Must Jurors Know the Stakes of Conviction? Sentencing, Encroachment, and Legal Proof
J. Colin Bradley
Law, New York University, US
Eleanor Gordon-Smith
Philosophy, Princeton University, US
In US criminal trials, fact finding and sentence imposition are separate tasks. Juries are asked to determine whether the prosecution has proved the defendant met the constituent elements of an offense; they are not permitted to know the sentence the defendant would face if convicted. Generally speaking, sentencing is considered a question of law and therefore a matter for the court to decide subsequent to a finding of guilt.1
A recent view of legal proof due to Sarah Moss argues that this feature of criminal procedure rests on an epistemic mistake.2 According to this view, knowledge is the standard for legal proof of criminal guilt: a defendant is proven guilty if and only if the jury knows that the defendant is guilty. Further, there is pragmatic encroachment on knowledge: whether S knows that p depends in part on the stakes surrounding p. Whether a jury has legal proof of a defendant’s guilt thus depends on what’s at stake—for instance, the sentence she would receive. Hence, in order to enable juries to better recognize when they have met the standards for conviction, we ought to make a significant change to the design of criminal jury trials: we should admit evidence of the sentence a defendant would receive. Call this the Pragmatic Encroachment Argument for Sentencing Evidence, or PEASE.3
In this article we argue against PEASE. First, we show that allowing pragmatic encroachment on legal proof would produce a distorted and unjust picture of trial practice such that any view that finds itself endorsing pragmatic encroachment on legal proof should turn around and go back. It would allow an unjust proliferation of threshold-affecting stakes, such that all kinds of costs would turn out to affect the standards of legal proof; it would risk denying that having legal proof that a defendant fulfils all the elements of a crime means having legal proof of their guilt; and it would risk allowing considerations that are not evidence of defendant’s guilt to make it easier for a defendant to go to jail. Second, we show that even granting pragmatic encroachment on legal proof, this does not suffice to establish that sentencing information should be shared with juries, since the basic rationale for excluding sentencing evidence does not even need to deny pragmatic encroachment on legal proof. It rests on the judgment that sentencing evidence is substantially more likely to be prejudicial than probative. Nothing about pragmatic encroachment on legal proof ipso facto undermines this judgment. If we’re right, PEASE’s story of legal proof produces an unacceptable picture of trial practice and is in any case not decisive for whether to admit sentencing evidence.
In our view, the effort to render an argument for sentencing disclosure in pragmatic encroachment terms reflects a more general hazard in legal epistemology, whereby legal phenomena are flattened by the attempt to assimilate them to epistemic framings. Juries perform a complicated function in the criminal legal system: they play a rule-of-law role by facilitating the rule-guided and evenhanded disposition of cases on the basis of fact-finding; and they play a populist, democratic role by inserting the votes of laypeople between a prosecutor and the deprivation of a person’s life, liberty, or property. These functions are interrelated, but distinct, and how we should think about juries in light of these roles and the different values they reflect raises difficult questions of political morality.4 Treating questions of the jury’s role as problems for epistemology not only fails to solve these problems, as we argue in the case of sentencing evidence, but also runs the risk of evading the real issues.
Before we lay out our charges against PEASE in earnest, some scene setting. Criminal trial procedure in the United States obeys what Justice Thomas describes in Shannon v. United States as a “basic division of labor,” whereby a jury “find[s] the facts and decide[s] whether, on those facts, the defendant is guilty … without regard to what sentence might be imposed,” and judges “impos[e] sentence.”
Courts most frequently exclude sentencing information on the grounds that it is substantially more likely to be “prejudicial” than “probative.”5 Per the Federal Rules of Evidence, evidence is considered “prejudicial” when it carries an “undue tendency to suggest decision on an improper basis, commonly, though not necessarily, an emotional one.”6
Moss (we think accurately) summarizes the rationale behind this rule as follows:
The meager probative value of sentencing information is easily outweighed by a danger of unfair prejudice—specifically, the danger that mentioning the consequences of conviction will encourage jurors to decide their verdict on an improper basis, such as undue sympathy for the defendant.7
Existing trial procedure thus seems to endorse:
Judicial Defense of Sentencing Ignorance (JDSI): Jurors should not be informed of sentencing outcomes because the pro-defendant prejudicial effect of informing juries of sentencing outcomes outweighs the probative effect of that information for helping juries acquire knowledge of the defendant’s guilt or non-guilt.
The Pragmatic Encroachment Argument for Sentencing Evidence (PEASE) offers a compelling-seeming way to deny JDSI. The first claim in PEASE is that knowledge is the standard for legal proof, that is, a defendant is proven guilty if and only if the jury knows that the defendant is guilty.8 The second is that there is pragmatic encroachment on knowledge: whether a subject knows p depends in part on the pragmatic stakes associated with believing p, paradigmatically the costs of error of acting on p if p is false. We’ll return to the mechanisms of pragmatic encroachment in more detail later, but here is a classic motivating case:
Train
A and B are waiting on a train platform. A asks B “is this next train an express, or does it make all the local stops, like Foxboro?” B says: “It’s going to Foxboro.” C, who will miss a career-making job interview if she does not get to Foxboro on time, overhears the conversation.
According to pragmatic encroachment, B’s testimony may suffice for A to know the train goes to Foxboro. Since C’s career will be ruined if he misses the interview, that same evidence may not suffice to mean C has knowledge.9 Two subjects alike in evidence and beliefs about p can differ in whether they know p depending on whether their practical situation imposes different stakes. Costs of error associated with falsely believing p can drive up the threshold for how much evidence a subject needs to know p.
In combination, PEASE’s claims suggest a promising way to deny JDSI. More severe sentences will tend to drive up the costs of false conviction, so will (via pragmatic encroachment) drive up the threshold for knowing that the defendant is guilty. If knowledge is the standard for legal proof, juries could deliberate to a certain evidential standard and vote to convict on that basis when in fact the threshold for knowing the defendant was guilty was higher given the actual sentencing stakes. If the pragmatic encroachment story of legal proof is correct, juries in the US are frequently not in a position to correctly assess whether they have met the standards for conviction, since they cannot correctly assess the costs of error. To enable juries to more accurately recognize whether they are in an epistemic position to convict, juries should be informed of the sentencing stakes.10
Note, PEASE’s claim is not that a jury must know the sentencing stakes in order to know that the defendant is guilty. Rather, the claim is that a jury needs to know the stakes in order to assess whether they know that the defendant is guilty. Whatever legal proof amounts to, its conditions are presumably not luminous; there will be situations in which a defendant’s guilt has been proven to the relevant standard but the jury does not recognize this fact (or vice versa). PEASE holds that ignorance of sentencing stakes is one form of this mistake. Per Moss, “when the jury is ignorant of sentencing information, they may misidentify the conditions under which they ought to convict or acquit.”11
Here is our plan. In Section I we show why we should not allow pragmatic encroachment on legal proof. In Section II we show that the soundness of JDSI as criminal procedure policy does not turn on whether there is pragmatic encroachment on legal proof. We close by reflecting on the false allure of epistemology-first arguments for legal reform.
