Aim or preference? Reflections on the commitment to the truth in the criminal process

Abstract

It is widely accepted that the criminal process aims at the truth. It is also widely accepted that convicting the innocent is worse than acquitting the guilty. While apparently unrelated, these two claims are in tension with one another. The latter claim is traditionally used to justify a standard of proof that is skewed in favour of the defendant, aimed at protecting the innocent from conviction. A skewed standard, however, is not the standard of proof that minimises expected errors; that is, it is not the standard to choose if truth-finding is indeed the aim of the criminal process. The article attempts to overcome this tension. It argues that, if someone seeks consistency between the commitment to the truth and the commitment to protecting the innocent from conviction, they should treat true outcomes as a preference on which the process is based, not as a/the aim of the process. Notably, a preference entails a more modest practical commitment than that entailed by an aim. Taking this more modest commitment to the truth has implications for the regulation of any phase of the criminal process in which the value of truth appears to be in tension with non-epistemic values.

Introduction

Imagine that you are a fact finder in a criminal trial, and that you are asked to decide truthfully. In other words, you should aim for decisions that are a true representation of the relevant state of affairs: you should aim at issuing a ‘not guilty’ verdict only if the defendant is, in fact, not guilty; and you should aim at issuing a ‘guilty’ verdict only if the defendant is, in fact, guilty. Imagine that you are also asked to reflect, in your decision-making, the value judgement that convicting the innocent is the worse of the two possible mistaken outcomes of a trial (the other being acquitting the guilty). Because of this, you commit to adopting a standard of proof that is markedly skewed in favour of the defendant. According to this standard, the prosecution cannot obtain a conviction merely by presenting a case that is stronger than the case supporting the hypothesis of innocence. The prosecution’s case must be substantially stronger than the latter for conviction to be warranted. Being a demanding evidential threshold, the standard you choose reflects the above value judgement: it contains the risk of convicting the innocent within limits that you find acceptable, even if this comes at the cost of a more significant risk of acquitting the guilty. As expected, while you adjudicate the long sequence of trials that have been assigned to you, the skewed standard requires that you issue a ‘not guilty’ verdict in many cases in which the evidence of guilt is stronger than that of innocence. These are cases in which you would have issued a ‘guilty’ verdict if the only directive you had received had been that of aiming for truthful decisions. After all, in these cases you have greater chances of getting it right if you convict the defendant.

This simple fictional scenario shows that there is a tension between the reasonable aim of truth-finding and the reasonable value judgement that false conviction is worse than false acquittal. This article is an attempt to clarify, and to overcome, this tension. In the next section I introduce and elucidate the claim that the criminal process aims at the truth, as well as the claim that false conviction is worse than false acquittal. With the claims properly laid out, I proceed to define the scope and the plan of the work.

Setting the stage

It is widely accepted that the criminal process aims at the truth: it aims to determine whether propositions concerning criminal behaviour are true. This claim about the process is so widespread to warrant the label of ‘platitude’, a claim repeated so frequently to be uninteresting.¹ It is due to this platitude that the analogy between legal fact-finding, on the one hand, and historical and scientific research, on the other, makes intuitive sense and has frequently occupied scholars.² These practices share the commitment to truth-finding. Ronald Allen put the point emphatically: ‘Notwithstanding the differences in fact finding among science, law, and history, one commonality stands out: without accurate fact finding, the rest of the process is useless’.³

As already pointed out by others,⁴ to make sense of the commitment to truth-finding it is unnecessary to delve into the longstanding, complex, philosophical discourse on the concept of truth. It suffices to rely on the plain Aristotelian correspondence theory of this concept, according to which ‘to say of what is that it is, and of what is not that it is not, is true’.⁵ Therefore, to claim that the criminal process – like historical and scientific research – aims at the truth is simply to claim that it aims at issuing only assertions such that things are as is asserted. Accordingly, a procedural outcome that involves a false assertion frustrates the aim, whereas a procedural outcome that involves a true assertion fulfils it.

The claim about truth-finding is often expressed with an ordinary verb alone (e.g., to aim, to pursue), as if it provided a description of the actual regulation and practice of criminal procedure. However, the claim is generally meant as a modal statement expressing a desideratum: the criminal process should aim at the truth. Truth is seen as a valuable goal, a goal worth pursuing, irrespective of whether the rules and practice of the criminal process being studied implement this normative judgement effectively. Indeed, these rules and practices are often critically assessed in terms of their truth-conduciveness. A telling example is Hock Lai Ho’s recent discussion of the claim at issue, where the claim is often expressed with an ordinary verb alone but is understood as a modal statement throughout.⁶

In this article I question the claim that the criminal process should aim at the truth; that is, the claim that the process should aim at issuing only assertions such that things are as is asserted. For brevity, I will refer to this claim as the ‘truth-claim’. To be sure, I am not interested in defending or attacking the claim. Undoubtedly, there are good reasons as to why the process should aim at the truth, the most obvious residing in the conditional form of criminal law norms: ‘if criminal behaviour, then punishment’.⁷ This form is underlain by the view that, whatever goods are realised by the infliction of punishment, the infliction of punishment realises these goods only if (or it best realises such goods if) criminal behaviour occurred. If this view is indeed correct, the process should aim at issuing punishment only if it is true that criminal behaviour occurred. Be that as it may, here I assume that the claim about truth-finding is, at the very least, reasonable. My focus is on the tension between this claim and another normative claim, which I will also take for granted: the claim according to which, in any given case, a false conviction (i.e., convicting the innocent) is worse than a false acquittal (i.e., acquitting the guilty). For brevity, I will refer to the latter claim as the ‘preference-claim’, in that it expresses a preference for one false outcome over the other.

While at first sight the truth-claim and the preference-claim appear unrelated, there is a straightforward relation between the two. The preference-claim is generally seen as the reason for choosing a standard of proof that is skewed in favour of the defendant, aimed at protecting the innocent from conviction. And yet, a standard of proof of this kind does not minimise the expected number of errors. In other words, in terms of truth-conduciveness it is not the best standard available. In this article I interrogate the tension between the truth-claim and the preference-claim. I show that, if the truth-claim is taken seriously, this tension is inevitable. To clarify, I am not arguing, as others have done already,⁸ that the criminal standard of proof is concerned with the distribution, as opposed to the reduction, of errors; that is, my contribution is not to point out that, unlike other rules of evidence, the standard is not ‘driven by the quest for the truth’.⁹ I make the distinct argument that, if someone endorses a standard of proof that is informed by the preference-claim, they cannot reasonably endorse also the truth-claim. To express endorsement of both is the sign of insincerity, cognitive dissonance or, at least, inaccuracy in one’s expression. However, I also argue that there is a suitable replacement for the truth-claim that avoids the tension with the preference-claim, while preserving a significant commitment to the truth.

