occur even if the best criminal procedure is flawlessly followed.
Even though participants in a democratic polity agree that all defendants on trial should be dealt with fairly and according to due process, they would not agree that no defendant may be convicted unless there is no doubt whatsoever, because this would require certainty, and the only way in which certainty can be achieved is to convict no one. But the consequences of convicting no one would likely undermine law and order, and consequently cause chaos in society. The participants agree, instead, that no defendant may be convicted unless it is beyond any reasonable doubt that he or she is guilty of the offense. Convicting defendants according to the "beyond any reasonable doubt" rule, however, does not mean that there will be no erroneous convictions.

Contractualism and the Death Penalty
Once we allow convictions in criminal trials-albeit under the "reasonable doubt" rule-we allow for the possibility of mistaken convictions. How can we justify or tolerate the possibility of mistaken convictions?
The most plausible answer is that the possibility of mistaken convictions is tolerated because we must enforce the laws in order to secure communal safety. Hence, mistaken convictions are seen as necessary evils. So far, this reasoning is not necessarily a utilitarian one. It is compatible with contractualism. However, some utilitarians might even use this position as a ground for arguing that this conclusion could be reached only via a utilitarian theory of deterrence. 72 The utilitarian move, I think, should be resisted.
Although security and deterrent effect are important factors for the court to consider before meting out certain punishments and for our tolerating the existence of mistaken convictions, it does not follow that there exist no limitations on the kind of punishment to be meted out for a given offense. Nor does it follow that any punishment consequent upon erroneous convictions can be tolerated, come what may. For one thing, even though we value punishment for its deterrent effect, we do not thereby become consequentialists or utilitarians, just as someone who buys a cellphone for its good consequences does not thereby become a consequentialist or utilitarian. To be a consequentialist or a utilitarian, we must also hold that morality is concerned with nothing else than consequences or utility. Certainly, we can value both deterrent effect and fairness, because doing so does not make us consequentialists or utilitarians. In subsection (3) below, I shall argue that a contractualist should reject the death penalty. In section IX, I argue that whether life imprisonment should be reduced to a shorter sentence will depend on the number of innocent murder convicts affected, as well as whether reducing life imprisonment to a shorter sentence will increase the number of murders.
(2) The Individualist Restriction and Pairwise Comparison In order to appreciate how Scanlon's contractualism works, we must deepen our understanding of his theory. As I pointed out in section VI, whether an act is justifiable, Scanlon argues, depends on the generic personal reasons raised by those affected by it. Whether a principle is justified or rejectable "depends only on various individuals' reasons for objecting to that principle and alternatives to it," though such reasons might involve an individual's well-being, or deontological considerations, such as fairness, or both. 73 Scanlon holds that "all objections to a principle must be raised by individuals," and that this feature of admitting only personal reasons "allows the intuitively compelling complaints of those who are severely burdened to be heard, while on the other side, the sum of the smaller benefits to others has no justificatory weight, since there is no individual who enjoys these benefits and would have to forgo them if the policy were disallowed." 74 Although what follows from the observation that "there is no individual who enjoys these benefits" is not, I think, entirely clear, Scanlon thinks that this fact provides grounds for holding that individuals' competing claims have to be judged against each other by pairwise Hon-Lam Li comparison. 75 His contractualist framework places the restriction that an individual can only put forward complaints or objections that concern only herself, whether or not such complaints or objections have to do only with her well-being, or fairness, or both. Derek Parfit labels this restriction the "Individualist Restriction." 76 The point of comparing individuals' claims in pairs is to avoid aggregating lesser claims held by a large number of people, which might override a weighty claim held by one person. Thus, if we can save one of two groups of strangers, but not both, we are to compare the weightiest claim in each group, and determine the priority of which group to save based on such a comparison. If either of these claims is weightier, then the group to which the weightier claim belongs will have the priority. In this event, the number of people in each group can be ignored. 77 Thus, on Scanlon's view, a single individual's life always outweighs the headaches of a huge number of people, since saving a life is more important than alleviating a headache and consequently the number of people suffering a headache never becomes relevant in the determination of priority. The point of the Individualist Restriction is to block groups from aggregating the importance of their claims.
I myself would reject the Individualist Restriction. Suppose we can save either one stranger's life, or alternatively save 200 other strangers from "near death," a condition almost as bad as death. We should save the 200 strangers from near death, but the Individualist Restriction would disallow us from doing so. 78 Given this implausible implication, we should reject the Individualist Restriction. If we reject the Individualist Restriction, would contractualism collapse into consequentialism or utilitarianism? The answer is no. Let me explain.
