A reconstructed dialogue

The Consequentialist: It is customary to justify the reasonable doubt standard by appealing to Blackstone’s maxim that “it is better that ten guilty persons escape, than that one innocent suffer.”^{Footnote 7} If the maxim is right, then we should want a standard of proof that makes false positives (i.e., false convictions) substantially less likely than false negatives (i.e., false acquittals).^{Footnote 8} Some argue that this high standard of proof is precisely the reasonable doubt standard.^{Footnote 9} This is a consequentialist justification,^{Footnote 10} as it derives the standard from considering exclusively the value of the state of affairs that a criminal trial may produce. It is rough, but it can be refined.

Scholars often associate the reasonable doubt standard with a probability of guilt in the vicinity of .9.^{Footnote 11} By adopting a decision-theoretic approach to the problem of selecting the criminal standard of proof, and by using Blackstone’s maxim as the source of the relative values of convicting the innocent and acquitting the guilty, we can provide a consequentialist justification for the reasonable doubt standard thus understood. According to this approach the decision maker should convict when the expected utility of convicting is equal to, or higher than, the expected utility of acquitting. Importantly, the expected utility of an outcome is calculated by multiplying the probability of that outcome with its utility. There is no need to set out the decision-theoretic model in full here.^{Footnote 12} It suffices to point out that if the values of Blackstone’s maxim are used to express the relative utility of the false positive and the false negative of a trial—thus indicating that the former is ten times worse than the latter—the decision-theoretic formula shows that a decision to convict is justified only if the probability of guilt is equal to, or higher than, .9.^{Footnote 13} Hence, the choice of the reasonable doubt standard.

The problem with this consequentialist justification lies precisely in Blackstone’s claim. It is doubtful that Blackstone engaged in a careful consideration of the relative costs of the false outcomes of criminal trials. And, even if he did, it is doubtful that his calculation would still be accurate nowadays. Consider the problem of recidivism. It was recently calculated with reference to trials for violent crimes that “every false acquittal enables more than thirty-six crimes (including on average seven violent ones) during the time when, but for the false acquittal, the defendant would have been [incarcerated and, thus,] incapacitated.”^{Footnote 14} If these statistics are accurate, endorsing Blackstone’s maxim would amount to maintaining that the conviction of an innocent defendant is as serious as the failure to prevent seventy violent crimes—not to mention the additional non-violent offences.^{Footnote 15} This is a preposterous claim to make; surely, one would be hard-pressed to deny that the latter evil is greater than the former. However, if we decide to depart from Blackstone’s maxim—claiming, for example, that a false conviction is just six times worse than a false acquittal—the standard of proof is bound to change, i.e., to become lower. True, the above statistics may be imprecise. Indeed, the underlying calculations were described by one scholar as “back-of-the-envelope math”^{Footnote 16} and subjected to serious criticism.^{Footnote 17} In particular, it is plausible that we should adopt a more conservative estimate of recidivism rates and of the incapacitating effect of imprisonment.^{Footnote 18} Still, if one considers a false acquittal’s undeniable potential for victimisation and suffering, it is very unlikely that they would endorse the ratio that Blackstone proposes.^{Footnote 19} What is more, not only does the consequentialist calculus sit uncomfortably with a standard of proof as high as the reasonable doubt standard; it seems to demand the adoption of different standards of proof depending on the circumstances of each case, e.g., the seriousness of the crime and the dangerousness of the defendant. After all, these circumstances impact on the utility measurements of the possible outcomes of the trial and, therefore, on the selection of the relevant standard of proof.^{Footnote 20}

I am, therefore, drawn to the conclusion that from a consequentialist perspective having a single evidential threshold as high as the reasonable doubt standard makes no sense at all.

The Deontologist: Let me suggest first that someone who accepts a consequentialist framework need not also accept your conclusion. You concede that your calculations are not immune from criticism and that, indeed, they have been strongly attacked.^{Footnote 21} This inevitably casts doubt on the soundness of your argument, such that I am tempted to ask you to provide a defence of your data if you want me to take the argument seriously. However, I am willing to assume that your calculations are accurate. My aim is to show you that your argument fails even if taken at its strongest, i.e., even if we grant this assumption. Indeed, your argument strikes me as wrong on grounds of principle. You see, if we were to abandon the reasonable doubt standard for a lower standard of proof we would be deliberately harming innocent people, namely, the innocent defendants who would be convicted as a result of such a change. It is morally unacceptable to harm an innocent intentionally, even if doing so brings about some good such as the improvement in crime prevention that you talk about. Thus, we should not lower the standard of proof.^{Footnote 22}

The Consequentialist: I beg to differ. Not only do I disagree with the principle that you just enunciated; I do not think that lowering the standard of proof as I previously suggested is a violation of such a principle. True, by lowering the standard lawmakers increase the risk that innocent people are convicted. However, in doing so they are merely accepting a degree of risk of false convictions. In no way are they intentionally harming the innocent. Similarly, adjudicators who convict on such a lower standard of proof are merely accepting a degree of risk that the conviction is false. They are not intentionally harming an innocent defendant.

The Deontologist: At a closer look, I concede that there may not be any intentional harming of the innocent in adopting—or in convicting on the basis of—a standard of proof that is lower than proof beyond a reasonable doubt. Your reply, though, invites more careful consideration of two distinct perspectives: that of the lawmaker, who sets the standard of proof, and that of the adjudicator, who determines whether the standard of proof is met and decides the case accordingly. Every day we perform actions that put statistical lives in danger, these being the lives of those that we cannot identify at the time of acting, but bear a risk of being affected by our agency.^{Footnote 23} For instance, when I put my garbage bin on the pavement for collection, I create the risk that any passer-by trips over the bin and sprains an ankle. Similarly, were lawmakers to lower the criminal standard of proof, they would impose a risk of false conviction on any future defendant. In other words, the lawmakers’ decision-making concerns statistical lives. Unlike lawmakers, though, adjudicators are confronted with identified lives: defendants in the flesh, with given histories, affections, and plans for the future.

Now, we have a powerful moral intuition that—ceteris paribus—it is worse to harm identified lives or put them at risk, than it is to harm or put at risk statistical lives.^{Footnote 24} This intuition has psychological ramifications, including the fact that we find it more difficult to resolve to act when our action harms or imposes a risk of harm on identified lives—especially when the harm is unjustified, as it is in the case of a false conviction. Because of this moral intuition your argument in favour of lowering the standard of proof strikes me as unacceptably reductive. You haven’t been explicit about this, but in presenting your statistics and in the resulting comparative assessment of false convictions and false acquittals, you seemed to consider exclusively the detached perspective of the lawmakers, a perspective from which defendants are statistical lives. You did not consider the engaged perspective of the adjudicators, a perspective from which defendants are identified lives. And yet, our moral intuition suggests that from the perspective of the lawmakers a (risk of) false conviction is not as bad as it is from the perspective of the adjudicators. Ignoring the perspective of the adjudicators means, therefore, that your assessment of the disvalue of a false conviction is grossly inaccurate; more precisely, a false conviction should be deemed considerably worse than you suggest. Notice, instead, that the value of a false acquittal should not vary depending on which of the two perspectives is adopted, given that those who may be victimised by the falsely acquitted are statistical lives from either perspective.^{Footnote 25} Because of your failure to attribute the correct disvalue to false convictions, your conclusion that we should abandon the reasonable doubt standard rests on shaky ground.

The Consequentialist: It may be true that some of us have the intuition that you mention. And yet, I cannot accept the normative significance that you attribute to it. When it comes to morality, I believe in agent-neutrality, i.e., “whether some consequences are better than others does not depend on whether the consequences are evaluated from the perspective of the agent (as opposed to an observer).”^{Footnote 26} I cannot accept that the same event—a false conviction—acquires different value depending on the perspective from which it is assessed, be it that of the lawmakers or that of the adjudicators. Thus, I do not share your view that my calculations are inaccurate.

What is more important, even if we assumed for the sake of argument that you are right about the different value that a false conviction acquires when considered from the perspective of the adjudicators and about the importance of accounting for this perspective in my calculations, it would be unclear why we should retain the reasonable doubt standard. You need to do more in order to justify this conclusion. I suggest that you work out a careful account of the relative value of the false outcomes of the trial, so as to try to confute my earlier argument that the standard should be abandoned.