I. AGAINST PRAGMATIC ENCROACHMENT ON LEGAL PROOF
Our strategy in Section I is to argue that allowing pragmatic encroachment on legal proof would yield such a strange and unjust picture of trial practice as to suggest something must have gone badly wrong in an argument that yields that result. We won’t take a view on which is the culprit premise.12 We’ll give three arguments to this effect: the problem from proliferating costs, the problem from closure, and the problem from symmetry.
A. The Problem from Proliferating Costs
The first challenge we present to the viability of pragmatic encroachment on legal proof is the problem from proliferating costs. Here’s the overview: pragmatic encroachment on legal proof would permit an unjust proliferation of threshold-affecting stakes, and once we allow pragmatic encroachment on legal proof we don’t have a principled way to avoid this.
Everybody should agree that it would be unjust to allow some kinds of stakes to affect the standards for legal proof. Compare:
SIX: Juror is deliberating about whether Defendant is guilty. She assumes that if convicted, Defendant’s sentence would be approximately six months. In fact, the sentence is six years.
JOB: Juror is deliberating about whether Defendant is guilty. She assumes that if convicted, Defendant’s sentence would be approximately six months. In fact, an omniscient demon has also made it the case that if Juror falsely convicts, Juror will lose her job.
In both SIX and JOB, Juror underestimates the costs of a false conviction. However, even if stakes affect the standards for legal proof, we must want to say they can only make a difference in SIX, not JOB. Is there a principled way to sustain this verdict once we allow pragmatic encroachment on legal proof?
Within the pragmatic encroachment wheelhouse, costs of error interact with the knowledge threshold via “Knowledge-Action principles,” which posit a connection between what a subject knows and what it is practically rational for her to do. Knowledge-Action principles license inferences about whether S knows p via whether it would be rational for S to act on p given her evidence. There are variations on how exactly to spell out a Knowledge-Action principle:
“If you know that p, then it shouldn’t be a problem to act as if p.”13
“Where one’s choice is p-dependent, it is appropriate to treat the proposition that p as a reason for acting iff you know that p.”14
‘‘It is appropriate to rely on p in practical reasoning iff you know that p.”15
Costs of error enter this picture by altering the badness of acting on p if p is false. They wield their influence on the knowledge threshold by altering whether S can rationally act as if p (or treat p as a reason in some p-dependent choice, or rely on p in practical reasoning, and so on) given her evidence.16
Turn back to Train for an example. C’s evidence with respect to p (hearing B say “this train goes to Foxboro”) supplies C with a certain epistemic position with respect to p. Is this enough for C to act as if p; can she rely on p in a choice like whether to get on the train? That will depend on the badness of getting on the train if p is false. Say that C is .7 on p. Say that in both Case 1 and Case 2, the payoff of getting on the train if p is true is 10. In Case 1 the cost of getting on the train if ~p is -100, and in case 2, -1. Via a Knowledge-Action link, the fact that it looks fine by the lights of practical rationality to get on the train in Case 1 and not in Case 2 tells us that C knows p in Case 1 but not Case 2.
Pragmatic encroachment per se is relatively capacious on which kinds of costs of error are eligible to play this role. Costs of error interact with the standards for knowing p via the functional profile of acting given one’s evidence about p; where the candidate cost of error functions to make it practically irrational for S to rely on p given her evidence, a Knowledge-Action Principle will get the result that S doesn’t know p. Examples of threshold-altering costs in the literature range from fatal allergies,17 to missed career opportunities,18 to disasters in space exploration,19 to failing to pay rent.20
Back to legal proof and the challenge from JOB. If PEASE is to be read as a claim about knowledge, it will be hard to see how SIX differs from cases like JOB in an important way. We know from Train that losing one’s job is the kind of cost that can affect the knowledge threshold via traditional Knowledge-Action principles, so long as the costs of losing one’s job are sufficiently bad that they make it irrational for a subject to φ given her evidence about p. By parallel, there could be cases in which it would be practically irrational for a juror to vote to convict given her evidence if she would lose her job if she falsely convicts. (If you think losing her job isn’t likely to generate the requisite disutility, there would presumably be some dollar amount loss the demon could cause at which it would be practically irrational to vote to convict even on outstanding evidence). If the standard for legal proof is knowledge, and a Knowledge-Action link tethers knowledge to practical rationality, it looks like it will be possible for costs of error that accrue to the juror herself to alter the standards for legal proof. However, we take it two jurors alike in evidence but with different personal costs of false conviction ought not have different thresholds for legal proof of a defendant’s guilt.
But this is law, not ordinary knowledge. Perhaps PEASE can find some principled way to say that some costs of error are ineligible to affect the threshold for legal proof even though they can affect knowledge more generally.21
The obvious culprit in JOB is that the norms governing the appropriateness of “acting on p” once the action is voting to convict a defendant in a court of law are not the norms of practical rationality. Whether it is “a problem” or “appropriate” for S to act as if p (Defendant is guilty) by voting to convict is not determined by the balance of her own goals or the accrual of costs to her. The conditions under which S can appropriately vote to convict are questions of legal justification, not individual utility as in Train.
To capture this, perhaps PEASE could offer the following restriction. PEASE could invoke a knowledge-action link which uses a legal sense of what it would take to be “a problem” to act as if Defendant is guilty by voting to convict. Since the norms governing when it’s appropriate for S to vote to convict are norms of legal justification and not practical rationality, perhaps only those costs of error that affect the legal propriety of voting to convict given S’s evidence are eligible to target knowledge in legal cases. Perhaps PEASE needs:
Knowledge-Action Principle (Legal Redux): If S knows that the Defendant is guilty, then it shouldn’t be legally problematic for S to act as if Defendant is guilty by convicting.
Costs that accrue to the juror would be irrelevant to this second relation if not the first, and we would have a way to rule out JOB.
However, there are problems. First, whether it is legally problematic to vote to convict a Defendant given slim-to-no evidence about sentence length is precisely what’s at issue in the debate over sentencing evidence, so while this restriction can rule out JOB, it only gives us a question-begging way to rule in SIX.
Second, it’s not clear that even this move can prevent a strange proliferation of threshold-affecting stakes. One unfortunate upshot of inequalities across society is that jail sentences of the same length can cause different degrees of suffering depending on idiosyncrasies of defendants’ lives: the quality of their life outside incarceration, whether this is the first incarceration, to what extent will it destroy friendships, will it mean a defendant gets divorced, and so on.
We take it that the fact that it would be experientially worse for defendant B to go to jail for x amount of time than for defendant A ought not make a difference to how hard it is to prove them each guilty—a defensible theory of legal proof should not show that people with already bad lives are easier to prove guilty than the extremely fortunate because the costs of error associated with false conviction are higher for the fortunate. When convicted rapist Brock Turner’s mother wrote that her son—“Stanford boy, college athlete”—would not “survive jail,” nobody thought that the fact that Turner had higher heights from which to fall should be relevant to the standards of conviction.22 However, it’s not clear how PEASE could avoid counting these as threshold-altering stakes even once it moves to the Knowledge-Action Principle (Legal Redux). By the time we’ve held that there’s a scalar difference in the badness of false convictions because some are longer than others, and that that difference affects the legal justification for voting to convict, it’s not clear why forms of comparative worseness to do with unpleasantness should be ineligible to do the same.