Scanlon's contractualism is to a great extent compatible with virtue ethics in that both theories rely on case-based moral reasoning, or casuistry. To this extent, both value the importance of practical wisdom (phronesis) and reject algorithm. However, Scanlon's Individualist Restriction-like the maximin principle and Rawls's difference principle -provides an algorithmic formula for decision-making involving numbers. Although, like these two principles, the Individualist Restriction initially seems plausible, all of them fail to withstand counter-examples. 79 As I see it, we can accept Scanlonian contractualism without the Individualist Restriction, and consider cases involving consequences on a case-by-case basis. For instance, we can reject any move to aggregate small utilities for outweighing a much weightier claim, and yet accept that a claim could be outweighed by marginally less weighty claims that are sufficiently numerous. I will now label the view that I favor, which rejects the Individualist Restriction, as "Scanlonian contractualism" in order to distinguish it from "Scanlon's contractualism," which embraces the restriction. 80 (3) What Follows from Scanlonian Contractualism? According to Scanlon's contractualism, the claim put forward by the innocent person convicted of murder by mistake (the "loser") would outweigh the claims held by other people. Since Scanlon does not take Contractualism and the Death Penalty into account the number of people who have a similar claim (e.g., those who place a lot of weight on deterrent effect), 81 the end result is a pairwise comparison between (a) the strongest personal generic reason on the one side (namely innocent murder convicts) to reject the death penalty, and (b) the strongest personal generic reason on the other, competing side (namely those who demand that a convicted criminal be severely punished). The personal generic reason of innocent murder convicts to reject the death penalty is grounded in the premise that being convicted of murder by mistake and then executed is possibly the worst kind of brute bad luck one could have. The strongest personal generic reason of those who demand murder convicts to be executed would likely be based on general deterrence. 82 Although Scanlonian contractualism (without the Individualist Restriction) is different from Scanlon's contractualism, both versions agree on this case. For Scanlonian contractualism, the plight of someone falsely convicted of murder and then executed is much worse than the plight of any other member in the community-even someone who stood to be murdered. 83 More important, if empirical research on deterrent effect is any guide, the reduction of the death penalty to life imprisonment would involve no loss of security. 84 If this is so, no one in the community could have any personal generic reason comparable in importance to that of innocent murder convicts. 85 Scanlonian contractualism departs from Scanlon's contractualism only when the competing claims being compared are sufficiently close, a condition that does not obtain in this case. So the conclusion reached here is that the death penalty should be banned because it is not justified to murder convicts who are actually innocent.
(4) Is it Ever Justifiable to Convict, Given Mistakes in Convictions? Is it ever justifiable to convict, if the general rate of mistaken conviction is between 3.3% and 5%? 86 It would be easier to begin our consideration with convictions for lesser offenses. Suppose one in every 20 convictions for careless driving is mistaken. Suppose further that John is convicted of careless driving but is actually innocent because the police arrested the wrong driver. Let us consider two questions: (1) Do we think that John should not have been punished (say, by way of a fine)? (2) Do we think that because there are erroneous convictions among careless-driving cases, the fine should be reduced or even dispensed with?
Clearly, John should not have been punished at all (because of negative retributivism). A similar conclusion would be reached if the offense were murder instead of careless driving. I take it that we do not think that because of the existence of cases like John's, the fine for careless driving should be dispensed with or even reduced, because we can assume that penalties meted out to careless drivers are an effective way to deter careless driving and to help ensure road safety. 87 Yet it seems that if the case under consideration were murder, the death penalty should be reduced to a lesser punishment. Why?
There are several reasons. First, careless driving is a very minor Hon-Lam Li offense, which does not carry with it any social stigma. Murder is a very different matter. Second, the death penalty does not allow any meaningful recourse if relevant evidence surfaces after the falsely convicted offender is executed. The case of careless driving is quite different. Third, there is no clear empirical evidence to support the view that the death penalty, compared with life imprisonment, would make society safer. On the contrary, there is evidence that the deterrent effect of the death penalty is no greater than that of life imprisonment. 88 Finally, the most important, but perhaps less obvious, reason is that the falsely convicted offender has too much brute bad luck to bear compared with other people in the community. In Scanlonian terms, her claim (based on generic personal reasons) outweighs the strongest claim of those in the community. 89 Because the punishment is the death penalty, her claim to not be executed would outweigh those who demand that a defendant convicted of murder should be put to death. 90

IX. Life Imprisonment
What if a person convicted of murder were sentenced to life imprisonment, instead of facing the death penalty? How could the problem be resolved? Obviously, the argument from lack of recourse would be irrelevant. Despite this fact, a case could be made against life imprisonment because, despite the possibility of appeals, a number of innocent persons would still be erroneously convicted of murder and consequently imprisoned for life. For one thing, DNA evidence is not available in all capital cases. In fact, DNA evidence is available in only a small fraction of all capital and non-capital cases. For another, regarding all capital cases without DNA evidence, there is no reason to believe that the rate of erroneous conviction would be lower than the rate of 3.3% to 5% estimated by Risinger for capital rape-murder cases in the U.S. in the 1980s. 91 Can life imprisonment be justified? A person falsely convicted of murder has strong personal generic reasons to reject life imprisonment.
In order to take account of this person's claim, we need to weigh these reasons against the personal generic reasons put forward by others (including those who claim that murder convicts should be sentenced to life in prison or to death) in the community. Since personal reasons have to do with individual well-being (besides fairness), we should ask how individuals would be affected. Since I do not accept positive retributivism, their case must (on my view) be based on the hypothesis that a reduction of life imprisonment to a long sentence (e.g., 25 to 30 years) would have an impact on communal security. 92 Those who proffer such a view would need to hypothesize that, for instance, someone would be murdered as a result of the substitution of a long prison sentence (viz. 25 to 30 years) in place of life imprisonment. 93 I think we can respond as follows. First, Roger Hood's empirical research shows that the effectiveness of the police in arresting the perpetrators of crimes has a much greater marginal Contractualism and the Death Penalty deterrent effect than the marginal deterrent effect of increasing the sentence from life imprisonment to the death penalty, because the latter marginal deterrent effect could not be measured. 94 (Almost certainly, the effect of increasing welfare would have a much greater effect on lowering crime than an increase in punishment. 95 ) Given these facts, the hypothesis that reducing the sentence from life imprisonment to a long sentence would result in a considerable rise in the murder rate is, at best, unreasonably speculative.