Notably, the truth-claim is often endorsed in a more sophisticated version than that discussed so far. It is widely accepted that truth should be the primary aim of the criminal process, not the only aim: there is a series of non-epistemic values that are in tension with the pursuit of truth and should nonetheless inform the entire course of the process, from the start of the investigation to the conclusion of the trial (and beyond).¹⁰ Evidence rules implementing these values abound and include exclusionary rules, rules of use, privileges, and compellability rules.¹¹ The classic example is that of a rule that excludes a confession obtained through torture, where the exclusion is not motivated just by the risk of unreliability, but also by respect for the physical and psychological integrity of the accused. Indeed, the confession would be excluded even if corroborated by other evidence.¹²

Now, my primary focus here is on adjudication, understood narrowly as the final decision problem of the trial. Adjudication takes place when the investigation is over, the decisions on the admissibility of evidence have been made, the parties have presented their case and, in the jury trial, the judge has concluded the summing up and given the instructions. What is left to determine is which verdict to issue. This decision problem – precisely, adjudication – is regulated by rules identifying the available verdicts and by rules setting the burden and standard of proof, therefore determining when a particular verdict should be issued. Of these rules it is the standard of proof that I am concerned with. The tension that I explore, then, is not the apparent tension, occurring in the phases of the process leading up to adjudication, between the truth-claim and non-epistemic constraints. It is, instead, the tension, occurring in the context of adjudication, between the truth-claim and a value judgement informing the standard of proof. This value judgement is the preference-claim. As I will argue in the article, the tension between the truth-claim and the preference-claim in the context of adjudication is markedly different from the apparent tension, occurring in earlier phases of the process, between the truth-claim and non-epistemic values. This difference is such that, unlike for the latter tension, it is not possible to dissolve the former, or to accommodate it within a consistent framework of goals and rules: assuming the preference-claim, a replacement for the truth-claim is needed. My focus on adjudication notwithstanding, I will argue that the proposed replacement for the truth-claim has implications for earlier phases of the process as well, in that it eases the tension between commitment to the truth and non-epistemic values that may characterise such phases.

Here is the plan of the work. In section 3 I offer a brief analysis of the rules that govern adjudication in the English and Welsh criminal process, considering whether the aim of truth-finding is implied by, if not expressed in, any such rule. I show that, while the standard of proof does not require knowledge (and, hence, truth) for conviction, this does not undermine the truth-claim. The tension between the truth-claim and the standard lies elsewhere. In section 4, I introduce the problem of the tension between, on the one hand, the truth-claim and, on the other hand, a claim informing the skewed standard of proof, the preference-claim. In section 5, I consider two correctives to the truth-claim, which promise to dissolve the tension. I argue that neither corrective succeeds. In section 6, I propose a replacement for the truth-claim that indeed avoids the tension, while maintaining a significant commitment to the truth. In essence, I argue that, if someone seeks consistency between the commitment to the truth and the commitment to protecting the innocent from conviction, they should treat true outcomes as a preference on which the process is based, not as a/the aim of the process. Notably, a preference entails a more modest practical commitment than that entailed by an aim. Section 7 offers some concluding remarks.

Truth-finding and rules of adjudication

Is the aim of truth-finding implied by, if not expressed in, any rule that governs adjudication in the English and Welsh criminal process? Rule 1.1 of the Criminal Procedure Rules 2020 states that ‘(t)he overriding objective of this procedural code is that criminal cases be dealt with justly’ and it goes on to offer a series of factors that are constituents of justice for the purposes of the code. At the top of the list is ‘acquitting the innocent and convicting the guilty’. Arguably, the phrase stands for accuracy, or truth:¹³ because acquittal involves the assertion ‘not guilty’, to acquit the innocent is to make a true assertion; similarly, because conviction involves the assertion ‘guilty’, convicting the guilty is to make a true assertion. Be that as it may, rule 1.1 is general, concerning the process in its entirety; and if it is indeed referring to truth as a constituent of justice, it does so in a way that does not directly affect the decision of any given case. In other words, rule 1.1. is a declaration of intent, not a decision rule.

When one looks at the rules that regulate specifically the problem of adjudication, as defined in section 2, they are hard pressed to find any explicit or implicit mention of the aim of truth-finding. The rule providing the binary verdict system, according to which the only possible conclusions of adjudication are acquittal or conviction, contains no such mention.¹⁴ This is not surprising, given that the rule itself says nothing about the conditions for issuing either verdict. It merely defines the choice set. These conditions are, instead, fixed by the rules setting the burden and standard of proof. The role of the rule allocating the burden of proof is to determine which party should make a reasons-giving effort with respect to which issue, so that a failure of the party to make such an effort will determine the verdict. However, the rule – which is unanimously read into the presumption of innocence –¹⁵ does not specify the extent of the reasons-giving effort that is required to reach the threshold of proof and, hence, to discharge the burden. The rule does not refer to probability thresholds, degrees of belief, confidence levels, not to mention truth. It is the standard of proof that defines the reasons-giving effort. The standard adopted in England and Wales, though, contains no express reference to the aim of truth-finding; in particular, it does not set truth as a condition for convicting. According to the Crown Court Compendium, ‘the prosecution proves its case if the jury, having considered all the evidence relevant to the charge they are considering, are sure that the defendant is guilty’¹⁶ (the Compendium adds that this standard ‘means the same thing as’ proof beyond reasonable doubt).¹⁷ Being sure is having a categorical or outright belief.¹⁸ That the fact finder is sure of guilt is itself strong evidence of guilt. But obviously the fact finder, as any human being, can be sure that something is the case when it is not, in fact, the case. One can be sure of propositions that are not true.