Let me restate our problem. Proof beyond a reasonable doubt is not equivalent to certainty.^{Footnote 27} Nor is it the highest standard of proof beside certainty.^{Footnote 28} Indeed, historians argued that the reasonable doubt standard was introduced precisely in order to make convictions easier than they were under the previous regime (which did not itself require certainty).^{Footnote 29} It follows that even the adjudicators who convict based on proof beyond a reasonable doubt are imposing a risk of false conviction on the defendant. In fact, nothing short of a standard of certainty would do if we want to neutralise this risk. Needless to say, should we require certainty of guilt for conviction, we may as well do without a criminal justice system, a conclusion that neither of us would accept. Now, if a risk of false conviction is unavoidable, the problem for us is to determine which degree of risk is justified and which isn’t. The consequentialist framework is capable of giving a principled answer to this question. You seem to suggest that the reasonable doubt standard reflects the appropriate degree of risk, but I have yet to see if you have a fully-fledged justification for such a claim.

The Deontologist: Well, my justification of the reasonable doubt standard rests in part on the intuition that I discussed earlier—concerning the difference between statistical and identified lives—and on the impact that this intuition has on the assessment of false convictions. It is frustrating that you do not recognise that this intuition has normative significance within the context of our debate. But I can still offer you a justification for retaining the reasonable doubt standard even leaving the intuition aside.

In your last contribution you make some obvious but important points. Convicting on the basis of proof beyond a reasonable doubt imposes a risk of false conviction on the defendant, and the only way to avoid this imposition is to require certainty of guilt. However, requiring certainty would make the administration of criminal justice impossible. A trade-off therefore seems inevitable. If we want a system that is capable of convicting the guilty, we must accept a quantum of risk of false conviction. As you argue, the problem lies in determining how much risk is justified.

Although I agree with how you restate the problem, I could not disagree more with the methodology that you suggest we adopt in order to solve it. You see, if we were to adopt an unfettered consequentialist calculus like yours, nothing would stop us from accepting a very high risk of false conviction when doing so would maximise value. In other words, nothing would prevent us from adopting a low standard of proof—possibly even a standard as low as the preponderance of the evidence! Even though we must live with the risk of false conviction, we owe it to the innocent defendant^{Footnote 30} to do the best that we can in order to minimise this risk.^{Footnote 31} Indeed, morality binds us to do our best in this respect, given how dreadful false convictions are—even when considered exclusively from the detached perspective of the lawmakers. It is because of this moral requirement that we must retain the reasonable doubt standard. This is not the highest conceivable standard of proof, but it is the highest standard of proof that we can feasibly satisfy in the circumstances of a criminal trial. It is, therefore, the standard that morality enjoins us to adopt.

The Consequentialist: So much is wrong in what you are saying. You argue that the reasonable doubt standard is “the highest standard of proof that we can feasibly satisfy in the circumstances of a criminal trial.” This claim is very difficult to defend. After all, the notion of ‘proof beyond a reasonable doubt’ is extremely hard to define; certainly, the interpretation of the formula is bound to be subjective.^{Footnote 32} As a result, the standard does not easily lend itself to comparisons with other standards in the same region, less still to the exact identification of a corresponding probability threshold. How can you exclude that there is a more stringent standard that we may be able to satisfy in the trial context?^{Footnote 33}

Also, if you are so concerned with protecting the defendant from false convictions, I do not understand why you are so focused on the standard of proof, at the expense of other relevant procedural devices.^{Footnote 34} The truth is that currently we are hardly doing our best to protect defendants from false convictions. For instance, why not increase the number of jurors? Why not impose unanimity in jury deliberations? Why not require judicial review of every guilty verdict? Why not exclude confessions and eyewitness testimony, on the grounds that they are often unreliable?^{Footnote 35} All these changes are fairly easy to implement and it is reasonable to expect that they would make false convictions less likely. If we do not go in this direction it is mainly because we consider that the reduction in the risk of false convictions that these changes may produce is not worth the resulting increase in the probability that guilty people escape punishment. As you can see, this is precisely the trade-off that we discussed earlier. You seem willing to accept the full implications of this trade-off with respect to some procedural devices, but not with respect to the standard of proof. I fail to see why this is so. This brings me to a more important criticism.

You assert that morality binds us to do the best that we can feasibly do in the circumstances of the criminal trial in order to protect defendants from false convictions. This may be true of your morality. It certainly isn’t of mine. But is your morality silent about the protection of innocent people from the risk of being the victims of crimes committed by the falsely acquitted? I doubt it. However, if your morality takes the risk of victimisation seriously, it cannot at the same time require that we do our best to protect defendants from false convictions. Given the phenomenon of recidivism, the more protection we give to innocent defendants, the less we give to innocent people at risk of victimisation by the falsely acquitted. Doing our best on one front may require doing very little—if anything at all—on the other.^{Footnote 36} You need to explain to me why we should give priority to the interests of the defendant and, in particular, why should we do so to the extent of doing our best to protect such interests.

The Deontologist: Before giving you the explanation that you ask for, let me first point out an imprecision of yours. You just said that “[g]iven the phenomenon of recidivism, the more protection we give to innocent defendants, the less we give to innocent people at risk of victimisation”. This is true only to some extent. In fact, I can think of measures that would protect both. Investing in forensic science and improving surveillance systems are just two of these measures; indeed, they would improve the quality of the evidence, thus decreasing the rate of false convictions and increasing that of true convictions.^{Footnote 37} So, our problem is not exactly a zero-sum game, as you suggest.

Now, let me explain why we should prioritise the interests of the defendant in being protected from false conviction over the interests of innocent people in being protected from victimisation. In falsely convicting a defendant, the state causes harm. In failing to protect an individual from being the victim of a crime committed by the falsely acquitted, the state is merely allowing harm to happen. Morality draws a clear distinction between doing and allowing harm,^{Footnote 38} which is reflected in the familiar distinction that our criminal law draws between act and omission.^{Footnote 39} To do harm is substantially worse than to allow harm to befall an individual. Why, may you ask, is this the case? Consider that “the basic intuitive idea behind the distinction is fairly clear: as an agent I must take special responsibility for what I do to other people as opposed to what merely befalls them or what other people inflict on them. When I do harm to someone I treat that person worse than I treat someone whom I merely allow to [be harmed]. Thus, although I have a duty not to stand by when people are being harmed, I must give greater weight to ensuring that I do not harm anyone myself.”^{Footnote 40}

Given our previous disagreement on the moral significance of an intuition that we seem to share, I am wary of relying on these intuitive thoughts alone in order to justify the moral significance of the distinction between doing and allowing. Fortunately, more can be said to provide such a justification. You see, refraining from doing harm generally involves less sacrifice of the agent’s individual freedom and resources than preventing other people or things from doing harm. Morality recognises this fact. A claim not to be harmed is considerably stronger than a claim to be protected from harm, precisely because of the difference in the extent of sacrifice that these two claims demand of the agent.^{Footnote 41} Thus, violating the former claim is considerably worse than violating the latter. This explains why we should give priority to the interests of the defendant in being protected from false conviction, when compared with the interests of innocent people in being protected from victimisation. It explains—or, at least, it contributes significantly to explaining—^{Footnote 42} why a false conviction is substantially worse than a false acquittal.

Now, you may be right in claiming that one cannot confidently assert that the reasonable doubt standard is the highest standard of proof that we can feasibly satisfy in the circumstances of the criminal trial. And I agree with you that in designing our criminal procedure we are not doing our best to protect the defendant from the risk of false conviction. This is partly because we accept a trade-off between false convictions and victimisations on the part of the falsely acquitted—or of any other criminal for that matter. In other words, we are willing to accept some instances of the former in order to avoid some instances of the latter. I share your reason as to why we do so, namely, having a criminal justice system that can protect citizens from crime. And I concede, therefore, that my previous claim that we should do our best to protect the defendant from the risk of false conviction was overstated; indeed, if we were to do our best in this regard, the criminal justice system may not afford citizens sufficient protection from crime. And yet, I insist that the way in which you understand the trade-off between the false outcomes of the criminal trial is profoundly mistaken. The distinction between doing and allowing harm is integral to this trade-off: it determines the weight to be assigned to the false outcomes, and consequently demands that—whilst we should not do our best to protect innocent defendants—we should direct most of our efforts towards protecting them. More precisely, it demands that we prioritise the defendants’ interests in being protected from false conviction, over the interests of innocent people in being protected from victimisation. Its imprecision notwithstanding, the reasonable doubt standard reflects the trade-off as informed by the morally weighty distinction between doing and allowing harm.