Second, even if (say) a reduction of life-imprisonment to 25 or 30 years resulted in an increase in murder rate, we would still need to compare the personal generic reasons in favor of the reduction against the personal generic reasons of those who are against it. An innocent murder convict has generic personal reasons in favor of the reduction on the grounds that he would be adversely affected if there were no such reduction. Those who are against the reduction would consist of (1) those who would be murdered had there been such a reduction ("the potential victims"), (2) relatives and friends of a victim of murder who are outraged that the reduced sentence is unreasonably lenient, and (3) people in society who think that the reduced sentence is unreasonably lenient.
In considering the case against the reduction, a utilitarian would argue that the feelings of the relatives and friends of the victim, as well as those of people in society, also count and might be aggregated to outweigh the claims in favor of the reduction. A contractualist would argue that lesser claims could not in principle be aggregated to outweigh a major claim. Moreover, it is plausible that anyone's feelings (against the reduction) should not count independently of how reasonable the claim against the reduction is. So we should focus on the personal reasons of an innocent murder convict and those of a potential victim (who would be murdered as a result of such a reduction).
Suppose such a reduction would cause an extra murder in the community. Let us analyze the situation in two stages: (A) At the first stage, we would need to weigh the personal reasons of the potential victim against the reduction, on the one hand, and the personal reasons of an innocent murder convict in favor of the reduction on the other. How are we to compare these two claims or the personal generic reasons behind them? To be murdered is a bad thing, because one's life is cut short, or because one's projects, commitments, plans, and relationships are terminated without notice. Not all murders are equally bad, since most of us would prefer being killed instantly to being tortured until dead, in order that pain and suffering be minimized.
It is probably foolhardy to speak in general terms, since every case is unique. Nevertheless, this might be unavoidable if we are to theorize and compare these two cases at all. Imagine that you are convicted of murder by mistake and sentenced to life imprisonment. You suffer (1) a false conviction, and (2) life imprisonment. First, although your life is not cut short, you have lost your freedom for the rest of your life. 96 Your years behind bars without any possibility of release make you feel that you are hopelessly rotting away without anyone's caring or attention.

Hon-Lam Li
Second, the public blame and curse you for something you have not committed. Yet, besides the legal appeals that you have already exhausted, there is no further appeal for you in the sense that no one would believe that you are innocent and no one would understand the psychological pain you suffer. Moreover, what you are suffering is not only psychological and physical pain, but also an extreme injustice.
Finally, it is a truism that the three most common causes of murder have to do with (1) love and passion, (2) money and greed, and (3) revenge. This being the case, at least to some extent you could do something to avoid being murdered. For instance, you should treat an intimate relationship with respect, love, and honesty. You should be fair to your business partners. And you should avoid being provocative. However, as the cases of Ronald Cotton and Darryl Hunt show, there is nothing that you can do to avoid being dragged into a murder case with which you are not involved at all. Thus, you feel that you have no control over your life, and have no fair and adequate opportunity to avoid being dragged into such a case. 97 We would need to form a judgment as to whether a murder is worse than life imprisonment consequent upon a mistaken conviction, or not. There are three possibilities: a murder is (1) worse than, (2) comparable to, or (3) better than life imprisonment consequent upon a mistaken conviction. While I recognize the risks of generalization, I believe that life imprisonment consequent upon a false conviction is probably comparable to-if not worse than-a murder.
(B) At the next stage, we need to factor in how the number matters. According to Scanlon's own version of contractualism, we should make a pairwise comparison between (1) the personal generic reasons against the reduction of sentence, and (2) the personal generic reasons in favor of the reduction. On a first approximation, this comparison translates into weighing (a) how bad a murder is, and (b) how bad a life imprisonment consequent upon a false conviction is. Yet, if we accept Scanlonian contractualism instead, as I argued we should, we should take into account numbers if the claims being compared are comparable. In other words, we should take into account how reducing the sentence from life imprisonment to a lesser sentence (e.g., a long sentence) would raise the number of murders, as well as how many innocent murder convicts would be affected. We should then compare these two numbers. If the number of innocent murder convicts is substantially greater than the number of murders, then life imprisonment should be reduced to a long sentence (of 25 to 30 years). If, on the other hand, the number of murders far exceeds the number of innocent murder convicts, then life imprisonment should not be reduced to a long sentence. Unless those who urge that a long sentence (25 to 30 years) is too light for someone convicted of murder have empirical data to show that a reduction of life imprisonment to a long sentence would cause a considerable rise in murder cases, such a reduction seems to me justified. 98

X. Lesser Sentences
Can the punishment for murder be further reduced? Can a long sentence of 25 to 30 years be further reduced to 18 years (a moderately long sentence), or 10 to 15 years (a moderate sentence), 5 to 10 years (a moderately short sentence), or 3 to 5 years (a shorter sentence)? As we leave behind the death penalty, life imprisonment, and long sentences, and turn instead to the consideration of shorter sentences, the brute bad luck suffered by an innocent defendant becomes less and less onerous, but at the same time the community faces a greater and greater security risk (if lower sentences result in more crime). In the broadest terms, the burden and risk the community is facing include (1) the feelings of the family members and friends of the victim (who was murdered), (2) the sense of outrage in the community at large that the sentence is too light, and (3) most significantly, the effect of general deterrence and incapacitation. As I said, I do not think that the feelings or emotions of the community and of the victim's friends and family members should weigh much or at all in the equation, since these feelings should depend on whether the sentence to be meted out is reasonable. 99 Another reason against counting these feelings is the contractualist principle that lesser personal reasons-even if very numerous-should not be aggregated if the contrary personal reasons are much weightier. After we set aside these feelings, the effects due to deterrence and incapacitation would tend to diminish as we travel down the scale of sentences.