The foregoing may baffle the reader. What would be the point of a standard of proof that sets truth as a condition for convicting? Truth is not a cognitive attitude. Rather, it is a correspondence between an assertion and the world, the latter providing the facts asserted.¹⁹ Requiring this correspondence for conviction could only be a shorthand for requiring a cognitive attitude about this correspondence. Let me clarify. Our cognitive attitudes inevitably mediate our relationship with the world, such that setting something in the world as a condition for the justification of an action could only mean setting as such a condition a cognitive attitude about that something. Now, the correspondence that truth consists of is itself something in the world. Therefore, requiring for conviction the truth of an assertion could only mean requiring a degree of belief in the truth of that assertion. The requirement of the truth of the assertion of guilt that were included in the standard of proof, then, would effectively be superfluous: it would add nothing to the requirement of a cognitive attitude about guilt that were also included in the standard. In fact, it would not specify the degree of belief in (the truth of the assertion of) guilt that the fact finder would have to entertain for conviction to be warranted, nor would it set a requirement of epistemic justification (that is, roughly, a requirement about the evidence needed to justify a conclusion).

Interestingly, in recent years the view that conviction requires knowledge of guilt has gained popularity in the literature.²⁰ According to this view, the fact finder is justified in convicting if, and only if, they know that the defendant is guilty. Under a mainstream analysis, which this view endorses, knowledge consists of a true and justified belief – plus some additional attribute to deal with Gettier-style cases, that is, cases in which intuition suggests that a true and justified belief is not sufficient for knowledge.²¹ Other scholars have defended the view that a justified belief, or a justified degree of belief, in guilt suffices for the purposes of conviction.²² According to this alternative view, truth is not a constituent of the standard of proof. At a close look, in fact, those outlined are not alternative accounts of the standard of proof. At least, they are not alternative accounts insofar as the implementation of the standard is concerned.²³ This conclusion follows from the considerations made earlier. To tell the fact finder that they can only convict if it is true that the defendant is guilty does not add anything to the directive – also included in the standard of knowledge – that they can only convict if they have a justified belief in guilt. Therefore, to direct the fact finder to ‘convict if, and only if, you know that the defendant is guilty’ is the same as directing them to ‘convict if, and only if, you have a justified belief in the defendant’s guilt’. There is nothing that the fact finder can do to satisfy the first standard that, if conscientious, they would not already do to satisfy the second. To put it with Dale Nance, ‘justification is our only access to the truth’:²⁴ there is no access to the truth that is independent of relying on the evidence at our disposal to form cognitive attitudes about the relevant proposition. If it is truth that we are aiming for, we’ll never know whether we got there. The best we can do is to form a degree of belief reasonably based on the available evidence; that is, a justified degree of belief.

Consider also that there is no difference between a standard requiring knowledge and a standard requiring a justified belief when it comes to the problem of reviewing a verdict. Imagine that at t1 the fact finder convicts based on a justified belief in guilt and that, at t2, significant exculpatory evidence is discovered. All that can be said with either standard in place is that, while conviction was epistemically justified at t1, it is not anymore at t2. Neither at t1 nor at t2 can we say that the assertion of guilt is false; we can merely say that the probability of its truth has decreased significantly with the discovery of the exculpatory evidence. Notably, neither standard of proof can alone provide an answer to the practical question as to what to do about the verdict issued at t1: if conviction is not epistemically justified anymore, should the verdict be upheld or reversed? The answer to this question requires considering non-epistemic values – such as finality – that underpin the doctrine of res judicata and, therefore, oppose revisiting decisions, notwithstanding that these decisions are not anymore supported by the evidence.²⁵ Overcoming these non-epistemic considerations, the English and Welsh system has opted for allowing the defendant to appeal a conviction based on fresh evidence. This can be taken as a clear sign of the system’s commitment to the truth. The case law of the Court of Appeal, though, indicates that non-epistemic considerations – such as respect for, and deference to, the institution of the jury – still play a role in the Court’s decision whether to allow an appeal.²⁶

The bottom line is that there is no inconsistency between maintaining that the criminal process should aim at the truth and defending a standard of proof that – unlike knowledge – does not include truth as a condition for the justification of convicting.²⁷ This is because the only way for the fact finder to pursue the truth concerning the issue of guilt is by heeding the available evidence and by forming, reasonably, a cognitive attitude based on such evidence. So, once the law has set as a standard a cognitive attitude like ‘being sure’ or ‘justified belief’, an additional directive to the effect that the fact finder can only convict if it is true that the defendant is guilty would not change the nature of the fact finder’s task; that is, it would not increase the truth-conduciveness of adjudication. If the standard of proof adopted in England and Wales is in tension with the truth-claim, then, this is not because the former merely requires a cognitive attitude about guilt.²⁸ As the next section explains, the tension follows from the fact that this standard of proof is skewed in favour of the defendant.

The tension between the truth-claim and the preference-claim

In England and Wales – but this is a common feature of Western criminal justice systems – the standard of proof is skewed in favour of the defendant. It is higher than the balance of probabilities standard adopted in civil cases. In other words – for those who understand standards of proof as probabilistic thresholds or, at least, rely on probabilities as convenient tools to convey the stringency of standards – the criminal standard of proof is higher than a probability of .5. In such a system one should expect a significant number of cases in which guilty defendants receive a ‘not guilty’ verdict notwithstanding that the fact finder has correctly applied the standard of proof.²⁹ In fact, in such a system not only is it possible, but also it is likely that a correct application of the standard leads to a verdict that makes a false assertion; in particular, a false assertion to the effect that the defendant is not guilty.

One may disagree with this way of portraying the functioning of the system. They may contend that the notion of guilt used in criminal verdicts should be understood in ‘probatory’ terms: it does not consist in the fact that the defendant committed the crime charged, but in the fact that, given a standard of proof, there is sufficient evidence to conclude, according to the standard, that the defendant committed the crime.³⁰ Under this construal, verdicts are about ‘probatory’ truth (whether or not the standard is met by the available evidence), not about ‘material’ truth (whether or not the defendant has committed the crime). So, irrespective of whether the defendant indeed committed the crime, a ‘not guilty’ verdict issued when the standard is not met would not be making a false assertion: it would be making the true assertion that the standard was not met. Under this alternative construal, then, a skewed standard may well produce a significant number of acquittals of guilty defendants, but the verdicts involved would all be truthful insofar as the standard of proof was not satisfied.