Last, even supposing there is a principled way to only “count” costs of error to do with sentence length, the bigger problem is that pivoting to a legally specified version of a Knowledge-Action principle seems to reverse the direction of explanatory priority.
The force of PEASE’s original argument for reform was supposed to be that without sentencing information, juries miss the standards for conviction—the threshold they deliberate to can be too low given the actual stakes. PEASE seems to claim here that it’s because sentencing stakes affect knowledge that they affect whether the jury can appropriately convict Defendant, and so should be made admissible at trial.
However, to avoid the problem of proliferating costs, it now seems PEASE would have to claim that it’s because certain costs of error do not affect the appropriateness of convicting (and so ought not be admissible at trial) that they do not affect knowledge of guilt. PEASE would wind up making two claims that run in opposite explanatory directions:
It’s because a candidate cost of error affects knowledge of guilt that it adduces to the legal legitimacy of convicting and so ought to be admitted at trial (claim with respect to sentencing stakes).
It’s because a candidate cost of error does not adduce to the legal legitimacy of convicting and so ought not be admitted at trial that it does not affect knowledge of guilt (claim with respect to JOB, or scenarios in which defendant would get divorced or suffer unusually high personal costs from imprisonment).
This seems to generate ad-hoc equivocations on what determines what. Moreover, by the time PEASE’s claim is that juries should be given access to sentencing information because they have to be able to know defendant is guilty, but only on the basis of the evidence presented at trial, and only taking in those stakes that are appropriately admitted at trial by virtue of how legally justifiable it would be to treat those stakes as relevant to whether to convict, it’s no longer clear in what sense the dispute over whether to admit sentencing evidence turns on a claim about knowledge or stakes at all, and not just a first order claim about what should be admitted at trial—that is, a first order claim about whether jurors are legally justified voting to convict absent evidence of the sentence.
To sum the first problem: allowing pragmatic encroachment on legal proof would permit an unjust proliferation of threshold-affecting stakes, and PEASE does not have a principled way to prevent this.
B. The Problem from Closure
The second problem we present for pragmatic encroachment on legal proof is the problem from closure. Here is the outline: we can’t simultaneously hold that there is pragmatic encroachment on knowledge and that knowledge is closed under entailment. If there is pragmatic encroachment on legal proof, we should expect that legal proof isn’t closed under entailment. But legal proof should be closed under entailment. So we should refuse to allow pragmatic encroachment on legal proof.
Let’s begin by rehearsing the tension between pragmatic encroachment and closure under entailment. Closure comes in two varieties, single premise and multi premise. Single premise closure holds that if S knows p and S competently deduces q from p, S knows q. Multi premise closure holds that if S knows a set of propositions p1 through pn and competently deduces q from p1 … pn, then S knows q. It has been widely noted that pragmatic encroachment is in tension with multi-premise closure.23 The problem is that if p1 … pn jointly entail q, but q is higher stakes than p1 … pn,, the standards for knowing q will be higher than the standards for knowing p1 … pn such that S could know p1 … pn and competently deduce q, but nonetheless fail to know q given q’s stakes. Defenders of pragmatic encroachment are often happy to accept this result.24
If there were pragmatic encroachment on legal proof, we should expect the same result: it should be possible for a jury to have legal proof of a set of propositions p1 … pn, rightly recognize that p1 … pn jointly entail q, and yet—because q is higher stakes than p1 … pn—fail to have legal proof of q. Consider a case:
California
Defendant is tried for statutory rape in California, which under California Penal Code §261.5 has four elements: (a) Defendant had sex with another person, (b) That person was not Defendant’s spouse at the time, (c) That person is under the age of eighteen, and (d) Defendant is more than three years older than that person. The jury has legal proof of (a). Jury has legal proof of (b). Jury has legal proof of (c). Jury has legal proof of (d). Given §261.5, and (a), (b), (c) and (d), jointly entail that Defendant is guilty. However, the stakes are higher for believing that Defendant is guilty than for believing any of the propositions (a), (b), (c) and (d), since if the jury concludes that Defendant is guilty in error they will sentence an innocent man to jail for many years.
Say that p1 … pn are the propositions that Defendant fulfils (a) through (d) and that being guilty of statutory rape is a matter of fulfilling (a) through (d), and the q that they jointly entail is “Defendant is guilty.” Pragmatic encroachment on legal proof should be happy to say that where the proposition “Defendant is guilty” is higher stakes than p1 … pn because it alone attracts sentencing stakes, a jury could have legal proof that Defendant fulfils the elements of the crime and yet fail to have legal proof that Defendant is guilty. As a matter of law, however, we think it is important that legal proof deny this.
If the elements of a crime have been proven, it has been proven that Defendant is guilty.25 It would be a legal mistake for a jury to say: “Look, we grant that we have legal proof that Defendant satisfies (a), (b), (c), and (d)—but nonetheless, we don’t grant that Defendant is guilty.”26 This is the point of delineating the elements of a crime in statute and instructing jurors to return a guilty verdict if and only if the Defendant satisfies the elements of the crime.27
Ours is not just the descriptive claim that law currently does not allow violations of closure. It is the normative claim that it is good for legal proof to obey closure. In large part the purpose of defining the elements of an offence in statute is to use democratic procedures to create rules about the conditions under which any subject of the state may be found guilty of a given crime and subjected to coercion. It is important to the rule of law that these rules be produced by and applicable to all, that is, that individual assessors like judges not be empowered to set variable terms regarding which conduct suffices for guilt of which crimes. For substantially the same reasons that we want statutorily delimited elements of a crime to be that (and only that) which establishes guilt of that crime, we should also want our concept of legal proof to obey closure. Indeed, it’s hard to see what else the statutory claim that a crime has elements (a) through (d) could be except a guide to the conditions under which guilt of that crime has been established. It would be distinctly odd for a legal system to say (in statute) that being guilty of elements of crime C is a matter of satisfying elements (a) through (d), and to then say (when defining legal proof) that having legal proof of the elements of a crime need not amount to having legal proof of guilt.
Of course, sometimes in cases of terrible injustice we may want juries to say something close but not equivalent to a violation of closure, namely, “we grant that (a) (b) (c) and (d) have been legally proven–but nonetheless we acquit.” If a defendant would face capital punishment for littering, it may be that this is what a jury should do all things considered. We don’t deny the possibility of such cases. What we deny is that in such cases, the defendant has not been proven guilty to the standard of legal proof. Instead, these are cases of straightforward nullification, in which a jury has legal proof that the defendant is guilty but judges it to be morally good that he goes free anyway. Juries can always nullify. That juries have the liberty to nullify is an unavoidable feature of a true jury system in which the finding of guilt-determinative facts is indeed left in the jury’s purview.28 It is also part of why we use them in decisions about deprivations of life and liberty; they are “the conscience of the community”;29 with the “power to follow or not follow the instructions of the court.”30 Whether it would be good to nullify is not internal to the concept of legal proof, however. The issue at stake for whether legal proof should obey closure is what licenses the conclusion, legal-proof-wise, that the defendant is guilty of a given crime. It would be odd and anti-rule-of-law for the legal system to wholesale deny that legal proof should transmit between “Defendant fulfilled all the elements of a crime” and “Defendant is guilty of that crime,” since this would mean using democratic procedures to set out the elements of a crime only to then system-wide deny that legal proof of those elements has anything to do with legal proof of guilt.