There are reasons for thinking that short sentences would be too light. First, if murderers were imprisoned for only three to five years, this would likely undermine the effects of deterrence and incapacitation, and the social order maintained by the rule of law would likely turn into chaos. 100 Another reason is that it would be difficult to differentiate the sentences for murder and those for armed robbery, rape, burglary, and theft. 101 The right sentence for a given offense should depend on the effect of incapacitation and deterrence as well as the kind of harm that innocent murder convicts would have to bear, with a cap on the maximum sentence (because anything that goes beyond the maximum would be unjustifiable). 102 In the case of murder, we have to weigh and compare the (marginal) suffering to be borne by a murder convict who is innocent, and the (marginal) harm posed to people if the sentence were reduced. If these kinds of suffering and harm are of different orders, then both Scanlon's contractualism and Scanlonian contractualism would agree to ignore the lesser harms. If, however, they are comparable, then even if the suffering borne by an innocent murder convict is more onerous, the number of people to bear the harm should be taken into account.
For example, if we have to choose between saving one life or alleviating numerous headaches, we should save the life. If, on the other hand, we have to choose between saving one life, or preventing 200 near deaths, we should prevent the near deaths. In

Hon-Lam Li
fact, if we can either save one life, or cure 10,000 cases of blindness, we should choose the latter course of action. This kind of reasoning under Scanlonian contractualism would reject the Individualist Restriction, but is not reduced to aggregational consequentialism. 103 My conclusion can only be imprecise. Based on Risinger's findings that 3.3% to 5% of all convictions in the U.S. in the 1980s were mistaken, we should seriously consider the following: moderate sentences (10 to 15 years), longer sentences (15 to 20 years), or long sentences (20 to 30 years). To determine the right length of sentence, we would need data from empirical studies. 104 It would be foolhardy of me to make a more precise proposal in the absence of such data.

XI. Concluding Remarks
The law with respect to careless driving helps ensure road safety and its repeal would undermine road safety. Similarly, the law with respect to murder helps protect communal security, and its repeal would make the community less safe. If someone falsely convicted of careless driving cannot reasonably reject the law imposing fines for the offense even though he can reasonably reject his own mistaken conviction, can someone falsely convicted of murder reasonably reject the law that imposes a sentence for murder (even though he himself can reject his own mistaken conviction)? 105 I think that no one could reasonably reject the criminal law and sentencing on murder, provided that the court takes appropriate account of the possibility that an innocent person could be convicted of murder by mistake.
Notice that there is no contradiction in holding both (1) that John can reasonably accept the criminal law and sentencing laws on murder and (2) that he can reasonably reject his own erroneous conviction. As a reasonable individual, John appreciates the rationale behind these laws. At the same time, he knows (from his own first-personal perspective) that he is innocent in the case in question. There is therefore an increase in knowledge shifting from (1) to (2). 106 The fact that there are mistaken convictions has significant implications for what is reasonably acceptable to people in the community. Scanlon's Individualist Restrictionlike the maximin principle-would allow the claim of the worst off ("the losers") to prevail over competing claims. Normally, those who would fare the worst in criminal cases are those who are erroneously convicted of murder, if the mandatory sentence for murder is the death penalty or life imprisonment.
Although Scanlon holds that we should consider the strongest personal generic reasons of competing parties only 107 and that we can ignore the numbers of people holding such claims, even when their significance is closer or more comparable or "relevant" 108 to each other, the implications of this view are highly implausible. That is why Scanlonian contractualism (without the Individual Restriction) is more plausible than Scanlon's contractualism. Thus, if we consider the Contractualism and the Death Penalty reduction of life imprisonment for murder to long sentences (25 to 30 years), moderate sentences (10 to 15 years), and eventually shorter sentences (3 to 5 years), there is no a priori reason to believe that an innocent murder convict could reasonably reject any or all of these sentences. 109 This is because, on the one hand, we need to take account of (1) the increased risk posed to communal security as a result of decrease in deterrence and incapacitation, as well as (2)  3 See Risinger, "Innocents Convicted," 768-80. In "Rate of False Conviction," Samuel Gross et al. present a "conservative" estimate that "the rate of erroneous conviction of innocent criminal defendants" sentenced to death is "at least 4.1%." Moreover, in Exonerations in the United States, Samuel Gross and Michael Shaffer point out that 8% of 873 recorded exonerations between 1989 and 2012 in the United States involved guilty pleas. Kent Roach also points out that a certain percentage of those who pleaded guilty in Canadian courts were in fact innocent. See "Wrongful Convictions in Canada," esp. 1475-6.