The Consequentialist: I have three concerns with your characterisation of the trade-off. First, even if one were to accept the distinction between doing and allowing harm and the fact that such distinction has implications for our debate, the choice of the reasonable doubt standard would not be necessitated. If the recidivism rate were very high and if we could confidently say that pre-trial procedures are good at shielding the innocent from wrongful prosecutions, the trade-off would clearly advise in favour of adopting a lower standard of proof.^{Footnote 43} After all, if you accept the trade-off you must accept that an instance of false conviction and an instance of victimisation—as well as the respective risks and the claims of those who bear such risks—are commensurable. There may be a situation such as that just described where the risk of victimisation is so high, and the risk of false conviction so low, that you must be willing to lower the standard of proof. The bottom line is that accepting the trade-off entails accepting that the reasonable doubt standard is not set in stone; rather, the standard of proof should vary as the relevant circumstances affecting the likelihood of each false outcome vary. But let me move to more important criticisms.

Second, I doubt that your argument relying on the doing/allowing distinction has much purchase in the context of state agency and, in particular, of criminal justice. In general, I am open to the possibility that there is a difference in cost between refraining from causing harm and preventing harm; and that, because of this difference, a claim not to be harmed should be given more weight than a claim to be protected from harm, other things being equal.^{Footnote 44} However, when it comes to state agency in the criminal justice domain, the difference cannot be sufficiently great to justify the reasonable doubt standard. Here it is decisive to consider that state agency—whether it concerns ensnaring criminals or trying defendants—always involves a complex institutional apparatus and a multitude of people.^{Footnote 45} Thus, it seems wrong to assume that pre-empting crime requires considerably more sacrifice on the part of the state than avoiding false convictions. This assumption seems particularly egregious if made with reference to the specific scenario that we are considering, one where all that the state needs to do to prevent the instances of victimisation at issue is to incapacitate individuals—i.e., the falsely acquitted—that have already been identified, investigated, and tried. Now, I concede that such calculations concerning state agency would require more careful attention. But the burden to provide calculations is on you, if you want your argument based on the doing/allowing distinction to bear any weight in our debate. Otherwise, I am left with my intuition that the difference in cost that you posit just isn’t as great as you assume.

Third, what puzzles me most is the fact that a deontologist insists on the importance of the doing/allowing distinction in the context of state agency in the criminal justice domain. People in your camp tend to emphasise the duties of the state towards its citizens.^{Footnote 46} They often advance social contract arguments in order to justify a duty on the part of the state to protect defendants from false convictions, and in order to defend the reasonable doubt standard based on this duty.^{Footnote 47} Whether or not we accept that the social contract is the right framework to discern the reciprocal obligations between the state and its citizens, don’t you think that we should conclude that if the state has any duty at all one of them must be the duty to protect its citizens from crime? You probably agree with this conclusion. After all, absent such a duty I doubt that you would be able to justify the criminal justice system. More importantly—a point for the contractarian—absent such a duty, the social contract would hardly be functional to overcoming a hypothetical state of nature, in particular, the profound existential uncertainty caused by an unbridled freedom of the individual. Now, consider that crime often inflicts harms that are greater than any false conviction. If your intuitions suggest otherwise, think of murder or of rape. Consider also that—as I have just suggested—discharging the duty to protect from crime, and, in particular, from the crimes of the falsely acquitted, is not much costlier for the state than discharging the duty not to inflict harm on citizens through false convictions (if it is costlier at all). In light of these considerations, how can you possibly argue that the latter duty is so much stronger than the former so as to justify the reasonable doubt standard?^{Footnote 48}

The Deontologist: Your criticisms are strong, and yet not insuperable. Your last comment invites a restatement of my position, with which I shall begin. I do claim that the state has both a duty not to inflict harm on its citizens and a duty to protect them from crime. Of course, under special circumstances the state may be justified in, or excused for, not discharging these duties—but this is an issue that need not detain us here. The former duty is stronger than the latter, precisely because discharging this duty is less onerous for the state—a point that you seem to concede, notwithstanding your speculations about the costs of state agency in the context of criminal justice. Again, the difference in strength between the duties ultimately depends on the fact that refraining from harming is a lesser sacrifice of one’s freedom and resources than protecting from harm; in other words, it follows precisely from the distinction that morality draws between doing and allowing harm.

In light of this restatement, my previous contribution can be interpreted as an argument to the effect that the reasonable doubt standard is justified by the very difference in strength between these two duties of the state: roughly, the avoidance of false convictions should be given substantial priority over the protection of citizens from the crimes of the falsely acquitted, because the duty not to inflict harm is substantially stronger than the duty to protect from crime. However, I agree with you that crafted in these terms my justification for the standard appears flimsy. Even though the duty to protect from crime is weaker than the duty not to inflict harm, I doubt that the difference is so pronounced as to justify a standard of proof as high as proof beyond a reasonable doubt.^{Footnote 49} In fact, I now see that the key to justifying the reasonable doubt standard is not the relative strength of these two duties of the state, or the underlying distinction between doing and allowing harm. It is, instead, the relative wrongness of the state’s behaviours consisting, respectively, in convicting the innocent and in failing to protect citizens from the crimes of the falsely acquitted. You may be wondering how I can disentangle the relative wrongness of these particular behaviours from the relative strengths of, respectively, the state’s duty not to inflict harm and the state’s duty to protect citizens from crime. The truth is that only in the case of false convictions a duty is violated.

Let’s assume that you are right about the phenomenon of recidivism and that a false acquittal involves a very high risk of victimisation. For the sake of argument, let’s even go as far as to assume that every false acquittal entails one or more crimes committed by the falsely acquitted. Now, compare the relative wrongness of convicting the innocent and of acquitting the guilty. Individuals are harmed in both cases, but the responsibility of the agent—the state—varies significantly. In convicting the innocent, the state harms an individual without justification.^{Footnote 50} This is clearly a violation of the state’s duty not to inflict harm. When the state acquits the guilty, instead, it is the falsely acquitted who should be held entirely responsible for the harm that they later inflict on innocent individuals. True, in letting the guilty at large the state has a causal impact on the resulting crime. But the crime is an intervening voluntary act of a third party, precisely, the falsely acquitted. This voluntary act ‘breaks’ the chain of causation linking the action of the state—i.e., the false acquittal—to the instance of victimisation, such that we cannot hold the state responsible for the harm caused by the criminal.^{Footnote 51} As a result, we cannot say that the state breached any of its duties towards those who are victimised. You see, here I am merely applying another basic principle that our criminal law^{Footnote 52} derives from morality: the principle according to which responsibility is negated by the presence of a novus actus interveniens. In calculating the acceptable trade-off between the false outcomes of a criminal trial, we must have regard to this principle. If we do so, we come to the conclusion that whilst the state is responsible for false convictions, it is not responsible for the crimes committed by the falsely acquitted. The state acts wrongfully in the first case, not in the latter. This explains why the state should strive to reduce the occurrence of false convictions—a task for which the reasonable doubt standard is perfectly suited.

The Consequentialist: Your reply is not satisfactory. To see why, we must consider more carefully the relevance of third parties’ voluntary interventions to the question of responsibility, but within the specific context of state agency vis-à-vis criminality.

Let me suggest, first, that when an agent anticipates the intervening act of a third party you should have reservations about the claim that such an act negates the agent’s responsibility for the resulting harm.^{Footnote 53} After all, in this situation the agent may be seen as accepting the causal role played by the third party. Notably, this seems to be the epistemic situation of the state vis-à-vis the action of the falsely acquitted. Indeed, given the high standard of proof, the state should expect many acquittals to be false. Also, given the data evidencing high recidivism rates, the state should expect that many falsely acquitted defendants will reoffend. This objection aside, the main problem with your new account is that it rests on the implausible claim that whilst the state has a duty to protect citizens from crime, this duty is not violated when crimes are committed by the falsely acquitted.