While prima facie appealing, this alternative construal does not eliminate the problem posed by a skewed standard. Even if a ‘not guilty’ verdict is understood as making an assertion about the evidence as opposed to the commission of a crime, the problem remains that in a system with a standard skewed in favour of the defendant there will be a significant number of cases in which the defendant is not punished even though they have committed the crime charged. In these cases, the system fails to treat the defendant as they should be treated given their criminal behaviour. What is more, the system incurs an epistemological failure, in that it fails to recognise the defendant as a criminal. Whatever the construal of the notion of ‘guilt’, then, a system with a skewed standard of proof will often authorise decisions that do not conform with the material truth. How can this be squared with the claim that the criminal process should aim at the truth? Understood in probatory terms, a ‘not guilty’ verdict issued against a defendant who has committed a crime does make a true assertion if the standard of proof is indeed not met. However, from a material perspective this verdict is a false negative. Notably, this is the perspective that informs the conditionals constituting substantive criminal law (‘if criminal behaviour, then punishment’). Substantive criminal law, without which criminal procedure would have no reason to exist, is indeed concerned with the commission of crimes: it punishes the commission of crimes and/or it aims at rehabilitating someone who has committed a crime and/or it aims at reducing the commission of crimes, etc. Insofar as the process is instrumental to the goals of the substantive law, the material perspective must also be the perspective adopted by the truth-claim.

There is nothing new in the realisation that a correct application of the standard of proof can lead to false outcomes. Indeed, whatever the stringency of the standard of proof, in the long run false outcomes are inevitable even if the standard is always applied correctly. And yet, choosing a standard skewed in favour of one of the parties is especially problematic from the perspective of truth-finding, because this is not the standard that minimises the expected number of errors. Let me clarify.

Larry Laudan argued that the ‘concept’ of the standard of proof pertains to the ‘soft-core’ of legal epistemology.³¹ According to him, while the ‘hard-core’ consists of rules that are concerned with error reduction, or truth-finding,³² the goal of soft-core rules – therefore, of the standard – is not to reduce the likelihood, or actual number, of errors. Rather, it is to make it likely that the distribution of the errors that do occur reflects our comparative assessment of the two possible error types, false acquittal and false conviction. For reasons that need not detain us here, I don’t agree with Laudan’s view that this is the goal of the standard;³³ but I accept that this is something that the standard does. A standard skewed in favour of the defendant, like that of the English and Welsh system, is informed by the values of the possible trial outcomes and, especially, by a preference for false acquittal over false conviction (the preference-claim). The standard may well lead to a significant number of errors (in particular, false acquittals), but it does so in a way that reflects this preference, such that the likely distribution of errors has a significantly higher proportion of false negatives than of false positives.³⁴ Notice, though, that the preference-claim determines a departure from accuracy: other things being equal, while a skewed standard is likely to decrease the number of errors of the type that is considered more costly, it is also likely to increase the overall number of errors. If all we cared for is truth-finding, that is, minimising errors, we should adopt the ‘balance of probabilities’ standard, which is not premised on a preference for one erroneous outcome over the other (its symmetrical nature follows from a state of indifference regarding the two erroneous outcomes). As consistently pointed out in the literature,³⁵ mandating conviction when the probability of guilt is over .5 (and prohibiting it, otherwise) minimises the expected overall number of errors made in fact finding.³⁶ Notably, what is being minimised by the balance of probabilities standard is the expected number of errors, possibly not the actual number. As Mike Redmayne put it, though, since ‘we are unlikely ever to know the actual error rate in the trial process, … minimising expected errors is the best we can attain’³⁷ in terms of adherence to the truth.³⁸

When Laudan made the conceptual point that the standard of proof does not aim at error reduction but is ‘best conceived as a mechanism for distributing errors’,³⁹ he was being imprecise at best. There may be value in his distinction between the hard-core and the soft-core of legal epistemology and one might consider it correct to claim, following Laudan, that the standard of proof pertains to the latter. Still, it is crucial to appreciate that the standard of proof does not just ‘distribute errors’. The choice of the standard is not neutral with respect to the aim of truth-finding. This choice does affect the expected overall number of errors, and a standard of proof skewed in favour of the defendant is not the standard that minimises this number, other things being equal: if we aimed at error reduction the standard of the balance of probabilities would serve us best.

The foregoing shows that it is problematic to claim that the process should aim at the truth while defending a standard or proof that is informed by the preference-claim.⁴⁰ If the process should indeed aim at the truth, then this is not the standard to choose. Of course, that the process should aim at the truth is the statement of an obligation, if not of a mere aspiration; and, even if they are not fulfilled, obligations and aspirations are valuable in that they provide targets. However, to honestly make the truth-claim, while setting the standard of proof in such way that the aim of truth-finding is outright thwarted or set back is a significant inconsistency. Notice that no similar inconsistency results from endorsing the truth-claim while maintaining – as many do –⁴¹ that also non-epistemic values should affect the regulation of the phases of the criminal process leading up to adjudication. Consider a rule excluding evidence obtained through torture or through other illegal means; or a rule sanctioning the non-compellability of the defendant and their spouse; or a rule screening exchanges between the defendant and their counsel. All these rules can be, and indeed have been, defended by appealing (also) to non-epistemic values, such as physical and psychological integrity, privacy, and autonomy. And it is possible that, in a given case, the application of one of these rules renders unavailable useful epistemic material. However, it is important to appreciate that, from the abstract point of view – the point of view of the lawmaker or the theoretician, that is – none of these rules warrant the expectation that the application of the rule will lead to a decrease in the overall fact-finding accuracy of the system; that is, to an increase in the overall number of errors. This is because it may well be that the evidence that is excluded by the rule in a token case, or the information that the rule prevents from coming to light, are just not reliable, such that ignoring them is instrumental to truth-finding. Indeed, the rules mentioned can also be, and have also been, defended on epistemic grounds: while they further non-epistemic values, arguably they also further overall accuracy.⁴² Moreover, consider that even if the targeted evidence and information were reliable, it may well be that there is sufficient additional evidence in the token case for the fact finder to reach an accurate verdict. Conversely, from the abstract point of view the expectation of a decrease in the overall accuracy of the system is well warranted by the choice of a standard of proof informed by the preference-claim: as seen earlier, other things being equal this standard yields an expected number of errors that is higher than that yielded by the balance of probabilities standard.