6 See State of North Carolina v. Ronald Cotton Jr., 318 N.C. 663, 351 S.E.2d 277 (1987). This case is narrated by the accused and the victim in Thompson-Cannino and Cotton, Picking Cotton. 7 A voir dire is a trial within a trial. Its purpose is to determine whether a piece of evidence is admissible, viz. whether it should be heard by the jury.

Hon-Lam Li
8 Especially unreliable is cross-racial identification. Of more than 230 cases exonerated (as of 2009) as a result of DNA evidence, more than 75% have to do with unreliable eyewitness evidence. See CBS News, Eyewitness Testimony.
9 See the interview with psychology professor Gary Wells in ibid.
10 Cotton's alibi evidence was discounted or even ignored, because this was a case in which a black man raped a white woman. Moreover, an accused is expected to lie in court, and hence an accused's word-if not corroborated with other evidence-is usually disbelieved. This is because an accused is naturally expected to say that he is innocent, whether or not he is guilty. For this reason, an accused used to have no right in English law to give evidence in a criminal trial, because it was thought that his evidence would be worthless. Although this prohibition was long ago discarded, the expectation that defendants will lie in an attempt to escape conviction remains common.
11 Through DNA testing the Innocence Project aims to help exonerate those convicted of serious crimes by mistake. See http://www.innocenceproject.org/ 12 A mistake could be made regarding a piece of evidence the admissibility of which is in question. In State of North Carolina v. Ronald Cotton Jr., Cotton's purported evidence that Bobby Poole admitted to other prison inmates that he had committed the rape was ruled inadmissible, although this was in fact the case.
13 See Gettier, "Is Justified True Belief Knowledge?" A crucial objection to this account is the Gettier Problem, according to which justified true belief does not necessarily constitute knowledge.
15 Very rarely, if ever, does anyone plead guilty to a murder charge. Gross and Shaffer, "Rate of False Conviction," as well as Roach, "Wrongful Convictions in Canada" show that in some cases innocent people plead guilty to criminal charges.
16 See Risinger's findings (to be discussed in section III below and esp. note 34) showing that 3.3 to 5% of all capital rape-murder "beyond any reasonable doubt" convictions in the U.S. from 1982 to 1989 were mistaken. I agree with Risinger in holding that although we do not know what the general rate of mistaken conviction is (viz. for all sorts of crimes), it is difficult to believe that such a rate would be much lower than 3.3% to 5%. See Risinger, "Innocents Convicted," 782-8. If this position is correct-if, say, 3% of all criminal convictions are mistaken-we can conclude that we do not know with certainty in any criminal conviction that the defendant must be guilty.
17 These include the scenarios where the defendant's story is unlikely; where the defendant is believed to have lied; where the defendant has changed his story; where circumstantial evidence suggests that the defendant met the victim, but the defendant denies it; where the defendant's demeanor in court is suspicious, such as when he appears to be nervous; where the prosecution has "circumstantial evidence" that leads to the conclusion that the defendant murdered the victim; and where the prosecution has evidence from a direct witness that the defendant murdered the victim.
18 See Nagel, View from Nowhere, chap. 11; Jeffrey, "Statistical Explanation vs. Statistical Inference"; Salmon, "Statistical Explanation"; and Mellor, "Probable Explanation." The fact that any one of us was born at all was unlikely to the extreme. Nevertheless, it is a fact that we were born. In any normal copulation, 450 million sperms compete for an ovum. It is widely believed that had a different sperm merged with a given ovum, a different individual would have emerged. The general point is that the extremely low probability of the occurrence of an event, E, does not entail that the credence of the proposition that E occurred must be low.
19 For one thing, the police officers were sometimes found to have lied with the intention of strengthening their case. One possible reason for lying on the defendant's part might be to enhance his case. He may wish to cover up an embarrassing fact; for example, that he had an affair. He may also have lied because he would wish to protect some people, such as the real culprit. 20 Circumstantial evidence is the sort of evidence that points to the defendant as the culprit by virtue of a number of coincidental factors. For instance, if the defendant denies that he was with the victim, whereas certain coincidental factors points toward the hypothesis that they did meet on or near the date of murder, suspicion would be cast on the defendant. However, highly unlikely events do take place.
21 See CBS News, Eyewitness Testimony. Our process of picking the right person is usually by elimination. If the real culprit is not at the lineup, the witness will usually pick the "next best" person, another reason why eyewitnesses can be unreliable.
22 In British law (except in Northern Ireland), a trial for non-capital crimes does not require a unanimous verdict in order to secure a conviction, although this is not so in federal courts in the United States. It is usually thought that in British law there is less chance of miscarriage of justice in a murder trial, compared with trials for noncapital crimes. However, I shall argue that, despite the requirement of unanimity, a murder trial is actually more prone to miscarriage of justice.
23 As the case of Ronald Cotton shows, even the victim can be mistaken about the identity of the real culprit. I am not saying that the victim is infallible; far from it. 26 The prosecution in State v. Hunt argued that even though the DNA in the rape kit did not match the defendant's, it was still possible that the defendant had committed murder. The Supreme Court of North Carolina as well as the federal Supreme Court accepted the prosecution's new position, and refused to acquit the defendant, until the defense lawyer was actually able to track down the real culprit. The fact that the courts allowed the prosecution to change its position amounts to requiring that the defendant prove his innocence beyond reasonable doubt.