My impression is that you are asking your principle of the novus actus interveniens to do work that it simply cannot do. If the voluntary act of a third party—more precisely, the crime of the falsely acquitted—were to play any role whatsoever in reducing or negating the responsibility of the state, this would be the role of severing the causal chain linking the false acquittal to the instance of victimisation. This role, however, would concern only the discharge of the state’s supposed duty not to inflict harm. In other words, the presence of the intervening voluntary act may be the reason why we should not conclude that, through falsely acquitting the defendant, the state has actively inflicted harm on its citizens, thus breaching its corresponding duty. If this understanding of your novus actus interveniens principle is correct, though, I find it puzzling that you seem to argue also that a crime committed by the falsely acquitted does not involve the state’s violation of a supposed duty to protect its citizens from crime. The intervening voluntary act at issue is precisely one of the acts that the state should prevent in performing such an obligation; how can this act possibly negate the state’s responsibility vis-à-vis the duty to protect? Consider also that it may be substantially easier for the state to prevent the crimes committed by the falsely acquitted than it would be to prevent the crimes of individuals who have not been identified, investigated, or monitored. And yet, a consequence of your reasoning seems to be that the state breaches its duty to protect citizens from crime only when it fails to prevent the latter crimes—for your argument relying on a break of the causal chain between false acquittal and crime applies only to the former.

If—as you argue—a state bears both a duty not to inflict harm and a duty to protect from crime, you have not proved that your framework can justify the claim that the state is not responsible for the victimisations on the part of the falsely acquitted, while it is for false convictions. True, you may still argue that the relative strength of the two duties has a bearing on the relative wrongness of the two false outcomes, and thus on the choice of the applicable standard of proof. However—as you conceded earlier—this difference in strength cannot alone justify the reasonable doubt standard.

The Deontologist: You seem to be right in saying that whilst the crimes of the falsely acquitted may negate the state’s responsibility for actively inflicting harm on its citizens through false acquittals, they may not negate the state’s responsibility for failing to protect its citizens from crime. I must concede that at this point I am left with an inadequate justification for the reasonable doubt standard, chiefly built upon the doing/allowing distinction and the resulting difference in the strengths of the two state duties at issue. I also think that the distinction between statistical and identified lives has a role to play in the justification, given that—as I argued earlier—it has a bearing on the disvalue of false convictions. In this regard, I must express my frustration at the fact that you fail to recognise the moral significance of this distinction. This said, I am not admitting defeat. My intuition that the reasonable doubt standard is justified—and that, therefore, your consequentialist argument is just wrong—is still strong. Without further reflection, though, I feel unable to unpack it and defend it…

The Retributivist: I have been listening to your exchange carefully and I would like to make a contribution. I am a deontologist as well. However, when it comes to punishment I take a distinctive approach, that is, retributivism. Retributivism is a family of theories, with their own differences.^{Footnote 54} But all retributivists would agree with the claims that it is intrinsically bad not to convict and punish the deserving (i.e., the guilty) and that it is intrinsically bad to convict and punish the non-deserving (i.e., the innocent). With the adverb “intrinsically” I signal the fact that these outcomes are bad per se, that is, irrespective of the consideration of their consequences, such as the fact that the falsely acquitted may recidivate, or the distress that punishment may inflict on the defendant’s family and friends. I must make clear from the start that I agree with you that the choice of the standard of proof must depend on a careful trade-off between the false outcomes that I have just mentioned.^{Footnote 55} I also accept that some of the consequences of these outcomes—whether for the defendants, their families, or society at large—are relevant to the trade-off.^{Footnote 56} In other words, that they are relevant to the calculation of the relative worth of the false outcomes, besides the intrinsic evil of each. Like the arguments of my friend the Deontologist, though, my argument aims to set moral limits to the consequentialist calculus. A crucial limit consists precisely in circumscribing the set of consequences that should be considered.

The trade-off should be designed so as to reflect two important moral considerations, relating to the retributivist claims that I mentioned earlier. First, even though false acquittals and false convictions are both intrinsically bad, the evil that is intrinsic in the latter outcome is considerably greater. There is more than one way for a retributivist to justify this claim.^{Footnote 57} First, a retributivist may appeal to the value of proportionality. You see, retributivists like me believe that punishment should be proportional to the crime. If we assume this proportionality, it seems to follow that the punishment of an innocent for crime x is at least as bad as it is for this person to be the victim of crime x.^{Footnote 58} Once we realise this, we are bound to conclude that punishing the innocent for x is substantially worse than acquitting the guilty of x.^{Footnote 59} Even leaving proportionality aside, the above claim about the relative intrinsic worth of the false outcomes of a trial may be justified by considering the rights or interests that these outcomes violate. One may believe that in failing to punish the guilty the state violates the rights of particular individuals, e.g., the victim, the law-abiding citizens, the defendants themselves.^{Footnote 60} Alternatively, one may believe that the duty of the state to punish the guilty is not owed to anyone in particular, but merely “reflects the value of doing justice”.^{Footnote 61} Now, if a false acquittal really violates rights, the losses that it imposes on the bearers of these rights are certainly much less severe than the loss imposed on the innocents by the violation of their fundamental right not to be punished.^{Footnote 62} If, instead, a false acquittal does not violate anyone’s right, consisting merely in the failure to do justice, we can claim a fortiori that such “abstract wrong” is substantially less severe than the violation of the innocent’s right not to be punished.^{Footnote 63}

Notice that my claim about the relative worth of the false outcomes does not concern their consequences; in particular, it does not concern the instances of victimisation resulting from false acquittals. This claim is, therefore, different from—and immune from the criticisms of—the claims previously defended by my friend the Deontologist, who, relying on different moral principles, was trying to show that from the perspective of the state the instances of victimisation resulting from false acquittals are less serious wrongs than may appear at first sight. As I mentioned earlier, though, consequences should play a role in the trade-off. This brings me to the second moral consideration that I want to make.

If all consequences were included in the trade-off, my position would be very similar to that of the Consequentialist, and—provided that the Consequentialist’s data are accurate—^{Footnote 64} I doubt that I would be able to offer a justification for the reasonable doubt standard. In fact, not all consequences should be included! I just argued that from a retributivist perspective punishing the innocent is a very serious misdeed. This is why we retributivists claim that the beneficial consequences of punishment—such as deterrence and, thus, crime prevention—may be treated as reasons to punish only if desert is present, lest we condone and promote the punishment of the innocent.^{Footnote 65} This means that, for example, crime prevention may justify punishing the deserving, but it may never justify punishing the non-deserving. Using an effective simile, one may say that “[d]esert functions like a gate in the normative equivalent of a transistor: until it is switched on, the potentially greater normative force of the [beneficial consequences of punishment] is switched off.”^{Footnote 66} The crucial point for our debate is that desert should play this transistor-like role not only in the justification of the institution of punishment, but also in the choice of the standard of proof.^{Footnote 67} Thus, the beneficial consequences that punishment may produce should not be factored into the calculus underlying this choice; only once the standard has been identified, and once guilt has been proven to its satisfaction, these consequences can be treated as reasons to punish. As for the detrimental consequences of punishment—such as its monetary costs, as well as the suffering and the loss of opportunities that punishment imposes on the convict’s family—retributivism does not object to factoring them into the calculus underlying the choice of the standard of proof.^{Footnote 68} After all, although retributivists are committed to punishing the deserving, they accept to incur the moderate evil of a false acquittal if going ahead with punishment may produce significant detrimental consequences.^{Footnote 69}

Now, combine the claim that 1) the intrinsic evil of a false conviction is considerably worse than the intrinsic evil of a false acquittal with the claim that 2) in setting the standard of proof the trade-off between the false outcomes of the trial should not incorporate the consideration of the beneficial consequences of punishment. Under these circumstances the value of a false acquittal is expressed exclusively in terms of its moderate intrinsic evil. True, a failure to punish involves a missed opportunity to produce general and specific deterrence, and, consequently, crime prevention. But the failure to bring about these consequences should not be considered in assessing the value of false acquittals, precisely because these are among the beneficial consequences of punishment. On the other hand, the value of a false conviction is expressed both in terms of its substantial intrinsic evil and in terms of, for example, the stigma, suffering, and lost opportunities that punishment imposes on the innocents and their dears—in other words, the detrimental consequences of punishing. Again, any beneficial consequence that punishment may produce—even if imposed on the innocent—should be ignored in calculating the value of a false conviction. As a result of this revised calculation, a false conviction turns out to be substantially worse than a false acquittal. The “retributive balance” between the two false outcomes “tips heavily, but not absolutely, in favour of protecting the innocent”.^{Footnote 70} The reasonable doubt standard is the best candidate to reflect this balance: it indicates that we should be mainly concerned with protecting defendants from false conviction, but without forgetting that false acquittals are still an affront to our ideal of justice.