A natural reply to the problem from closure would be to deny that there is a difference between the stakes of finding Defendant has satisfied all the elements of the crime and finding Defendant guilty. Perhaps precisely because fulfilling the elements of a crime suffices for guilt of that crime, the stakes for affirming (a), (b), (c), and (d) inherit the stakes of finding guilty. There would thus be parity between the stakes of p1 … pn and the stakes of the jointly entailed q (“he’s guilty!”), and pragmatic encroachment on legal proof could be reconciled with the important commitment that legal proof that a defendant has fulfilled the elements of a crime means legal proof of Defendant’s guilt.
How might PEASE establish parity between the stakes of p1 … pn and the stakes of q? It won’t do to say that the stakes surrounding each element of the crime should be considered in isolation. In California, for instance, the stakes surrounding (a)—whether defendant had sex with another person—are plainly not as high as the stakes for finding Defendant guilty, and it is hard to see how any of the other questions (whether Defendant and the sexual partner were married, how old they each were at the time) could attract the requisite costs of error if they are assessed independently from one another.
One strategy would be to hold that the stakes of each element ought not be considered in isolation, but ought to take account of that element’s place in the series. In California, for instance, if three elements have already been affirmed, the stakes of the remaining element will be the stakes of finding Defendant guilty, since affirming the last element would complete a set of propositions that together entail Defendant’s guilt. This might secure the parity in stakes required to avoid violating closure. But it gets an implausible picture of legal reasoning that law cannot sensibly endorse. For instance, if jurors consider elements in a different order from one another, their collective findings may not respect the sequential raising of the evidential bar. Juror 2 and Juror 3 may agree about whether Defendant had sex with the person in question but use quite different evidential standards depending on whether they take this answer to seal Defendant’s fate. To avoid falling short of the standards for legal proof in this way, juries would presumably need instruction on which element to consider last, as well as the fact that it should be treated as higher stakes given the sentence. This is a peculiar picture of trial practice, and in violation of a standard requirement on rational decision-making. In California, it would turn out that the highest evidential bar in the whole case—the proposition it is hardest for the prosecution to prove beyond a reasonable doubt—concerns which year Defendant was born.31
An alternative way to secure parity between the stakes of p1 … pn and the stakes of q would be to hold that jurors should deliberate over whether the whole conjunction block of (a), (b), (c), and (d) has been proven taking the stakes of the conjunction to be the stakes of conviction. This strategy has the best hope of dodging the problem from closure. But again, it produces a picture of trial practice that we should not endorse.
There is a live and difficult question in trial procedure about whether criminal juries should deliberate about each element in turn or about the whole conjunction of elements.32 Those in favor of having juries consider each element sequentially say this method increases juror understanding of and convergence with the law and gives affected parties access to the reasoning behind a defendant’s conviction or acquittal, which they hold is a requirement of justice.33 Those in favor of instructing juries to consider the whole conjunction of elements instead say that deliberations structured around each element interfere with a jury’s liberty to nullify and tend to be too favorable to prosecutions. Each side faces the additional problem that instructing juries to deliberate one way rather than the other disrupts their autonomy to select not only a verdict but a method of reaching a verdict. Whether to instruct juries to deliberate over each element or over the whole conjunction of elements (or neither) seems to us to turn on substantive justice considerations and empirical questions, including (i) the actual effects of different styles of deliberation, e.g. whether they tend to favor prosecutions, (ii) the relative normative weight of those effects, and (iii) whether defendants’ entitlement to a trial by a jury entitles them to a trial in which the jury determines for themselves how to deliberate. It would be a bad order of explanation if this controversial question in trial procedure needed to be resolved not by engaging with or balancing these considerations, but by noticing that—since there is pragmatic encroachment on legal proof—juries must deliberate over the conjunction of elements in order to avoid the problem from closure.
Second, in any event, PEASE should not want to go this way. Instructing juries to deliberate over the whole conjunction of elements given a stakes-sensitive concept of legal proof risks systematic overestimations of the standards of proof. Consider a charge of murder in the first degree. One element that would need to be proven is that Defendant killed a person. Another is that Defendant did so with malice aforethought. If the jury finds that the first element is satisfied, but the second is not, Defendant may be convicted of second-degree murder or perhaps manslaughter. Trials are frequently structured in just this way, such that one or two elements of a more serious crime would suffice to convict Defendant of a lesser included charge: first-degree murder trials prosecute questions of fact that would be sufficient for manslaughter, grand larceny for petit larceny, battery for assault, and so on.34 However, according to a pragmatic encroachment story of legal proof, the threshold for a manslaughter conviction should be lower than for a conviction of murder in the first degree, since the sentence will typically be less severe. So too for other lesser included charges. If the jury’s deliberation about whether Defendant killed a person takes place within the conjunction block for murder in the first degree and hence under the stakes for murder in the first degree, they will risk systematically overestimating the standards for legal proof required for convicting Defendant of the lesser included charge.
To sum the problem from closure: it’s very hard to reconcile pragmatic encroachment and closure under multi premise entailment. A view that posits pragmatic encroachment on legal proof should predict that, as in ordinary knowledge, a jury could have legal proof that a defendant has satisfied each element of a crime but not have legal proof of the criminal guilt that these jointly entail. Law does not and should not permit this possibility.35
C. The Problem from Symmetry
The final challenge we present is the problem from symmetry. The best epistemology is currently divided over whether pragmatic encroachment is a symmetrical phenomenon, that is, whether pragmatic considerations can lower the threshold for knowing p as well as raise it; can the costs of failing to believe p if true make it correspondingly easier to know p?36 In the legal domain, the analogous question is: If the costs of falsely convicting a defendant can drive the standards for legal proof up, could the costs of failing to convict a guilty defendant drive them down?37 The law cannot allow that result, and needn’t wait for the epistemology.
The stakes that drive up the threshold for knowing Defendant is guilty were supposed to be the costs of false conviction, i.e. convicting an innocent. The costs that lower the threshold would be the costs of false acquittal, i.e. failing to convict the guilty. Paradigmatically these include harm to the community if the defendant reoffends, miscarriages of justice, harm to witnesses and victims who are disbelieved, and so on. These costs will tend to be more serious the more serious the crime: the badness of acquitting a rapist exceeds the badness of releasing a small-time drug dealer. Hence if the costs of false acquittals can lower the threshold for conviction, we may wind up with the counterintuitive result that more severe crimes have lower standards for conviction. Sometimes this effect will be cancelled by the stakes-raising fact that more serious crimes tend to carry more serious sentences, but this will not always be available. Symmetrical encroachment on legal proof could in principle allow situations like:
Double Defendants
A is on trial for murder in Montana. B is on trial for trafficking marijuana in Texas. Because of discrepancies in sentencing law, A and B each will be sentenced to 10 years in prison if convicted. The cost of acquitting A if guilty would be .8. The cost of acquitting B if guilty would be .2.