27 Thus, neither a 6-year-old child nor a mentally insane person could possibly commit a criminal offense.
28 "Before the attempted assassination of Reagan, Harvard Law School professor Alan Dershowitz said in a telephone interview Monday, 'this would be a clear case of insanity, because the pre-meditation would not be seen as undercutting insanity, it would be part of demonstrating insanity.' But under the post-Hinckley rules, he said, 'that's a very uphill battle.' " CBS News, "Insanity Defense Difficult." The point I wish to make is that the possibility that Loughner was insane could not be ruled out. In this case, after a process of plea-bargaining, Loughner pleaded guilty to 19 counts and waived his right to the insanity defense in exchange for the prosecution's not seeking the death penalty. 29 Black,Capital Punishment,chap. 2, 30 In Steinbeck's novella, Of Mice and Men, Lennie is an autistic man who inadvertently killed Curley's wife. He did not have the mens rea to commit murder. However, had he been arrested and tried, he would likely have been convicted of murder.  (from 1982 to 1989) in which the convictions were later found to be erroneous by virtue of exculpatory DNA evidence. To be especially safe, he allows for the possibility that 5% of these defendants were actually guilty, and uses 10.5 as the numerator. The denominator should not be all death penalties imposed from 1982 to 1989, of which there were 2235. Instead, it should be all capital rapemurder cases in which there was a request for DNA evidence and such evidence was available. This number was estimated to be 319. This yields a percentage of 3.3%. This is, however, the minimum because there must have been cases in which post-conviction legal counsel might not have requested DNA evidence. Risinger estimates the maximum percentage of erroneous convictions to be 5%. Moreover, relying on a different method, Gross et Gross et al.] the benefit of the doubt: let's assume that he understated the number of innocents by roughly a factor of 10, that instead of 340 there were 4,000 people in prison who weren't involved in the crime in any way. During that same 15 years, there were more than 15 million felony convictions across the country. That would make the error rate .027 percentor, to put it another way, a success rate of 99.973 percent ("The Innocent and the Shammed"). " 36 See Markman and Cassell,"Protecting the Innocent," 122,126,157. 37 Ernest van den Haag is a well-known defender of the death penalty. See Haag, "Ultimate Punishment." 38 Ibid. Because I have already displayed skepticism about positive retributivism elsewhere, I shall not try to discuss the argument that executing murder convicts would contribute to just desert. See Li, "Contractualism and Punishment," sections 6, 8, and 9 as well as notes 86 and 96 therein. 39 See Carter, "Letter to the Editor": "Hugo Bedau had best come in out of the heat. As a defender of capital punishment, I have no problem in admitting innocent people can be executed and couldn't care less what happens to Gary Graham [who had admitted to armed robbery but denied murder]. He should have been executed for what he confessed to. There is a war going on in our own country-against crime and thugs like Graham. It is sad that innocent people get killed in war, but that is the way it is. Ask any wartime veteran. Try 'em, give 'em 90 days for appeal and then hang 'em slowly at noon on the courthouse lawn. Just maybe killers-to-be will get the message, just as Japan did when we dropped the A-bomb (cited in Radelet and Bedau, "The Execution of the Innocent," 123)." Even though William Blackstone said that "[i]t is better that ten guilty persons escape than that one innocent suffer" (Commentaries on the Laws of England, Book 4, 352) and Benjamin Franklin said that "it is better a hundred guilty persons should escape than one innocent person should suffer" ("Letter to Benjamin Vaughan. March 14, 1785″), it is arguable that their common unspoken assumption is that punishing guilty cases, if they are sufficiently numerous, can outweigh the moral badness of (unintentionally) punishing an innocent person.
40 See Ehrlich, "Deterrent Effect of Capital Punishment." 41 Hood, "Capital Punishment," 6. See also Tonry,"Sentencing in America,and Nagin,"Deterrence, Utility would be maximized if criminals would be so afraid of execution by dismemberment that they would be more deterred from committing certain crimes than if a less cruel form of punishment were in place instead, and if the pain incurred before and during dismemberment would be outweighed by the gain in terms of deterrent effect. 43 Let us suppose, for the sake of discussion, that the gain in deterrent effect outweighs the pain and suffering of the criminals (or, in the case of vicarious punishment, the pain and suffering of their closed relatives and friends).
44 One might oppose utilitarianism on various grounds. Besides contractualism, various forms of nonconsequentialism also reject utilitarianism. However, contractualism is not the same as nonconsequentialism. For one thing, as Scanlon points out, Robert Nozick's entitlement theory is a form of nonconsequentialism, but one that a contractualist would reject. For another, even though contractualism coheres with certain form of nonconsequentialism (e.g., Frances Kamm's nonconsequentialism), contractualism is a meta-ethical theory with normative implications, whereas nonconsequentialism is a normative theory.
45 See Scanlon, "Contractualism and Justification." Moreover, Scanlon's position is actually more complex. For him, citizens have the obligation to obey laws only if the state is legitimate. But a state cannot be legitimate if it grossly fails to take care of its citizens by failing to provide for education, welfare, and security. Nevertheless, even as a victim of a grossly unjust system, a citizen might still have the moral duty not to harm others, depending on whether her act to steal (say) could be justified on grounds of necessity, and whether her act would impose too high a cost on others. See also Scanlon, "Individual Morality," 13-19; and Shelby, "Justice, Deviance and the Dark Ghetto." For the present purpose, I am ignoring this complication. And I will expound on what counts as "justifiable" to individuals in section VI below.