The Deontologist: Having listened carefully to your argument, I praise it for its elegance and coherence. It seems that you have finally produced a justification of the reasonable doubt standard that is solidly anchored in a deontological framework. You have succeeded where I previously failed. I see, however, that our friend the Consequentialist is not yet persuaded.

The Consequentialist: In fact, I am not persuaded by your argument, Retributivist. I find it puzzling, incomplete, and blunt.

It is puzzling because it calls into question existing practices of punishment, notwithstanding that they are informed by the reasonable doubt standard. Thus, the argument ultimately seems to call into question the very standard of proof that it defends. You see, our criminal justice system produces a very high number of false convictions.^{Footnote 71} I wonder whether this rate is justified under your theory; in other words, whether it is a proper reflection of the relative value that you attribute to trial outcomes in the calculation underlying the standard of proof.^{Footnote 72} I fear that it may not be, for your value attributions would probably demand a lower rate of false convictions. I appreciate that false convictions have a complex aetiology, and cannot be imputed exclusively—or even mainly—to the standard of proof and to how it is applied.^{Footnote 73} And yet, I cannot help thinking that the very standard that you defend—i.e., the reasonable doubt standard—may be an important contributing factor to the high—for you, unacceptably high—rate of false convictions. A possible reason why the standard may play such a role is that—as discussed in my exchange with the Deontologist—its formula lends itself to different interpretations. In particular, empirical research^{Footnote 74} has shown that mock jurors match the reasonable doubt standard with probability thresholds that you would certainly consider too low. Are you sure that the standard of proof that you defend may not be among the causes of a state of affairs that you cannot condone?

Your argument also strikes me as incomplete. You present as a tenet of retributivism the claim according to which the beneficial consequences of punishment may be treated as reasons to punish only once desert has been established. You then argue that desert should play this transistor-like role also when it comes to selecting the appropriate standard of proof. In your view, this means that the standard of proof should result from a calculus that does not take into account the beneficial consequences of punishment. These consequences come into play—i.e., are treated as reasons to punish—only once the standard of proof is satisfied. Based on these premises, you conclude that the appropriate standard for criminal trials is proof beyond a reasonable doubt. Now, it is evident from your theory that you treat proof of guilt beyond a reasonable doubt as being equivalent to the establishment of desert. This equivalence, though, needs elaboration and defence.

Are you making an ontological claim here? In other words, are you saying that desert is, or is constituted by, proof beyond a reasonable doubt that the defendant is guilty? If so, you must explain why one should understand desert in these terms. You see, retributivism may well say that desert has the transistor-like role that you describe; but such a role of desert does not entail that desert is, or is constituted by, proof of guilt to the satisfaction of a probability threshold selected without considering the beneficial consequences of punishment—that is, in your view, the reasonable doubt standard. You may have very good reasons to draw such an inference from the role of desert to its constitution, but you should spell these reasons out. Indeed, I would not be surprised if many of your retributivist fellows rejected this ontological view, arguing instead that desert is actual guilt—i.e., that it requires a probability of 1 that the defendant committed the crime.

It may be that in expressing the above equivalence you are making an epistemological claim, instead. You may be arguing that proof of guilt beyond a reasonable doubt—although not identical to, or constitutive of, desert—is the concept that gives us the best epistemic access to the notion of desert; in other words, the concept that brings us closest to knowing what desert is. This claim too would need justification. In fact, a retributivist may reasonably consider that higher standards of proof—including certainty—give us better epistemic access to the notion of desert. As I suggested earlier, it is possible to conceive of these standards. Now, you may agree with your fellow retributivist on this point; and yet you may insist that proof beyond a reasonable doubt has epistemological precedence with respect to such higher standards, because of pragmatic reasons, i.e., you may say that it represents the best that we can feasibly do in order to ascertain desert in the circumstances of the criminal trial. This claim, though, would bring us back to square one. Indeed, I already addressed a similar claim in my exchange with the Deontologist and need not rehearse my counterargument here.

If you are not making either an ontological or an epistemological claim about the notion of desert, it is not clear why a retributivist should agree with you that proof beyond a reasonable doubt is the probability threshold the satisfaction of which justifies the consideration of the beneficial consequences of punishment.

As mentioned earlier, I also find your argument blunt. In fact, this criticism equally applies to the arguments of our friend the Deontologist, who, to be fair, seemed to voice a similar, and equally justified, criticism against my arguments.^{Footnote 75} I always enjoy to hear arguments and to assess their internal coherence. However, the temptation is strong to just shrug all of your arguments off. After all, I am a consequentialist and your arguments are based on premises that I simply do not accept. For instance, you resort to the idea of proportionality in order to calculate the relative intrinsic values of false acquittal and false conviction. I may recognise a value to the proportionality between crime and punishment, but only if—and to the extent that—proportionality is instrumental to securing a more valuable set of consequences than we may obtain without proportional punishment. Also, I certainly cannot accept your claim that the beneficial consequences of punishment are screened off until desert is established. The bottom line is that from my perspective there is no such a thing as a moral principle or rule that transcends the logic of value maximisation.

During the exchange with you and the Deontologist I had to put myself in your shoes for us to be able to have a debate at all.^{Footnote 76} I have tried to show you that even if one were to accept the premises of your theories, a justification of the reasonable doubt standard would not follow.^{Footnote 77} The alternative for me would have been to remain silent and shake my head in denial, refusing to test the validity of your arguments because of an a priori objection to the soundness of their premises. I now see that engaging in the debate was the right thing to do, as I believe that I have been able to appreciate the strengths of yours and the Deontologist’s arguments, but also to uncover their weaknesses. However, I am conscious of the fact that you will always be able to adjust your premises, tweak your deontological theory by bringing in other moral principles, and give the argument for the justification of the reasonable doubt standard another go; after all, the exchange with the Deontologist was an example of this trial and error. Should you endeavour to do so, I would be ready to put on your shoes again, and try to spot some new inconsistency in your argument. But this is the most that I can do here, unless we are ready to abandon the narrow topic of our current debate and to engage in an arduous—and possibly endless—discussion on the broader merits of consequentialism and deontology. The truth is that I doubt that I will be able fully to convince you that we should abandon the reasonable doubt standard, as I doubt that you will be able fully to convince me of the contrary.^{Footnote 78} We are somehow stuck with our fundamental disagreement.

The Intermediary: I have waited for an appropriate time to make my contribution to this most interesting exchange, and no time seems better than this. For the most part, I agree with the Consequentialist’s objections to the Deontologist’s and the Retributivist’s justifications of the reasonable doubt standard. At the same time, I experience uneasiness when I consider the Consequentialist’s argument in favour of lowering the standard of proof. I—like the Deontologist—have a strong intuition that the reasonable doubt standard is justified.

You probably wonder whether I am a consequentialist or a deontologist myself and, if I am neither, what am I. In truth, these questions are simply beside the point and do not need answering here. What the Consequentialist just said concerning the limitations of the debate so far is right. Your disagreement about the justification of the reasonable doubt standard is rooted in a fundamental disagreement about normative frameworks. As a result, whilst you are able to criticise each other’s arguments for lack of internal coherence, you cannot provide knockout arguments without getting bogged down in the dispute on the alternative between consequentialism and deontology. The challenge for us lies in finding a common ground that would allow us to push forward—and possibly bring to an end—the current debate, without triggering our fundamental moral disagreement. I will show you that it is possible to justify the reasonable doubt standard independently of which normative framework one adopts. The common ground on which I will build this justification is twofold. It comprises: first, the value of respect for the defendant and its relationship with the reasonable doubt standard; second, basic rules of instrumental rationality, which I expect all of us to endorse. I trust that you will all be able to relate to my argument notwithstanding your different standpoints, and to embed it in your own normative framework. Should you not be persuaded that my argument justifies the reasonable doubt standard, at the least I hope that it gives you strong reasons in favour of retaining the standard, which you should account for in your own future arguments.