Given the sentencing stakes are the same, PEASE holds the cost of false conviction would be the same for A and B. Given A’s crime is more serious than B’s, it would be more costly to falsely acquit A than B. However, we take it that it would be bizarre for law to hold one needs less evidence to prove A guilty of murder than to prove B guilty of drug trafficking.
Of course, PEASE could deny that the result in Double Defendants is all that peculiar; it could hold that what it takes to have “proof beyond a reasonable doubt” can drop with the stakes of false acquittal. However, it’s worth noting just how radical a claim this would make PEASE. It is one of the most widely accepted and constitutionally protected elements of the law that just conviction requires proof beyond a reasonable doubt (“BARD”). The Supreme Court in Winship calls BARD “that bedrock axiomatic and elementary principle” that realizes the presumption of innocence.38 When explicating what BARD amounts to, courts do not speak in ways compatible with the suggestion that BARD could be lowered on pragmatic grounds—or even lowered at all.39 Every court in the United States presently operates on the standard of proof that considerations that have nothing to do with whether Defendant committed the crime cannot make it easier to send them to jail. It would abdicate the spirit of BARD to say, “look, we agree guilt must be proven beyond a reasonable doubt, but given the stakes of acquitting, one needn’t have very much evidence to meet that standard.”
If PEASE was the right guide to the standards for legal proof, we should expect the answer about whether legal thresholds can be lowered as well as raised to turn on final epistemology about whether pragmatic encroachment is symmetrical. But it seems considerations of justice intervene.
II. PRAGMATIC ENCROACHMENT ON LEGAL PROOF WOULDN’T UNDERMINE JDSI
Let’s check in with the dialectic. In Section I, we argued that the conclusion that legal proof is subject to pragmatic encroachment leads to unacceptable implications for criminal trial procedure. In Section II, we’ll now turn to argue that PEASE misses its target in any event: even if there were pragmatic encroachment on legal proof, this would not of itself provide reason to give up on JDSI, for the soundness of JDSI as policy in criminal trials is independent of whether there is pragmatic encroachment on legal proof.
As we’ve seen, a fundamental principle of American evidence law is that evidence should not be admissible at trial if the probative effect of admitting that evidence would be outweighed by the prejudicial effect of admitting it, that is, if the fact-finding value of that evidence is outweighed by its tendency to prejudice juries. JDSI is a straightforward application of this general rule. As a refresher, JDSI holds:
Judicial Defense of Sentencing Ignorance (JDSI): Jurors should not be informed of sentencing outcomes because the pro-defendant prejudicial effect of informing juries of sentencing outcomes outweighs the probative effect of that information for helping juries acquire knowledge of the defendant’s guilt or non-guilt.
US courts distinguish between “relevant” and “irrelevant” evidence when adjudicating what is admissible at trial. Relevant evidence is defined in the Federal Rules of Evidence as evidence that “has any tendency to make a fact [of consequence to the charge] more or less probable than it would be without the evidence.”40 Sometimes courts exclude sentencing evidence not via JDSI but by holding that sentencing evidence is “irrelevant” and hence categorically inadmissible. If evidence is simply irrelevant, there is no need to consider how prejudicial it might be. However, JDSI does not say that sentencing evidence is irrelevant. JDSI excludes sentencing evidence on the basis of a balancing test, concluding that it is more likely to be prejudicial than probative. Importantly, prejudicial evidence is a subset of relevant evidence: it is evidence that bears on a fact to be proved at trial but that nevertheless carries an undue tendency to suggest decision on an improper basis. Consider a particularly gruesome autopsy photo. It is relevant to what needs to be proven—the victim’s brain was shredded by a bullet—but admitting it as evidence may be more likely to induce revulsion in the jury than to help the jury determine the facts. When courts say sentencing evidence is inadmissible because its probative effect (its relevance) is outweighed by its prejudicial effect, this is what they have in mind, as when the Shannon Court worried that “providing jurors sentencing information … creates a strong possibility of confusion.”
The core defense of JDSI therefore does not even need to deny the possibility of pragmatic encroachment on legal proof. The core defense of JDSI is that sentencing evidence is more prejudicial than it is probative: it is an institutional judgment about trade-offs. A defender of JDSI does not need to claim that sentencing evidence is not probative at all. All she needs to hold is that, however probative sentencing evidence is, it is less probative than it is prejudicial. It’s therefore consistent with JDSI that sentencing information would make a difference to the standards for legal proof in just the way that the pragmatic encroachment model describes. All JDSI claims is that the probative value of this sentencing information does not make up for the pro-defendant prejudicial effects of allowing juries to know about possible sentences.
PEASE, therefore, does not of itself provide an argument against JDSI. For that, PEASE would need to say something about the interaction between the probative value of sentencing information and the tendency to prejudice juries. We can see two ways a defender of PEASE could discharge this burden and turn her argument into a denial of JDSI. First, she could argue that pragmatic encroachment on legal proof shows that the probative effect of sentencing evidence at trial is higher than has been historically imagined. Second, on the other side of the trade-off, she could argue that the pro-defendant prejudicial effect of sentencing evidence is not as high as JDSI supposes. When sentencing evidence leads a jury to acquit where it otherwise would have convicted, she could say, we are not observing pro-defendant prejudicial effect, but a properly epistemic effect. On this approach, the probative effect of sentencing evidence outweighs the prejudicial effect because the latter is de minimis.
We think each of these strategies fails and will take each in turn.
A. Overestimating the Probative Effect of Pragmatic Encroachment
The first strategy for defenders of PEASE to show that the probative effect of sentencing information outweighs the prejudicial effect is to hold that in light of pragmatic encroachment on legal proof, the probative effects of sentencing information are much higher than has historically been imagined.41
However, even granting pragmatic encroachment on legal proof, the probative effect of sentencing evidence is likely to be marginal. Only in a narrow range of cases is it likely that sentencing evidence could be outcome determinative, i.e., that the actual sentencing stakes would make the difference between a jury knowing that a defendant is guilty and lacking knowledge that the defendant is guilty. It is true that juries regularly underestimate the sentences that defendants face, in part because of mandatory minimums, three-strikes laws, or other disproportionately high sentences for minor offenses. But we doubt it is routine or even all that plausible in practice that this makes the difference in whether a jury has cleared the knowledge threshold or not.
First, juries already know that stakes are high, even if they are mistaken about exactly how high. Plenty in the ceremony of jury trials—the judges’ robes and gavel, the bailiff’s handcuffs, the judge’s instructions to the jury, the oath jurors swear—indicate the seriousness of the affair. In criminal trials, every juror has sufficient reason to know that the forced deprivation of liberty is a likely outcome of a guilty verdict, and that sending someone to jail for any amount of time is highly consequential.