So, how can one explain the widespread inconsistent endorsement of both the truth-claim and the preference-claim? Perhaps we are not being sincere after all. We – scholars and practitioners – repeat the truth-claim as a mantra not because we actually endorse the claim, but because we appreciate its legitimising power with respect to the enterprise of the criminal process: the simple, straightforward, teleological appeal to the truth gives the process – at least, in its ideal form – an aura of respectability. A more charitable explanation is that we are affected by a cognitive dissonance of sorts. We consider truth desirable, especially in the administration of criminal justice. Its desirability leads us to the view that the criminal process should aim for it, or the truth-claim. At the same time, we consider false conviction to be the worst outcome of the process by far, that is, we endorse the preference-claim. This leads us to defend a standard of proof that reduces the risk of false conviction at the expense of the overall accuracy of the system. Caught between these two weighty, and reasonable, considerations we somehow lose sight of their inconsistent implications. Yet another possibility is that we are simply inaccurate in our expression. We imprecisely refer to the truth as an aim when we genuinely mean to make a less significant practical commitment to it: perhaps we have a strong preference for true outcomes but, our choice of words notwithstanding, we don’t really see such outcomes, nor do we really mean to portray them, as the aim of the process. As I will show later, this less significant commitment to the truth is consistent with adopting a skewed standard of proof. Now, when it comes to the fundamental values of the criminal process, being precise in scholarly expression is surely not an idle, pedantic, exercise. Precision on such matters can only improve communication both within our community of scholars and with our external interlocutors.

The question to address is whether it is possible to correct the apparent misalignment between the commitment to truth-finding and the skewed standard of proof, such that the charges of insincerity, cognitive dissonance and inaccuracy could be avoided.

Two correctives to the truth-claim that don’t take us far

In the next section I propose a replacement for the truth-claim that avoids the tension with the preference-claim, while maintaining a significant commitment to the truth. Before doing so, it is worth considering and discarding interpretations of, or correctives to, the truth-claim that promise to achieve the same result.

Notice that a different understanding of the propositional content of criminal verdicts would allow to fulfil the truth-claim while endorsing the preference-claim. This is a possibility I have already criticised in the previous section, but I now want to explore further so as to strengthen the case for rejecting it. Perhaps verdicts should not be construed as making an assertion about whether the defendant committed the crime charged; rather, they should be construed as making an assertion about whether the available evidence is sufficient to satisfy the standard of proof for conviction. Under this construal the ‘guilty’ verdict would be stating that the standard is met, whereas the ‘not guilty’ verdict would be stating that the standard is not met. This understanding of the propositional content of criminal verdicts produces an alignment between the aim of truth-finding and the skewed standard of proof. A ‘guilty’ verdict asserts that the standard of proof for conviction is met and, if the assertion is true, it entails compliance with the standard: the defendant is convicted, as the standard requires. Similarly, a ‘not guilty’ verdict asserts that the standard of proof is not met and, if the assertion is true, it entails compliance with the standard: the defendant is acquitted, as the standard requires. Whether or not the standard is skewed in favour of the defendant, a correct application of the standard cannot ever frustrate the truth-claim: issuing the verdict required by the standard always involves making a true assertion.

One may raise doubts about whether criminal verdicts are understood in these terms by fact finders, lawyers, the system, or the polity.⁴³ Irrespective of these doubts, though, accepting this construal renders the truth-claim utterly uninteresting. To claim that the criminal process should be pursuing the truth about the satisfaction of the standard of proof, is essentially the same as claiming that it should be pursuing the correct application of the standard. Under the proposed corrective, the ‘guilty’ and the ‘not guilty’ verdicts make true assertions insofar as the standard is, respectively, satisfied and not satisfied; that is, insofar as each is the verdict required by the standard. This means that to aim for a true verdict is to aim for the verdict that is required by the standard of proof, that is, it is to aim for the correct application of the standard. The aim of truth-finding, thus, collapses into the aim of rule-compliance. As a result, truth-finding ceases to be a distinctive feature of the criminal process, or an interesting feature of the process for that matter. Indeed, any rule, or system of rules, that is created in good faith is created so that the rule(s) be complied with.⁴⁴

An alternative corrective to the truth-claim would be to argue that to claim that the criminal process should aim at the truth is really to claim that the process should aim at making an assertion of guilt if, and only if, the assertion is true. So redefined, the truth-claim apparently says nothing about assertions of non-guilt, or innocence. Therefore, it seems to avoid the problem raised in the previous section. Let me clarify. ‘Not guilty’ verdicts may well be compliant with the standard of proof but, since the standard is skewed in favour of the defendant, it is likely that these verdicts make a false assertion or – for those who understand verdicts in probatory terms – it is likely that these verdicts instantiate a failure of the system to recognise the guilty defendant as a criminal. If the truth-claim concerns guilty verdicts only – that is, if not-guilty verdicts are irrelevant to the satisfaction of the claim – then a skewed standard is apparently better than the balance of probabilities at fulfilling the claim. Other things being equal, under a more stringent standard it is more likely that, if guilt is asserted, the assertion is indeed true.

At a closer look, though, one notices a significant problem with the proposed reframing of the truth-claim. In such reframing, the truth of the assertion of guilt is treated as both a sufficient condition for the assertion to be due and a necessary condition for the assertion to be justified. If, however, truth is a sufficient condition for the assertion of guilt to be due, given the skewed standard of proof one reasonably expects there to be many cases in which the defendant is acquitted pursuant to the standard but the truth-claim is frustrated, since the sufficient condition is met: the defendant is actually guilty. In other words, false acquittals still frustrate the truth-claim under the proposed corrective, and this brings us back to square one. Redefining the truth-claim in terms of an obligation/aspiration to make an assertion of guilt if (and only if) the assertion is true, reproduces the issue of a misalignment between the truth-claim and a standard of proof informed by the preference-claim.

A possibility would be to treat the truth of the assertion of guilt only as a necessary condition for the assertion to be justified. Accordingly, the aim of the criminal process would be to respect this necessary condition, thus to make an assertion of guilt only if the assertion is true.⁴⁵ This means that false acquittals cannot frustrate the truth-claim: the fact that the defendant is guilty, after all, is not treated by the claim as a sufficient condition for the assertion of guilt to be due. Only false convictions frustrate the truth-claim since they involve an assertion of guilt in the absence of a necessary condition for the assertion to be justified. Notice, though, the paradoxical outcome of this redefinition of the truth-claim: a criminal justice system that were to acquit all defendants would be compliant with it. By acquitting all defendants, the system could never fall foul of the necessary condition for conviction: it would fulfil the aim of respecting such a condition.⁴⁶ Of course, a conscientious application of the available rules of adjudication would surely bring about some convictions, including convictions of the guilty. But an aim that can be satisfied by letting all guilty defendants go free is hardly a commitment to truth-finding; in any case, it surely isn’t the sort of commitment that scholars have in mind when they utter the truth-claim.