We all agree that there is value in punishing those who have committed crimes, and thus disvalue in letting them go free. True, we may disagree as to why this is so, e.g., the Consequentialist stressing the preventative function of true convictions, the Retributivist drawing our attention to their intrinsic goodness. We may also disagree on the contents of punishment and on how it should be tailored to the crime. But if we did not consider that punishing the guilty is valuable—and, therefore, letting them go free is deleterious—we would not support the existence of the criminal justice system in the first place.^{Footnote 79} Thus, we would not bother to debate the choice of the criminal standard of proof. We also agree that there is value in avoiding false convictions. Again, we may disagree as to why this is so. For instance, the Consequentialist seems to recognise that the instrumental costs of a false conviction—in terms of suffering, erosion of trust, monetary expense, etc.—outweigh the possible instrumental benefits—in terms of deterrence. The Retributivist ignores the instrumental benefits of a false conviction and focuses exclusively on its instrumental costs and, especially, on its intrinsic evil. All of you—the Deontologist included, at last—seem to agree that the relative values of trial outcomes^{Footnote 80} should determine the standard of proof,^{Footnote 81} given that the standard of proof has a bearing on the relative frequencies of these outcomes.^{Footnote 82} This claim, however, is wrong. In particular, it is wrong to maintain that the only—or even the primary—function of the standard of proof is to reflect and operationalise such value judgments. I don’t deny that, to some extent, the standard of proof may perform this role. My claim, though, is that the standard’s primary role is different and consists in realising another value that we all cherish, namely, the respect for the defendant. The remaining of my contribution will proceed as follows. I will first clarify the relationship between the reasonable doubt standard and the respect for the defendant. Then, I will appeal to basic rules of instrumental rationality to show that we should retain the reasonable doubt standard, precisely in virtue of this relationship.

I am not interested in giving a general account of the concept of respect, of its elements or its distinct kinds.^{Footnote 83} My account is narrow in two ways. First, I focus only on the respect that the state owes the defendant qua defendant. Of course, in doing this I do not intend to deny that the state or other agents owe the defendant respect also qua person.^{Footnote 84} Second, I am not offering you a comprehensive account of what respecting the defendant means and requires. It is sufficient for my purposes to show that a particular behaviour on the part of the state constitutes a clear instance of disrespect for the defendant. Now, I expect that you will all agree that this behaviour is an instance of disrespect, which is why in the following I will refer to it as such. However, it is worth clarifying that the success of my argument in favour of retaining the reasonable doubt standard does not depend on this conceptual agreement. You may endorse a different conception of respect from mine and argue that the behaviour at issue is not disrespectful. And yet, my argument may still succeed. For its success does not depend on a shared understanding of respect, or of respect for the defendant for that matter. It depends, instead, on whether you attribute value to avoiding such state behaviour, and I will appeal to your respective normative frameworks to show that you do so. What state behaviour is at issue here, then?

The state is clearly disrespecting defendants when it denies them the opportunity to defend themselves.^{Footnote 85} This behaviour is disrespectful because it betrays a failure on the part of the state to acknowledge defendants as reasonable interlocutors in the decision-making process—that is, as individuals who are capable of making a valuable contribution to such a process—and/or it denies the defendants’ reasons the possibility of making such a contribution.^{Footnote 86} Failing to regard people as reasonable interlocutors and/or blocking out their reasons from decision-making may well be instances of disrespect when they concern persons other than defendants and contexts other than the criminal process. What makes this behaviour especially disrespectful if it concerns defendants in criminal cases is that defendants are the “subjects” of the decision-making process within which such behaviour takes place. In other words, the process is chiefly about the defendants’ past agency and it may have a severe impact on their present and future lives. There is something particularly disrespectful in not regarding people as capable of contributing to a decision-making enterprise, and/or in not allowing them to contribute to it, when it is their agency that is at issue and it is their lives that may be impacted upon by the process and by the final decision.^{Footnote 87}

The Deontologist and the Retributivist probably explain the above instance of disrespect in terms of a violation of one of the defendants’ fundamental rights (i.e., the right to defend oneself from a criminal charge) and of the state’s correlative duty towards them. It is on this ground that they see such state behaviour as intrinsically bad; in other words, that they attribute value to its avoidance.^{Footnote 88} Whether the Consequentialist recognises the existence of this right and correlative duty, the Consequentialist must recognise that treating defendants in conformity with them is valuable.^{Footnote 89} Of course, for the sake of producing a state of affairs that they considered particularly valuable, the Consequentialist may well accept some instances of denying defendants the opportunity to defend themselves. However, I doubt that in the Consequentialist’s calculations such instances would be considered as neutral. I trust that the Consequentialist would, instead, factor them as costs, which, again, shows that value would be attributed to their avoidance. After all, this state behaviour frustrates an interest that most defendants understandably have; and the resulting one-directionality of the decision-making process is likely to undermine the truth-finding function of the trial and to produce more miscarriages of justice than would occur in a system allowing for a dialogue between parties offering different perspectives. Moreover, as I will soon show, abandoning the reasonable doubt standard involves routine denial of the opportunity for defendants to defend themselves. I am confident that the Consequentialist would consider it very costly to disrespect defendants across the board. Such behaviour on the part of the state is bound to erode the trust of its citizens, who would see their reasonable demand to be heard and protected by the state frustrated. This erosion of trust is likely to have negative consequences for the functioning and the stability of the criminal justice system itself.

My argument is that convicting without proof beyond a reasonable doubt realises the instance of disrespect for defendants that I have just mentioned, namely, it denies defendants the opportunity to defend themselves; and that, therefore, applying the reasonable doubt standard is a necessary condition for respecting defendants—a behaviour which, as just suggested, we all value, at least as far as my minimalist notion of respect is concerned.

The standard of proof beyond a reasonable doubt requires that the adjudicator convict only when there is no reasonable innocent account available. In other words, the standard would be misapplied if the defendant were convicted notwithstanding that such a reasonable account exists. Let’s now consider what would happen if the adjudicator were allowed to convict on the basis of a less stringent standard of proof. Here it is crucial to consider that in our system conviction involves the assertion that the defendant is ‘guilty’. This is, indeed, the content of the fact-finder’s verdict. Whilst any practicable standard of proof inevitably falls below certainty, the assertion of guilt is categorical, implying a probability of 1 that the defendant committed the crime charged. There is no acknowledgment of innocent accounts in the ‘guilty’ verdict; no expression of doubt whatsoever. The verdict attributes responsibility to the defendant as if it were a matter of certainty.^{Footnote 90} Now, making such an assertion in the face of a reasonable innocent account—as it would often^{Footnote 91} happen were the adjudicator to convict on the basis of a standard lower than proof beyond a reasonable doubt—involves discarding the account as unworthy of consideration; in particular, as unworthy of influencing our reconstruction of the relevant facts. Again, consider that no mention of this account can be made in the categorical statement of fact constituting the guilty verdict—a statement that represents the result of such a reconstruction.

The problem here is that the discounted innocent account is—ex hypothesi—reasonable. Admittedly, the attribute “reasonable” is not an easy one to define with precision. However, we may all agree that at the very least this attribute indicates the presence of reasons in support of the claim or argument that it describes.^{Footnote 92} If so, discarding a reasonable account as unworthy of consideration involves treating in this very way both the individual reasons underlying such an account and the combination of these reasons. More precisely, it involves denying to these reasons their status of reasons. They are denied their natural capacity to influence our view of the world,^{Footnote 93} not to mention our ensuing agency.^{Footnote 94} Whilst the presence of reasons against an assertion should—and normally does—lower one’s confidence in the fact that the assertion is true, a categorical statement of guilt leaves no room for such an adjustment: it simply ignores conflicting reasons.

Crucially, the discounted reasons speak in favour of the defendant; indeed, we are talking about a reasonable innocent account, after all. Therefore, denying them their status of reasons amounts to dismissing the defendants’ interest in being heard,^{Footnote 95} and thus in being given the opportunity to defend themselves. In fact, a system that accepts conviction through categorical verdicts without proof beyond a reasonable doubt is designed in such way that whether or not defendants give—or merely have—reasons in their defence may make no difference whatsoever to the outcome. The system may still make a categorical assertion of guilt even if there would be good reasons to doubt this assertion. Such a system, therefore, fails to treat defendants as reasonable interlocutors, capable of making a contribution to the enterprise of fact finding, and of trial adjudication more generally. Of course, defendants in this system may be provided an actual opportunity to contribute, by presenting their case at trial. However, whilst the case presented may be reasonable, there would be no trace of it in the categorical guilty verdict that may follow. The defendants’ contributions would be denied the role that we would expect them to have in any instance of fact finding that is responsive to reasons. Such an opportunity to defend oneself would thus be a sham,^{Footnote 96} not the constituent of a respectful interaction.^{Footnote 97}

Categorical guilty verdicts deny the presence of reasonable doubts; if we want a system that respects defendants, we should allow such verdicts only when guilt is proven beyond a reasonable doubt. Put differently, my claim is that the “leap” from the satisfaction of the standard of proof—which inevitably has to be lower than certainty for the system to operate—to the categorical verdict—which, instead, expresses certainty—can only be justified when the standard is proof beyond a reasonable doubt. Indeed, when only a lower standard is satisfied, the leap would disrespect defendants, as it would prevent the recognition of possible reasons in their defence, thus denying them the opportunity to defend themselves. Therefore, a system adopting a lower standard of proof is disrespectful of any defendant, because by design it is simply unfit to do justice to the reasons they may have to support their claim; thus, by design it undermines the defendants’ capacity to contribute to decision-making.