Compare this to a classic case in which underestimated costs of error make a difference to what a subject knows. S—a lover of chocolate and carob—hurriedly purchases a cake at the supermarket for a dinner party. She takes herself to know that the cake is made with chocolate, not carob, based on the statistical prevalence of chocolate cakes over carob and the sight and smell of the cake. This might ordinarily suffice for knowledge. However, if a guest at the party has a deadly carob allergy, it may be that S’s evidence is inadequate for knowledge. The mechanism here as in many other stakes-shifting cases is that the threshold for knowledge given the stakes is significantly higher than the threshold to which the subject deliberated. There is a wide gulf between the threshold to which she deliberated and the knowledge threshold, and the wide gulf makes it relatively likely that any given evidence set available to S will fall somewhere between the two thresholds such that S would knowLOW STAKES but does not knowHIGH STAKES. Unless she realizes she is in a high stakes case, she is quite likely to be mistaken when she takes herself to have knowledge.
However, given all jurors know that incarceration is likely at stake, criminal trials are different. When we say that a juror is “ignorant of the stakes,” we are typically pointing to a smaller distance between the threshold she deliberated to and the actual threshold given the actual stakes. The difference between assumed stakes and actual stakes is a difference between high stakes and higher stakes. It is less likely that a given evidence set available to a jury would surpass the threshold given the presumed high stakes, but fail to clear the threshold given the higher stakes. S’s underestimation of the stakes is less likely than in ordinary cases to make the difference to whether she cleared the knowledge threshold and she is less likely to be wrong about whether she has deliberated to the appropriate standard.
Compare:
One might worry: aren’t there cases where the gap between high stakes and higher stakes in criminal trials is just as wide as in traditional cases–for example, if the presumed sentence is one year but the actual sentence is life? Such cases are possible, but not the paradigm.42 Moreover, even where there is a large gap between the presumed sentence and the actual sentence, it is not clear that a large difference between high stakes and higher stakes means a correspondingly large difference between knowledge thresholds. Past a certain point, additional increases to already-high stakes have diminishing marginal impact on the threshold required for knowledge. When the guest has a fatal carob allergy, this is a high stakes case: it gets harder to know whether “this cake is chocolate.” Do things go all that differently under higher stakes? Suppose that in addition to having a fatal allergy, S’s guest is also on the cusp of a breakthrough discovery in her cancer lab, and also has a son who would be without a caregiver if she died, and so on, such that her death would be really spectacularly bad. How much more evidence does S need for knowledge; how many more times must she double-check? It seems that as the stakes increase from very high to even higher, additional stakes can have diminishing marginal impact on the knowledge threshold.
If so, then underestimating the exact length of the sentence need not mean proportionately underestimating the standards required for proof. Because in all criminal trials jurors already have sufficient reason to know that the stakes are quite high, if they are deliberating responsibly, they will already be deliberating to a very high knowledge threshold. Correctly determining that the stakes are higher rather than high is unlikely to make a difference to whether the jury has deliberated to the appropriate standard. So even if there is pragmatic encroachment on legal proof, the probative value of that effect will likely be negligible given that juries already deliberate to high-stakes thresholds and given the diminishing effect of additionally high stakes on an already-high knowledge threshold.
B. Prejudice, Nullification, and Courts
If sentencing evidence is unlikely to be probative to such an extent that it falsifies JDSI, the other way PEASE might threaten JDSI is by denying that the prejudicial effects of admitting sentencing evidence are significant.43 When jurors are presented with sentencing evidence, they are more likely to vote to acquit. While this may appear to be a prejudicial effect, PEASE could argue that we should construe this as a properly epistemic effect instead: when jurors refuse to convict upon receiving sentencing evidence, it is because of the interaction they judge to be occurring between that evidence, the rest of the evidence at trial, and their ability to know the defendant is guilty.44
Consider the fact that jurors sometimes express doubts and misgivings once they learn the actual sentence imposed after a guilty verdict. Moss cites the example of a juror who voted to convict a thirteen-year-old boy of felony murder without knowing that there was a mandatory sentence of life without parole. The juror later learned of the mandatory minimum and remarked: “I heard on the radio that he would be spending the rest of his life in prison. I didn’t know that was mandatory.”45 The following character seems to us to represent a real-life phenomenon:
Regretful Juror:
Juror voted to convict Defendant. Juror believed at the time of conviction that it was probable that Defendant would be sentenced to five to ten years in prison. Later, Juror learns that Defendant’s conviction was subject to a mandatory sentence of life in prison without the possibility of parole. If Juror had known the actual sentence, he would not have voted to convict Defendant.
How should we interpret Regretful Juror and real-life cases like it? One possible interpretation is that the juror experiences epistemic dissatisfaction with her verdict in the same way that we may want to “take back” epistemic judgments or assertions once we learn of additional stakes. Such an interpretation of Regretful Juror and related phenomena would allow PEASE to explain how what appear to be significant prejudicial effects are in fact legitimately epistemic effects. This would require the following interpretation of Regretful Juror:
New Doubts:
In light of the evidence presented at trial and the assumption that the sentence is five to ten years, Juror takes herself to have knowledge that Defendant is guilty. If Juror had known that a guilty conviction would automatically sentence Defendant to life in prison without parole, then Juror would no longer have taken herself to know that Defendant was guilty, and Juror would have voted to acquit.
But consider a different possible interpretation of Regretful Juror:
Nullification:
In light of the evidence presented at trial and the assumption that the sentence is five to ten years, Juror takes herself to have knowledge that Defendant is guilty. If Juror had known that a guilty conviction would automatically sentence Defendant to life in prison without parole, Juror would not have voted to convict, because she believes this would be a grossly unjust outcome.
We submit that Nullification is a more plausible interpretation of Regretful Juror cases than is New Doubts. In Nullification there is no suggestion that Juror ceases to take herself to know that the defendant was guilty. She simply does not believe that it is just to impose such a harsh sentence and does not want to participate in what she takes to be an injustice. We think it is more natural to interpret Juror (here and in Moss’ cited cases) as expressing regret that her vote to convict had the harsh outcome it did. Juror can experience this kind of moral regret without downgrading her epistemic position on Defendant’s guilt.
We do not need to deny that something properly epistemic like New Doubts ever happens. But we think it is implausible as a dominant interpretation of what happens in Regretful Juror cases. At a minimum, if PEASE’s strategy for arguing against JDSI is to argue that the pro-defendant prejudicial effects of sentencing evidence are minimal, we would need considerable empirical evidence that New Doubts and not Nullification is the right account of what happens in cases like Regretful Juror. A critic of JDSI can still argue that such effects are good, all things considered. We take no stance on whether nullification is good. Our point is only that this would be a very different strategy than the one PEASE promises. Even granting pragmatic encroachment on legal proof, the question whether sentencing evidence should be admitted at trial threatens to collapse into the question whether it would be good if juries engaged in more nullification; that is, if they more often exercised their democratic function over their rule-of-law function. This is precisely the kind of debate in political morality that PEASE was designed to avoid.