46 See Li,"Contractualism and Punishment,. Let me clarify the relation between contractualism and the idea of negative desert. The idea of negative desert says that people cannot be justifiably punished for what they have not done, or for what is beyond their control. Within the broad framework of contractualism, this is a principle which no one can reasonably reject. (Of course, a non-contractualist can also accept the idea of negative desert.) I have argued in "Contractualism and Punishment" that, based on the importance of deterrence and negative desert, a contractualist can offer an attractive theory of punishment.
47 Ibid, 49 See Li, "What We Owe to Terminally Ill Patients," 238-9. Some of the ideas in this section are drawn from this paper of mine.
50 Can a utilitarian avoid the repellent conclusion? To do so, she might argue that an individual's life has infinite utility, which even an astronomical number of interpersonal headaches cannot outweigh. However, a life cannot have infinite utility unless (1) each minute of such a life has infinite utility, or alternatively (2)  57 See Scanlon, "Contractualism and Justification," 6. We use "generic reasons" because we want to settle "questions of right and wrong in the abstract, before we know which individuals, if any, we will actually interact in these ways." Note that this focus on abstract requirements does not mean that the circumstances under which an act takes place are morally irrelevant to the determination whether a principle is reasonably rejectable or not. 66 See ibid. Harsanyi used the concept before John Rawls, although the contents of their respective concepts are different and Harsanyi did not use the term "veil of ignorance." Contrary to Harsanyi, Rawls interprets this condition as the one under which self-interested parties are deprived of particular information, including any objective basis for estimating probabilities. Yet Scanlon has doubts about this very feature of Rawls' Original Position, which he finds "slightly puzzling" (Scanlon, What We Owe, 148). Rawls' theory can be found in Theory of Justice, esp. chap. 3. Rawls' objections to Harsany's view are presented in chap. 3, sections 27 and 28. Harsanyi criticizes Rawls' refusal to admit "subjective probability" or even "logical probabilities" in the original position. See Harsanyi, "Can the Maximin Principle," esp. 599. 67 Scanlon, "Contractualism and Utilitarianism," 145-6. This supposition is common to both Harsanyi and Rawls. Insofar as this is so, Scanlon's objection to Harsanyi given below also applies to Rawls. 68 Ibid.,143-7. Further, Scanlon's contractualism differs from Rawls' in another two respects. The contracting parties in Rawls' Original Position do not have any conception of the good, whereas in Scanlon's theory parties contracting hypothetically to reach an agreement have access to moral reasons. In addition, whereas Rawls' theory is designed for use in arriving at principles of justice governing the basic structure, Scanlon's theory aims at individual actions and (with modifications) at policies and laws. On this last point, see Scanlon, "Individual Morality." 69 See also Li,"Contractualism and Punishment, While there are other objections to the death penalty, the current objection is grounded in the premise that there are mistaken convictions. According to this objection, the death penalty cannot be justified to innocent people convicted by mistake. 71 This case is, however, different from "cruel and unusual" punishment and vicarious punishment in that whereas we know the identity of those who are punished by cruel and unusual punishment and vicarious punishment, we may never know the identity of those who are executed as a result of erroneous conviction. According to Risinger, "Innocents Convicted," 3.3% to 5% of those convicted of capital rapemurder charges in the U.S. in the 1980s were actually innocent. Even though we do not know who these people were, would the moral badness or wrongness of killing innocent persons be absolved if we do not know who they are? The answer is clearly no. To see this point, consider a bomber about to drop bombs on the civilians of enemy country. Would the badness or wrongness of killing civilians be absolved if the pilot did not know the identity of those who were about to be killed? The answer is no.
72 Some utilitarians might even argue that because utilitarianism is correct, "cruel and unusual" punishment and vicarious punishment can be justified if the empirical context is such that these punishments would promote utility under the circumstances. 79 The maximin principle "tells us to rank alternatives by their worst outcomes: we are to adopt the alternative the worst outcome of which is superior to the worst outcomes of the others" (Rawls, Theory of Justice, 152-3). The difference principle is to apply to the basic institutions ("basic structure") of society, and permits inequality of goods ("social primary goods") only if such inequality benefits the worst off. See Harsanyi, "Can the Maximin Principle," for counterarguments against the maximin principle and the difference principle. Glover defend the plausible view that it is generally much worse to be convicted of murder by mistake and then executed, than to be murdered. 84 See Hood, "Capital Punishment, Deterrence and Crime Rates," 6. See also Tonry, "Sentencing in America," 141-98; and Nagin, "Deterrence in Twenty-First Century," 199-263. Andrew von Hirsch et al. also support the view that the key factor for deterrence of punishment to work is certainty of punishment, and not severity of punishment. 85 I assume that the family members of the victim do not have claims comparable to the claims of murder convicts who are actually innocent.