Let’s briefly take stock. So far, I argued for the conclusion that it is valuable to retain the reasonable doubt standard, because the standard is a necessary condition for respecting defendants. This conclusion rests on the following claims: 1) that a state that denies defendants the opportunity to defend themselves is disrespecting them; 2) that it is valuable to avoid such behaviour; 3) that guilty verdicts make a categorical attribution of responsibility; 4) that issuing a guilty verdict in the presence of a reasonable doubt denies defendants the opportunity to defend themselves; and that, therefore, 5) a criminal justice system adopting a standard of proof that is lower than proof beyond a reasonable doubt is by design undermining the defendants’ opportunity to defend themselves. Importantly, for you to endorse these claims, you don’t need to relinquish any moral commitment characterising your respective normative framework. Again, you may disagree with my minimalist notion of respect for the defendant. If so, you may expunge the language of respect from my argument. The conclusion would remain that retaining the reasonable doubt standard is a necessary condition to realise something that we all have reasons to value, i.e., giving defendants the opportunity to defend themselves.

The Deontologist: I beg your pardon for interrupting your argument midway. But your interpretation of guilty verdicts does not persuade me. You claim that these verdicts are categorical; in other words, that they attribute responsibility as a matter of certainty. However, we all know that any practicable standard of proof must be lower than certainty. By “we” I don’t mean exclusively the participants in this dialogue; I also mean jurors, judges, lawyers, indeed, the public in general. If so, in reading the verdict we certainly do not interpret it as making a categorical assertion; rather, we interpret it as a statement to the effect that guilt was proven to the satisfaction of the applicable standard of proof. In other words, in our understanding the statement expresses a probability of guilt that is lower than certainty. Such a statement inevitably leaves room for doubt. Therefore, it does not discount possible reasons in favour of a defendant; contrary to what you claim, it does not disrespect defendants by denying them the opportunity to defend themselves.

The Intermediary: This objection is welcome, for it allows me to clarify my argument. A possible response would be to challenge your claim about how “we”—understood in the broad sense that you use—allegedly interpret guilty verdicts, by saying that this claim requires empirical testing. Although a fair response, it is hardly the strongest I can give. After all, for my purposes, how guilty verdicts should be understood matters more than how “we” actually understand them.

You see, there is a precise reason as to why guilty verdicts are worded in categorical terms. This reason lies in the concept of ‘punishment’. Among the essential elements of this concept is that punishment is issued because of what the defendant did. In other words, punishment is for a deed.^{Footnote 98} A corollary of this element is that, as a conceptual matter, we cannot punish the defendant “for having probably committed a crime”, or “because there is evidence showing that she may have committed it”. We can only punish the defendant ‘for having committed the crime’, without adding any qualification.^{Footnote 99} True, we cannot be certain that the crime occurred; in fact, as was recognised several times in this dialogue, any practicable standard of proof falls below certainty. However, if we want to see ourselves as punishing an individual, we must behave as if we were certain;^{Footnote 100} we must make the leap which I mentioned earlier, that from the satisfaction of the standard of proof to the categorical guilty verdict. The question arises as to when we are justified in making such a leap. I answered it by saying that the leap is justified only when proof of guilt is beyond a reasonable doubt, lest we disrespect the defendant.

Our practices of criminal justice and, in particular, their conceptual basis indicate that the categorical assertion of guilt made in the guilty verdict is no mere window dressing. It has, instead, a fundamental role to play. Without it, we could not see ourselves as being in the business of punishment. In light of this, not only does it seem unlikely that “we” understand guilty verdicts as you suggest. Even if shared by some, such an understanding would be incompatible with a key concept informing our own criminal justice practices, namely, the concept of ‘punishment’.

Let me now move to the second part of my argument.

The Consequentialist: Before you proceed, I have a different objection to make. For the sake of argument, let’s concede that you persuaded us that applying the reasonable doubt standard is a necessary means to realising a value that we all cherish: respect for the defendant. The problem is that the reasonable doubt standard is also an obstacle to punishing the guilty. In other words, lowering the standard of proof is a means to realise another value that we all cherish. If so, whilst I may recognise that you have correctly pointed our attention to the value of respecting the defendant, I contend that you have not given us a justification for the reasonable doubt standard. In fact, as we are willing to accept the punishment of some innocent individuals for the sake of bringing about more true convictions, we would certainly be willing to accept some instances of disrespect in order to achieve this very goal. Thus, respect for the defendant should be just another consideration to be factored into the calculus underlying the choice of the standard of proof. I am confident that, even with this amendment, the calculus would suggest that we adopt a standard lower than proof beyond a reasonable doubt.

The Intermediary: This objection could not have been timed better. In the second part of my argument—to which I now move—I will show you that the objection fails. I will do so by appealing to rules that I expect all of you to endorse, i.e., basic rules of instrumental rationality.

Let me first clarify that abandoning the reasonable doubt standard to adopt a lower standard does not mean accepting “some instances of disrespect”, as your objection misleadingly contends. It means accepting that every defendant is disrespected: every defendant’s ability to influence fact finding and the resulting state agency is dismissed and undermined by the design of the system, irrespective of the outcome of their trials. This is because a system with categorical verdicts, but without the reasonable doubt standard, is structurally incapable of recognising all the defendant’s reasons as such. This explains why I said earlier that abandoning the reasonable doubt standard involves a routine failure to respect defendants.

The Consequentialist: Fine, but this leaves open the question as to whether we are willing to sacrifice respect for defendants across the board for the sake of obtaining more true convictions. In other words, the fact that abandoning the reasonable doubt standard produces routine disrespect may not be—in fact, I suggest that it is not—enough to justify such a standard. We may still conclude that routine disrespect is a cost we should bear in order to promote true convictions.

The Intermediary: I am afraid that your rejoinder still misses a fundamental point about the instrumental function of the reasonable doubt standard. As a result, it mistakenly presents increasing the rate of true convictions and avoiding routine disrespect of defendants as necessarily conflicting goals that must be subject to a trade-off. It is no surprise that you understand the relationship between these two goals in such terms. After all, your argument against the reasonable doubt standard gained traction precisely by assuming that the two goals on which the debate has focused thus far—namely, the aim of increasing the rate of true convictions and that of protecting the innocent from false convictions—conflict with one another and must, therefore, be subject to a trade-off. Once this assumption permeated the debate, it became particularly challenging for the Deontologist to produce a justification for the reasonable doubt standard—although, to be fair, the Deontologist expressed some reservations about the assumption.^{Footnote 101} At the end of my contribution I will show you that not even the goal of increasing the rate of true convictions and that of protecting the innocent from false convictions should be understood as necessarily conflicting. And yet, I will also suggest that an alternative understanding of their relationship may not by itself provide the justification for the reasonable doubt standard. But first things first…let’s now consider the instrumental role of the reasonable doubt standard vis-à-vis the goal of increasing the rate of true convictions and that of avoiding routine disrespect of defendants.

I concede that adopting a standard of proof that is lower than the reasonable doubt standard is instrumental to increasing the incidence of true convictions. This is the very reason why you have been arguing so vigorously for a lower standard of proof. However, it is crucial that we understand that abandoning the reasonable doubt standard is not a necessary means to achieve this goal—a point that the Deontologist previously hinted at, without however giving it sufficient weight.^{Footnote 102} Other means are available and include increasing police resources, investing more in the development of forensic sciences and in order to make the respective evidence available at trial, improving surveillance systems. This is just an exemplification. But you will probably agree with me that the means in this short list promise to be more effective at increasing the rate of true convictions than abandoning the reasonable doubt standard.^{Footnote 103} After all, they promise to improve the quality of the evidence and to bring more guilty people to court in the first place, things that a standard of proof cannot do. Crucially, none of the means in the list entails rejecting outright a value that we all cherish.^{Footnote 104} Abandoning the reasonable doubt standard, instead, does so, given that applying the standard is a necessary means to respecting the defendant. Given this state of affairs, what are we to do about the standard of proof?