PEASE, on its own, is silent on JDSI, which is a balancing claim. So even granting pragmatic encroachment on legal proof, PEASE does not supply an argument against JDSI. Further, resources internal to PEASE look unlikely to provide that argument by showing that either (a) the probative value of sentencing information is significantly higher than has been imagined or (b) the prejudicial effects are minimal.
III. CONCLUSION
PEASE is an epistemic argument for substantive legal reform. We see the appeal of such a strategy. As fact-finding bodies, juries are bound by epistemic norms. But the fact that we have juries at all—the fact, for instance, that we entrust fact finding not to experts or to witnesses with first-hand knowledge but to people specifically chosen to have neither—reflects much else besides epistemic propriety. The adoption of the jury was largely the outcome of a centuries-long series of disputes over who should hold power over life, limb, and property.46 The jury has long been seen as a crucial democratic safeguard against the overreach of the King, the federal government, and the local prosecutor by inserting a popular institution into the judicial process.47 Accordingly, the jury, like other legal institutions, manifests a constant push and pull between several sets of values: the democratic values of popular decision-making and representation, and rule-of-law values including the importance of accurate, consistent applications of law. For example, debates over reforms meant to make nullification more or less likely, or over the usefulness of “special verdicts” in criminal trials reflect different judgments about the relative value of jury power and autonomy versus adhering to the rule of law or the primacy of legislative decision-making. Special verdicts may increase uniformity and predictability, but at the cost of jury autonomy in deliberation; empowered juries can nullify more often, but at the cost of legislative control over defining and punishing offences.
There is no doubt something attractive about the effort to express the admissibility of sentencing evidence in criminal trials as primarily a problem of and for epistemology. Framing the issue as one that implicates only the epistemic function of juries seems to allow us to intervene in practical debates about legal reform without having to address head-on these long-running debates of political morality. The question then appears to be one of facilitating juries in their core fact-finding task, and everyone agrees that whatever else juries do, they are fact finders. It would be an elegant argument to show that it turns out that to perform this core, uncontroversial task, juries must have access to the same information they would get if the point were merely to give them greater power. On such an argument, what appears to be a question primarily of political morality—the division of labor between courts and juries in the criminal trial—can be resolved on epistemic grounds. Though specific epistemic positions and arguments will remain contestable, they are more tractable than related moral disputes, and so the epistemic framing itself may seem to constitute progress. Moreover, an epistemology that accepts pragmatic encroachment on knowledge may seem especially well-suited to treating the admissibility of sentencing evidence as a properly epistemic problem.
But we have argued that this effort is unsuccessful. Pragmatic encroachment on legal proof produces unacceptable results for trial practice and in any event does not provide an argument against JDSI.
If law were in the pragmatic encroachment game, that would give us a straightforward argument that American trial practice should be more concerned with its costs of error than it currently is, given the extravagant lengths of its sentences. One of the attractive features of pragmatic encroachment is that it humanizes knowledge. Opposing a tradition according to which evidence is all that counts, encroachment connects knowledge to things that matter to us deeply, like the gravity of certain errors or the horror of certain costs. Fortunately, in law, unlike in knowledge, we did not lack for a way to bring those considerations in absent the encroachment story. It’s perfectly intelligible to criticize legal policy for having bad effects, imposing unacceptable harms or horrifying costs. Opponents of JDSI should make a justice case on justice terms. Those problems need not tamper with knowledge to matter.
Notes
- There is an exception for capital trials, where juries may be given sentencing responsibilities. See Simmons v. South Carolina, 512 U.S. 154 (1994). [^]
- Moss 2021; 2022. [^]
- We’ll talk about what “a jury” knows, though we don’t take a view on whether a jury is a group agent. Following Moss, we’ll also talk about disclosing “the” sentence, though since it would be impractical to hold sentencing hearings before trials juries would more likely be told the range of possible sentences including mandatory minimums or maximum. [^]
- The role of juries was a recurring sticking point in ratification debates over the US Constitution. One leading Anti-Federalist spokesperson made the characteristic argument that juries “secure to the people at large their just and rightful controul in the judicial department” (Federal Farmer 1788/2004). Tocqueville 1840/1966 later highlighted this in his influential description of the political role that juries played in American democracy. See Alschuler and Deiss 1994. [^]
- For helpful discussion of the legal rationale see Bellin (2010, pp. 2237–42). [^]
- Federal Rule of Evidence 403, cmt. [^]
- Moss 2021, p. 261. [^]
- For views on which legal proof is or requires knowledge see: Moss 2022, p. 2; Thomson 1986; Littlejohn 2020; and Smith 2018. [^]
- Train is adapted from Fantl and McGrath (2002, p. 87). Other canonical defenses of pragmatic encroachment include: Stanley 2005; Fantl and McGrath 2007; Schroeder 2012; and Ross and Schroeder 2014. There are a variety of debates within the pragmatic encroachment literature about (a) whether costs of error are the only costs to affect knowledge thresholds and (b) the mechanism by which the costs of believing p shift the knowledge threshold; see Worsnip 2020. We will set aside these finer details. An analogous position is moral encroachment, according to which the moral stakes of the belief p (e.g. the risk of thereby wronging someone) can affect its epistemic credentials. See: Basu 2020; Moss 2018; and Jackson and Fritz 2021. One can imagine an analogous argument that there is moral encroachment on legal proof since plausibly it would wrong S for a jury to find her guilty if she is innocent. Everything we say with respect to pragmatic encroachment should import to this argument also. [^]
- Since Moss (2021; 2022) is the architect of this recent and novel contribution in legal epistemology, we will engage substantially with those papers. However, we intend our argument to extend to anyone who thinks a rejection of JDSI can be pulled from the premises that there is pragmatic encroachment on knowledge and the standard for legal proof is knowledge, e.g. Biebel (2023, p. 53), Epps and Ortman (2023, pp. 837–838). [^]
- Moss 2021, p. 260. [^]
- For arguments against knowledge as the standard for legal proof see: Lackey (2021), who argues that knowledge is neither necessary nor sufficient for legal proof; Smartt (2022), who argues the view cannot sustain the right results in probabilistic cases; and Enoch, Fisher and Spectre (2021) who argue it condemns legal systems to objectionable “knowledge fetishism.” For arguments against pragmatic encroachment on knowledge see Kvanvig (2011). [^]
- Fantl and McGrath 2002, p. 71. [^]
- Hawthorne and Stanley 2008, p. 9. [^]
- Brown 2008, p. 1137. [^]