86 Three remarks. First, I suppose here that the general rate of erroneous conviction is not much lower than 3.3% to 5% (see note 16). Second, the main purpose of this subsection is to try to understand why it is more difficult to justify sentences for serious crimes, than penalties for minor offenses, even if the risk of erroneously conviction is the same in both categories. Finally, my third remark is that the question whether it is justifiable to convict is possibly relevant to the Gatecrasher Paradox. In L. Jonathan Cohen's original version of the paradox, a civil trial is involved: "Consider … a case in which it is common ground [between the organizers Plaintiff and the Defendant, who attended a rodeo] that 499 people paid for admission to a rodeo, and that 1,000 are counted on the seats, of whom A is one. Suppose no tickets were issued and there can be no testimony as to whether A paid for admission or climbed over the fence. So by any plausible criterion of mathematical probability there is a .501 probability, on the admitted facts, that he did pay. The mathematicist theory would apparently imply that in such circumstances the rodeo organizers are entitled to judgement against A for the admission-money, since the balance of probability … would lie in their favour. But it seems manifestly unjust that A should lose his case when there is an agreed mathematical probability as high as .499 that he in fact paid for admission" (Cohen,Probable and the Provable,75). We can modify this example into a criminal case. Suppose the jury is faced with a criminal trial in which out of 1000 attendees, 30 of them have paid for entry whereas 970 have not. Can the jury justifiably convict all attendees of an offense for not paying for entry, if the mathematical probability of truly convicting any of them is 0.97? Most experts would think that the answer is no. Assuming that they are correct, there might be an analogous or related (though different) problem about the justifiability of convicting any defendant if the mathematical probability of truly convicting the guilty is only 0.97 (if the rate of erroneous conviction is, say, 3%). I mention this problem in order to set it aside. For discussion on the Gatecrasher Paradox, see Kaye, "Paradox of the Gatecrasher," 104; Rhee,"Probability,Policy,and the Problem," 289,and Enoch et al.,"Statistical Evidence,Sensitivity," 207. 87 For the view that traffic laws deter traffic offenses, see Zaal, "Traffic Law Enforcement." 88 See Hood, "Capital Punishment;" Tonry, "Sentencing in America;" and Nagin, "Deterrence." See also von Hirsch et al. and note 84.
89 See Li,"Contractualism and Punishment, This scenario resembles to some extent the case in which we can either save someone's life, or alleviate numerous headaches, but not both. Should an agent decide to alleviate the headaches, instead of saving the life, he is doing something wrong because the person whose life is not saved would have too much to bear, compared with each of those suffering a headache. Not to save her would be morally impermissible. 91 See Risinger, "Innocents Convicted," esp. 768-80. Moreover, there is anecdotal evidence that the rate of mistaken conviction in serious cases is about 5%. According to Mark Rabil, the defense lawyer of Darryl Hunt, about 5% of prison inmates serving long sentences persistently deny that they were involved in the crimes for which they have been convicted. Although this evidence is merely anecdotal, it corroborates with Risinger's considered judgment that the rate of erroneous conviction in capital rape-murder cases in the US from 1982 to 1989 was 3.3 to 5%, as well as the "conservative" estimation of 4.1% (among those sentenced to death) by Gross et al., "Rate of False Conviction." 92 I myself do not accept positive retributivism apart from condemnation and gratitude as deserved. I discuss this in Li,"Contractualism and Punishment, My position is roughly similar to Scanlon, "Giving Desert its Due." (For my minor disagreement with him, see Li,"Contractualism and Punishment," 194.) For criticism of positive retributivism, see Scanlon,What We Owe,chap. 6;and Hart,Punishment and Responsibility,postscript,. I shall set aside the question of positive desert, since I do not accept it. Consideration of positive desert would require another paper. In section X, I shall consider whether people's emotions or feelings should be given any weight.
93 To see this point, consider the principle, P, that an innocent person should not be mistakenly convicted and punished for what he does not deserve. On the contractualist model, people belonging to different interest groups would raise generic personal reasons in favor of, or against, the principle P. Those falsely convicted would support P on the grounds that they would be worse off. Those who would reject P would be grounded in the generic personal reason that accepting P would undermine communal security. This might be a valid reason if they could show that the way communal security would affect each individual of the community-viz. the harm some individuals would each bear because of insecurity-is comparable to what an innocent person has to bear (which is not merely harm, but also injustice), and if the number of these individuals would greatly exceed the number of the innocents convicted. Those who want to justify life imprisonment would have to show this. I am willing to concede, however, that whether or not life imprisonment is justifiable would, in part, depend on empirical data. 97 See Hart, "Legal Responsibility and Excuses," where he argues that legal excuses are grounded in individuals' need to have control over their lives; and Scanlon, What We Owe, chap. 6, which also stresses the importance of the fair and adequate opportunity for individuals to avoid committing crimes.
98 If life imprisonment consequent upon a mistaken conviction is worse than murder, then Scanlon's contractualism would ignore the number of murder cases caused by such a reduction and conclude that life imprisonment should be reduced to a long sentence. Scanlonian contractualism would, however, take into account the number of murder cases, since a murder case is comparable to life imprisonment based on an erroneous conviction.
99 If the sentence is indeed reasonable, then after the fact of mistaken conviction is taken into account, the relatives of the victim and the community should not strongly object to the sentence meted out.
100 See Li,"Contractualism and Punishment,. I would be the first person to admit that this is an empirical issue, and hence would depend on a particular context. It is totally conceivable, though extremely unlikely in our cosmopolitan cities, that a short sentence could be adequate in terms of deterrence or incapacitation.
101 If a conviction of murder merits three years of imprisonment, how long a sentence would be appropriate for the offense of armed robbery? Suppose armed robbery carries a two-year sentence. One problem is that we would be hard-pressed to determine the difference in gravity between murder and armed robbery, because the difference in imprisonment is only one year. I am indebted to Joe Lau for this point.