It may help to restate our problem as follows. We value two goals, among others: avoiding routine disrespect of defendants and securing more true convictions.^{Footnote 105} Now, performing a certain action—i.e., applying the reasonable doubt standard—is a necessary means to achieving the former goal. Performing an alternative action—i.e., lowering the standard of proof—is a means to achieving the latter goal, although not a necessary means, nor even the most effective means available. Common-sense understanding of means-end relationships advises that we perform the first of the two actions, namely, applying the reasonable doubt standard, and pursue the latter goal through means other than the alternative action, namely, the action of lowering the standard of proof. This would hold true even if we were to conclude that avoiding routine disrespect is a less important goal than securing more true convictions—a point raised by the Consequentialist, and one on which I want to remain non-committal. In fact, once we heed the different instrumental role that the two actions concerning the standard of proof play in pursuing our two goals, we realise that these goals do not conflict with one another; they can both be achieved, and there would be no need to sacrifice what we may consider as the lesser goal.

My conclusion that we should retain the reasonable doubt standard can be reinforced by expressing our decision-making in terms of intentions and beliefs. Let’s hypothesize that all of us intend both to secure more true convictions and to avoid routine disrespect of defendants. Of course, intending a result involves a level of psychological commitment that is greater than merely treating it as a valuable goal. Whilst I believe that we all consider these two results as valuable goals, I am not sure whether you actually have any greater commitment to them. Thus, my starting point is a hypothesis, not an assertion, about our shared intentions. Now, I may well be wrong about this, but I trust that by this time we all believe that applying the reasonable doubt standard is a necessary means to avoid routine disrespect, whilst lowering the standard of proof is not a necessary means to increase the rate of true convictions. This being the case, our failure to intend to apply the reasonable doubt standard would be an instance of instrumental incoherence, as it would violate what has been dubbed the “instrumental principle”.^{Footnote 106} The principle tallies with common-sense reasoning. It states that if one intends to achieve a goal and believes that a given action is a necessary means to achieving it, then—unless they give up the intention to achieve such a goal—they ought to intend the necessary means as well.^{Footnote 107} Assuming that we don’t give up the intention to avoid routine disrespect, the instrumental principle enjoins all of us to retain the reasonable doubt standard. Notice that the same reasoning would not apply with regard to the means-end relationship between lowering the standard of proof and securing more true convictions. This is because none of us believes that lowering the standard is a necessary means to such an end. Therefore, the instrumental principle does not demand that we intend to lower the standard. The different instrumental role played by the standard of proof with respect to the intention to avoid routine disrespect, on the one hand, and that of securing more true convictions, on the other, explains why the two intentions are compatible. In other words, it explains why an agent can hold both at the same time without being vulnerable to the charge of incoherence—as the agent would be if the necessary means to fulfil one intention were not compatible with the necessary means to fulfil the other. This consideration on the compatibility of the two intentions, together with the discussion we had so far about the value of respecting defendants and that of convicting the guilty, support the conclusion that a reasonable agent should hold both intentions; therefore, that a reasonable agent should defend the reasonable doubt standard.

I said earlier that one of the merits of this justification is that it is in keeping with both a consequentialist and a deontological normative framework. I hope that I have succeeded in showing this. Another merit is that, unlike some consequentialist and deontological arguments, my justification does not concern a supposed numeric or verbal equivalent of the reasonable doubt standard, e.g., a probability threshold of .9 or “the highest standard humanly possible.” Instead, it concerns the standard directly, giving prominence to its formulation and, especially, to the notion of reasonableness contained in it. Thus, not only is this justification more ductile than the ones that you have offered so far—for it is not embedded in a single normative framework; it is also more precise—for its object is the formula of the standard itself. It is difficult to overstate the importance of this last point. You see, someone may think that it was unnecessary for me to elaborate on the relationship between the reasonable doubt standard and the value of respecting defendants. I could have simply considered the goals of securing (more) true convictions and of protecting the innocent from false conviction—these being the goals on which you focused your attention prior to my contribution, and goals that I certainly value. Then, I could have appealed to considerations of means-end rationality to show that we may pursue both goals without lowering the standard of proof. True, if we want to establish a criminal justice system that is capable of securing true convictions, we must choose a standard of proof that is lower than certainty, thus accepting that some instances of false conviction will occur. However—the hypothetical argument continues—this is as far as the trade-off between the above two goals should go. Once our practicable system is running, we should not tinker with the standard of proof in order to secure more true convictions; we should, instead, maintain a high standard. After all, whilst a high standard of proof gives protection to innocent defendants—one may even claim that it is necessary to achieve such a goal effectively—lowering the standard of proof is not a necessary means to securing more true convictions. As discussed earlier, there are other means that promise to be more effective in this regard, for example, increasing police resources, enhancing the quality of forensic evidence, improving surveillance systems. Notably, these measures seem to be instrumental also to protecting the innocent from false conviction, especially because they promise to improve the quality of the evidence available at trial.

Now, I concede that the argument that I just sketched is promising. However, at the very best it may provide the justification for a high standard of proof, whatever this may be. I fail to see how—without further elaboration—it would justify the reasonable doubt standard. An engagement with the formula of the standard and its meaning is needed to provide such a justification. My discussion of the relationship between the standard and the respect for the defendant relies precisely on such an engagement.

A final note: you may find it odd that the goal of protecting the innocent from false conviction plays no role in my justification of the reasonable doubt standard. Rest assured that I do value this goal. As with the goal of securing (more) true convictions, though, I believe that not even the goal of protecting the innocent from false conviction should be treated as the driving force behind the choice of the standard of proof. And yet, I do recognise and value the fact that the ‘alternative means’ to securing (more) true convictions, advocated for earlier in my argument, are also instrumental to protecting the innocent from false conviction. More importantly, I do recognise and value the fact that this latter goal seems well served by the reasonable doubt standard. Not only protecting the innocent and respecting the defendant are compatible goals; they can both be pursued using the same tool.

Concluding remarks

Needless to say, the debate is not over, but merely on hold. It will be for the retributivists who participated in it to respond to the criticisms of their contributions, which I put in the Consequentialist’s mouth. Also, the Intermediary’s justification will hopefully spur new contributions. At this stage, I can only identify some aspects of this justification that may attract attention and criticism. Consequentialist thinkers may question the value that the respect for the defendant is alleged to have within their framework. In particular, they may try to argue that avoiding routine disrespect is a considerably less important goal than increasing the rate of true convictions. They may also try to argue that lowering the standard of proof increases the rate of true convictions more effectively than any of the alternative means mentioned by the Intermediary.^{Footnote 108} If successful with these arguments, consequentialists may conclude that the good is maximised by increasing the rate of true convictions through lowering the standard of proof, even if this comes at the cost of routine disrespect for defendants. Of course, acting on this conclusion would require that they explicitly repudiate the intention to avoid routine disrespect, lest their decision-making is instrumentally incoherent. Deontologists, instead, may be interested in providing further elaboration on the necessary connection between the reasonable doubt standard and the respect for the defendant; they may as well reject this connection—as consequentialists may, of course—and argue that the justification of the standard lies elsewhere…

Whether or not consequentialists and deontologists will make the Intermediary’s justification their own, its main value resides in the challenge that it raises: finding a common ground between normative frameworks, where to discuss the merits and demerits of the reasonable doubt standard. According to the Intermediary, a possible common ground is provided by the value of respect for the defendant—and its relationship with the reasonable doubt standard—and by rules of instrumental rationality. There may be another, more fruitful common ground on which to build an attack to, or a justification of, the standard. Perhaps, only when this common ground will be found we will see an end to this debate. Importantly, though, the common ground is not only functional to the scholarly dispute on the reasonable doubt standard. We should expect that those who are tried and punished endorse different normative frameworks: a discourse that transcends these frameworks is necessary if we are to be capable of justifying to all of them a crucial condition for punishment.^{Footnote 109}