A replacement for the truth-claim: truth as preference rather than aim

My proposal for overcoming the tension between the truth-claim and the preference-claim is based on a structural feature that is widely shared by normative theories of the criminal standard of proof. Virtually all theories of the standard that one can find in the academic literature argue that the choice of the standard depends on the consideration of the values of the possible outcomes of the trial.⁴⁷ This is evident in accounts that rely on decision-theory and, therefore, identify the standard based on an interval scale where the four possible outcomes are arranged as a function of their respective value.⁴⁸ But it is also true of accounts that do not understand the standard of proof in decision-theoretic terms – that is, as a rule for the maximisation of expected value in each criminal case – but rather conceive of the standard as a rule to bring about a desirable distribution of the trial outcomes churned out by the system. According to these accounts, of course, the desirability of the distribution depends on the values of the outcomes that it contains; therefore, the standard of proof is chosen based on such values.⁴⁹ Finally, the values of the possible outcomes of the trial play a crucial role also in those accounts that try to resist the apparent consequentialist reasoning that informs the approaches mentioned so far.⁵⁰ In a nutshell, these last accounts argue that, because false conviction is a serious wrong – to put it with Dworkin, it inflicts a ‘moral’, as opposed to just a ‘bare’, harm –⁵¹ the State has a duty to do its best to avoid it; and they add that endorsing a stringent standard of proof (the reasonable doubt standard) is required by the State’s duty to do its best.⁵²

What all these accounts of the standard of proof have in common is that they are based on a value function, whether numerical or not, cardinal or ordinal. In other words, they attempt to justify the standard of proof by reasoning from a comparison of sorts between the values of the possible outcomes of the trial. Notice that those among the accounts surveyed that seemingly focus on the value of one outcome only are, in fact, concerned with the values of all outcomes. I am referring to the avowedly non-consequentialist accounts mentioned earlier, which give centre stage to the value of false conviction. To identify false conviction as the outcome that the State must do its best to avoid, these accounts must have first established that all other outcomes are better than false conviction, lest the argument is a non-starter. If any of the other outcomes were on a par with, or worse, than false conviction, why shouldn’t the State do its best to avoid this outcome as well, or instead? And what would this mean for the stringency of the standard of proof?

In previous work I have endorsed the decision-theoretic approach to the selection of the standard of proof.⁵³ Here, I am not interested in defending this approach against its competitors. I am interested, instead, in the common denominator of the main approaches offered in the academic literature: they all rely on a value function for the identification of the standard of proof. The important point to consider is that this is also true of the traditional – and mainstream – argument for a standard of proof skewed in favour of the defendant. Whether correct or not, this is the argument that one is likely to hear from any practitioner supportive of such a standard, if not from any lay person that has minimum familiarity with the criminal justice system. The argument draws on a revered version of the preference-claim, the maxim by William Blackstone according to which ‘it is better that ten guilty persons escape, than that one innocent suffer’.⁵⁴ The maxim expresses a comparative value judgement concerning the possible false outcomes of a trial. This judgement – the traditional argument goes – warrants a standard of proof that, compared to the symmetrical standard, decreases the likelihood of the worse outcome, even if at the cost of increasing the likelihood of the other. Because the worse outcome is false conviction, this is a standard of proof skewed in favour of the defendant: a standard that makes it harder to obtain a conviction than it would be under the balance of probabilities standard.

Perhaps the best-known example of this common justificatory strategy is found in Justice Harlan’s concurring opinion in In re Winship,⁵⁵ the celebrated decision of the US Supreme Court that laid the foundations of the reasonable doubt standard in the US Constitution. There, Harlan claims to derive the reasonable doubt standard from ‘an assessment of the comparative social disutility’⁵⁶ of the two erroneous outcomes of the trial – false conviction being, according to this assessment, the worse of the two.

A similar argument for a standard skewed in favour of the defendant is found in the leading text on English and Welsh criminal evidence, according to which ‘[t]he steeply asymmetrical criminal standard of proof ... is rationalized as reflecting the widely-held belief that the relative disutility of error tilts strongly against wrongful conviction in criminal trials’.⁵⁷ The preference-claim is also evident in the argument advanced in another influential text. Referring to the English and Welsh system, the text states: ‘[i]t is generally accepted that some … acquittals will be of persons who are in fact guilty of the offences charged, and who would be convicted if the standard of proof were the lower civil standard of the balance of probabilities. Such acquittals are the price paid for the safeguard provided by the ‘beyond reasonable doubt’ standard against wrongful conviction’.⁵⁸ To conclude this brief survey, notice that the preference-claim is implicit in what is probably the most famous judicial elaboration of the 'reasonable doubt standard'  in the English and Welsh system. In Miller v Minister of Pensions⁵⁹ Lord Denning says about proof beyond reasonable doubt:

It need not reach certainty, but it must carry a high degree of probability. Proof beyond reasonable doubt does not mean proof beyond the shadow of a doubt. The law would fail to protect the community if it admitted fanciful possibilities to deflect the course of justice. If the evidence is so strong against a man as to leave only a remote possibility in his favour which can be dismissed with the sentence ‘of course, it is possible but not in the least probable’, the case is proved beyond reasonable doubt, but nothing short of this will suffice.⁶⁰

A reasonable interpretation of the passage is that, according to Lord Denning, a standard of proof higher than the being sure of guilt, or reasonable doubt, standard ‘would fail to protect the community’ since it would lead too easily to an acquittal; and since many of these acquittals would likely be of guilty defendants. While a high rate of false acquittals lessens the deterrent effect of the law, it also means that individuals who are likely to recidivate are left free. According to Lord Denning, though, ‘nothing short of [the reasonable doubt standard] will suffice’, since a lower standard would increase the likelihood of false conviction to an unacceptable extent, this being the worse false outcome. As in the examples previously given, then, in Lord Denning’s argument too, the standard of proof results from an assessment of the comparative disutility of the two false outcomes; more precisely, it results from the preference-claim, the claim that a false conviction is worse than a false acquittal.

The traditional argument for a standard skewed in favour of the defendant, then, relies on a value function. Apparently, the function includes only the values of the false outcomes. The true outcomes, in fact, are not explicitly mentioned in the argument only because, in the value function on which the argument is premised, the values of these outcomes are sufficiently high; high enough to consider the outcomes desirable. In other words, the traditional argument is underpinned by the idea that, because true outcomes are desirable, they are unproblematic and, therefore, they need not play a central role in the selection of the standard of proof, notwithstanding that the standard will likely affect the rate of these outcomes as well. All the attention is given, instead, to the undesirable, false, outcomes. From a decision-theoretic perspective this makes very little sense. The decision-theoretic formula for the selection of the standard of proof indeed includes the values of all trial outcomes;⁶¹ and the attempts to rely on a simplified formula, including only the values of false outcomes, have been convincingly criticised.⁶² Be that as it may, even if the values of the true outcomes do not play a central role in the traditional argument for a skewed standard, they surely play a role in the background. Or – one may say – they do not play a central role in the argument precisely because they exhaust their role in the background. The argument is allowed to focus on the false outcomes by its proponents, since the preliminary construction of a complete value function has identified true outcomes as desirable and, therefore, unproblematic. In a warped world in which true outcomes were undesirable and false outcomes were desirable, Blackstone’s – perhaps I should say Whitestone’s – maxim and the argument built upon it would be about the former outcomes, not about the latter.

The value function on which the traditional argument for a skewed standard is premised, then, arranges the outcomes from most to least valuable according to either of these orderings: 1)true conviction; true acquittal; false acquittal; false conviction; 2)true acquittal; true conviction; false acquittal; false conviction.⁶³ Both orderings reflect the view that only false outcomes are undesirable and, hence, reflect the preference for true outcomes over false outcomes. Moreover, both orderings reflect the preference for false acquittal over false conviction. The latter is what I called the preference-claim. There is no incoherence in a value function that reflects both preferences. In fact, the plausibility of the two orderings – evidenced by the explicit support that they have received in academic works on the standard of proof –⁶⁴ arguably derives from the two preferences they reflect. Importantly, there is no incoherence between a preference for true outcomes and a standard of proof skewed in favour of the defendant, notwithstanding that the latter does not minimise the expected number of errors. In fact, decision theory – which, according to its supporters, effects coherent reasoning – can deliver a skewed standard based on a value function that reflects both the above preferences; that is, it can deliver a standard that does not maximise expected accuracy based on a function in which true outcomes are preferred to false outcomes.⁶⁵ Because in the traditional argument for a skewed standard of proof true outcomes only operate in the background, one may reasonably question whether, in this argument, a preference for true outcomes actually informs such a standard. Earlier I attempted to show that it does to an extent: the argument’s focus on false outcomes is allowed by an implicit initial evaluation of true outcomes as desirable. Those who are unpersuaded by this interpretation of the argument (and those who reject the argument itself) can look at decision theory for a much clearer illustration of how both the preference for true outcomes over false outcomes and the preference for false acquittal over false conviction can inform a standard of proof skewed in favour of the defendant.⁶⁶

It is in a value function reflecting these two preferences that lies the key to the alignment between the criminal process’s commitment to the truth and the preference-claim – the claim that, in any given case, false acquittal is preferable to false conviction. As seen earlier, if we understand the commitment to the truth in terms of the truth-claim (that is, in terms of an aim to reach true outcomes) there is no hope for an alignment. The alignment is realised, instead, once the truth-claim is replaced by understanding the commitment to the truth in terms of a preference for true outcomes over false outcomes. To be more precise, the commitment should be expressed as the requirement that, in the value function on which the selection of the standard of proof is premised, true outcomes be preferred to false outcomes. As just seen, such a function can also reflect the preference-claim without incurring incoherence: the alignment between the commitment to the truth and the preference-claim is, therefore, possible. Notwithstanding its apparent focus on the problem of selecting the standard of proof, the reworked commitment concerns the entire process. Before discussing this point, though, it is important to elucidate the difference between an aim and a preference and, therefore, between the truth-claim and what I have proposed as a replacement for it.

Aims and preferences entail different practical commitments. To have an aim is to have a commitment to achieve something. Failure consists in not achieving that which the agent is committed to achieve. To have a preference is to have a commitment to give due weight, in one’s deliberations, to the value judgement represented by the preference. Failure consists in not giving this value judgement such weight. To exemplify, compare the scenario in which an agent aims to use public transport for all their holiday trips, with the scenario in which the agent merely prefers to use public transport for all their holiday trips. If, in the first scenario, the agent uses their car to reach a holiday destination, they have clearly incurred a practical failure: they failed to achieve their aim. In the second scenario, instead, the agent may well reach a holiday destination with their car without incurring a practical failure. Perhaps the agent also prefers sunny holidays to rainy holidays; this preference is weightier than that about transportation; and, at the time of their annual leave, there is no holiday destination with a sunny forecast that the agent can reach with public transport. All things considered, therefore, the agent reasonably chooses to use their car to reach a sunny destination. A failure of practical reasoning would be incurred, however, if the agent in the second scenario chose a holiday destination without considering, or giving due weight to, their preference for public transport. Similarly, compare the truth-claim – according to which true trial outcomes should be an aim – with my proposed replacement for it – according to which true trial outcomes should merely be preferred to false outcomes. If true outcomes should be an aim (even if only one of several aims, then), any false outcome would represent a failure.⁶⁷ If, instead, true outcomes should merely be preferred to false outcomes, no failure would necessarily be incurred when a false outcome occurs. If the preference for true outcomes is given due weight in the selection of the standard of proof and if a correct application of the standard leads to a false outcome, the outcome would be in kee** with the practical commitment represented by the preference.

The commitment to the truth is, therefore, more modest under the proposed replacement for the truth-claim than under the truth-claim itself. While under the truth-claim the commitment is fulfilled in case of a true outcome and frustrated in case of a false outcome, under the proposed replacement the commitment is fulfilled insofar as the value function on which the standard of proof is premised incorporates the value judgement that true outcomes are preferred to false outcomes. Instances of false outcomes, therefore, do not frustrate the commitment to the truth. Insofar as the standard of proof is informed by a preference for true outcomes over false outcomes, instances of false outcomes produced by a correct application of the standard are in kee** with such commitment. The acquired modesty of the commitment to the truth, then, means that the commitment is also in kee** with the value judgement that informs the traditional argument for a skewed standard of proof, or the preference-claim. After all, as we have just seen, a value function can reflect this value judgement while also reflecting a preference for true outcomes over false outcomes. This is precisely the kind of value function on which the traditional argument for a skewed standard is based. To be sure, here I leave open the question whether the traditional argument indeed gives due weight to the preference for true outcomes. As said already, unlike the decision-theoretic approach, the traditional argument appears to give true outcomes a limited role in the selection of the standard, and this can be seen as unsatisfactory from the perspective of the reworked commitment to the truth. The important point, though, is that if the standard is effectively informed by a value function where true outcomes are preferred, there is no additional standpoint provided by the commitment to the truth from which to criticise the standard, or a false outcome that were to occur as a result of a correct application of the standard. As seen in section 4, instead, the truth-claim is in tension with the preference-claim, insofar as the latter is responsible for a skewed standard of proof. This is because the truth-claim treats the truth of the assertions involved in criminal verdicts as an aim, and a skewed standard is not the standard that minimises the expected number of false assertions; in other words, it is not the standard to choose in order to achieve that aim.⁶⁸

Notice that the proposed replacement for the truth-claim nonetheless maintains a significant commitment to the truth: a preference for true outcomes over false outcomes. Unlike under the truth-claim, this commitment to the truth is realistic, in that it can consistently coexist with the reasonable commitment to protect the innocent from conviction. Moreover, the proposed replacement eases the apparent tension, occurring in the phases preceding adjudication, between non-epistemic values and truth-finding. The criminal process should be underlain by a preference for true outcomes over false outcomes, or so the reworked commitment to the truth requires. Indeed, while so far I have referred to the value function relied upon for setting the standard of proof, this function is the set of preferences over outcomes of those who so defend that standard – be this the lawmaker or the theoretician. In other words, on pain of incoherence those who defend a given standard based on a given value function must also heed this very value function whenever it is relevant – as it surely is – to the regulation of earlier phases of the trial. Insofar as the preference for true outcomes over false outcomes informs the standard of proof, then, it should also inform the rules that regulate fact finding in earlier phases. And undoubtedly it does inform such rules, starting with the fundamental rules that irrelevant evidence is not admissible and that relevant evidence that risks biasing the jury should be excluded. However, as shown by the role of the preference-claim in the traditional argument for a skewed standard of proof, the rules of the process should not be driven exclusively by the value of truth. Other values should play a role in framing these rules and influencing their interpretation. This is the case for the phase of adjudication as it is for the previous phases of the process. In section 4 I mentioned a series of rules, governing the phases of the process leading up to adjudication, which are also informed by non-epistemic values. I have argued that, from the abstract point of view, none of these rules warrants the expectation that the application of the rule will lead to a decrease in the overall fact-finding accuracy of the system. I have, therefore, suggested that rules of this kind are not in tension with the truth-claim. To be sure, should there be rules which, like a skewed standard of proof, do warrant such expectation, they would be inconsistent with the truth-claim. The important point here is that, if one endorses the proposed replacement for the truth-claim, even these supposed truth-thwarting rules would be in kee** with the commitment to the truth insofar as they are justified by a reasoning where the preference for true outcomes has been given due weight. This preference may well be weighty; but, in the decision problem concerning how to regulate a particular issue (e.g., a question of admissibility), it may be less weighty than one or more non-epistemic values (e.g., trial fairness), in the same way in which the preference for public transport may be less weighty than the preference for a sunny holiday destination in the example given earlier. Insofar as the preference for true outcomes is given its due weight in determining the balance of reasons, the fact that it is ultimately outweighed by countervailing considerations surely does not mean that it is ignored, violated, disrespected. Rather, the preference is there as a constant commitment to reckon with whenever non-epistemic values would support the adoption of a truth-thwarting rule.

Concluding remarks

This article studied the widely accepted claim that the criminal process should aim at the truth. I called this the truth-claim. As is well known, the criminal standard of proof is skewed in favour of the defendant based on what I called, instead, the preference-claim. This is the judgement that false acquittal is better than false conviction. Due to being skewed, the standard does not minimise the expected number of errors. This number is minimised by a less demanding standard, the balance of probabilities. This means that the truth-claim is inconsistent with a skewed standard and with the underlying preference-claim. To dissolve this tension, I have first considered two correctives to the truth-claim, but to no avail. Then, I called attention to the fact that virtually all arguments for the selection of the criminal standard of proof are based on a value function, that is, a scale that arranges the possible outcomes of the trial as a function of their value. Notably, this is also true of the traditional argument for a standard of proof skewed in favour of the defendant. I argued in favour of replacing the truth-claim with the requirement that the value function on which the standard of proof is premised should express a preference for true outcomes over false outcomes. Because this preference can coexist with the preference of false acquittal over false conviction – indeed the two preferences do coexist in the value function relied upon by the traditional argument for a skewed standard – the proposed replacement avoids the tension characterising the relationship between the truth-claim and the preference-claim.

The gist of my argument is that, if someone seeks consistency between the commitment to the truth and the commitment to protecting the innocent from conviction, they should treat true outcomes as a preference on which the process is based, not as a/the aim of the process. Taking this more modest commitment to the truth also has implications for the regulation of other phases of the criminal process in which the value of truth may be in tension with non-epistemic values. I pointed out that the preference for true outcomes over false outcomes – indeed the complete value function – is likely to be relevant to the regulation of other phases of the criminal process; and I suggested that, being a preference rather than an aim, the possible tension between it and relevant non-epistemic values may well lead to adopting a truth-thwarting rule without disrespecting the commitment to the truth that the preference represents.

Of course, one may choose to overcome the tension between the truth-claim and the preference-claim by refusing to follow through on the implications of the latter; that is, by abandoning a skewed standard of proof and choosing the symmetrical standard instead. They might consider that their preference for false acquittal over false conviction is not sufficiently strong to warrant a departure from accuracy, after all. This option would secure normative consistency, but it would be an unpopular option indeed. Instead, understanding the commitment to the truth in terms of a preference for true outcomes over false outcomes would secure consistency without upsetting the mainstream sentiment about where the standard of proof should lie. While not as lofty as the truth-claim, this preference can still be a significant commitment, reverberating through the whole criminal process.