- One wants to be a little careful about what it means to “act as if p” or “rely on p in practical reasoning,” etc. The fact that S knows p cannot license any way of acting on p, since there are actions that p would not rationally support even if true; for discussion see Worsnip (2020, p. 7), and Ichikawa (2012). If S believes that vaccines will protect her from polio, and on the basis of that belief protests outside CDC headquarters, the fact that she should not protest does not tell us that she does not know that vaccines protect from polio. The problem with her “acting on p” is not an epistemic defect but that the fact that there is no rationally approvable connection between p and this mode of acting. So strictly, Knowledge-Action principles should be understood as positing connections between S knowing p and it being not a problem for S to act as if p by φ-ing, or relying on p in practical reasoning as a reason for φ-ing, where p would, if true, rationally support or legitimize φ-ing, and so on. Various pragmatic encroachers have ways of spelling out this connection. For brevity, we’ll use the shorthand of “acting on p” or “relying on p,” since we’re happy to spot PEASE the idea that where p is “the defendant is guilty,” and φ-ing is voting to convict, p would, if true, bear the right rationalizing connection to φ-ing (though see Ichikawa 2012, pp. 52–53). [^]
- Ross and Schroeder 2014, p. 261. [^]
- Fantl and McGrath 2002, p. 67. [^]
- Schroeder 2012, p. 281. [^]
- Stanley 2005, pp. 3–4. This capaciousness has been something of stalking horse for pragmatic encroachment: if what matters for whether a candidate cost of error can affect whether S knows p is that it targets the practical rationality of actions relying on p, it’s difficult to say why allergies or rent are in, but demons and infinite Pascalian rewards are out. See Worsnip 2020. [^]
- Moss makes a similar restriction when arguing that the claim the standard for legal proof is knowledge means jurors must know Defendant is guilty only on evidence admitted at trial (Moss 2022, p. 2, note 2). [^]
- Levin 2016. [^]
- At least, and some have recently defended a tension with single premise closure also (see Anderson and Hawthorne 2018; 2019). We’ll focus on multi premise closure, since most cases of legal fact finding are multi premise. [^]
- Indeed, many deny that knowledge is closed under multi premise entailment on independent grounds, see e.g. Zweber (2016). [^]
- For simplicity, we leave aside the issue of affirmative defenses. [^]
- Evidence that this would be a legal mistake is that in a civil trial, this would be a mistake that a judge could overturn by issuing a judgment notwithstanding the jury’s verdict. [^]
- To be clear, the claim here is that fulfilling the elements of a crime suffices conceptually or logically for guilt of that crime, not the claim that criminal juries should deliberate over each element one by one as they do in “special verdicts.” We’ll return to this issue shortly. [^]
- Nullification is what Philip Pettit calls a “spandrel right” (Pettit 2023, p. 183). No positive provision of law specifically protects the jury’s right to nullify, but any system that leaves the jury undisturbed as the ultimate finder of fact cannot avoid providing the jury with the liberty to nullify. [^]
- United States v. Spock, 416 F.2d 165, 192 (1st Cir. 1969). [^]
- United States v. Ogull, 149 F. Supp. 272, 276 (S.D.N.Y. 1957). [^]
- We think this looks equally funny whichever element the jury is instructed to consider last. [^]
- One acute version of this arises for legal probabilism i.e. the view that legal proof requires proof over a certain probability: it can easily be the case that the probability of each element is over the threshold while the conjunction is not (Hedden and Colyvan 2019; Schwartz and Sober 2017). If probabilism is the right account of legal proof, the legal system will have to say either that legal proof of each element is sufficient for guilt, or that proof of the conjunction is necessary, but not both. This is largely orthogonal to PEASE’s knowledge account of legal proof, since it’s a lot more orthodox to hold that if one knows a and knows b one knows a^b. [^]
- See Sood 2021; 2023. [^]
- The importance of this procedural safeguard has been recognized by the US Supreme Court, which held in Beck v. Alabama, 447 U.S. 625 (1980), that defendants in capital murder trials have a constitutional right to a lesser included offense jury instruction. [^]
- One need not have the same view on closure in legal proof and knowledge. We can and do decide which account of legal proof to prefer in light of its procedural and democratic upshots. If one thinks that knowledge as an analysand does not share this feature, one may reject that knowledge is closed under entailment while affirming that legal proof should be. This section would then suggest grounds to deny that knowledge is the standard for legal proof. [^]
- See e.g.: Jackson and Fritz 2021; Basu 2020, p. 205; Gordon-Smith forthcoming; von Klemperer 2023; and Crewe and Ichikawa 2021. [^]
- Moss (2021, pp. 269–270) expresses openness to this possibility. [^]
- In re Winship, 397 U.S. 358, 363 (1970). [^]
- Some explicate BARD in credal terms, e.g. McCauliff (1982). Others reject numeric thresholds and hold BARD is a function of rational or deductive upshot, e.g. “[such] proof which is wholly consistent with the guilt of the accused, and inconsistent with any other rational conclusion,” Corpus Juris Secundum, vol. 23, Criminal Law, §1502 (quoted in Gardiner 2019, p. 298). Laudan (2011), who does defend the revisionary claim that costs of falsely acquitting repeat offenders could lower the threshold for conviction in second-offence trials, accepts that this would mean rejecting BARD. [^]
- Federal Rule of Evidence 401. [^]
- We take it this is Moss’ primary strategy. Though Moss (2021, p. 261) does not make explicit how PEASE is meant to deny JDSI, she characterizes current defenses of JDSI as appealing to the thought that “the meager probative value of sentencing information is easily outweighed by a danger of unfair prejudice—specifically, the danger that mentioning the consequences of conviction will encourage jurors to decide their verdict on an improper basis, such as undue sympathy for the defendant.” The bulk of her argument is devoted to showing that sentencing stakes can enter into jury deliberation not by inducing “undue sympathy” but instead by raising the threshold for knowledge. [^]
- As an empirical point, many costs of incarceration (disfranchisement, the ruling-out of certain employment and housing opportunities, stigma, trauma, and so on) are realized on the first day of one’s sentence, and empirical work suggests that the first year of incarceration is experientially the hardest (Bronsteen, Buccafusco, and Masur 2009), such that the move from “no sentence” to “some sentence” represents the largest leap in stakes. [^]
- It is important to distinguish this strategy from a strategy that engages with JDSI over the moral valence of pro-defendant prejudicial effects. For example, one thing a critic of JDSI might argue is that JDSI mistakenly assumes that pro-defendant prejudicial effects are morally bad whereas, perhaps because sentences are too long and incarceration too brutal, pro-defendant prejudicial effects are actually good. On such a view, the prejudicial effects don’t outweigh the probative effects but point in the same direction: in favor of admitting sentencing information. This involves a highly contestable moral assessment, and we assume that this is not what defenders of PEASE have in mind, though some things Moss says do point in this direction. For example, Moss (2021, p. 272) observes that “jury nullification in response to sentencing information may provide a valuable democratic safeguard against overcharging by prosecutors or excessively harsh statutes adopted by legislatures.” This suggests that nullification is an upshot of a prejudicial effect, and to the extent it is a “valuable democratic safeguard,” that this effect is at least pro tanto good. But Moss does not develop this argument about the first-order value of nullification. [^]
- Despite Moss’s use of the terminology of “nullification,” this better fits with what Moss argues. This is also the strategy taken by Epps and Ortman (2023, pp. 825–26). [^]
- Moss 2021, p. 263. [^]
- Thayer 1892. [^]
- Amar 1998. [^]
ACKNOWLEDGEMENTS
We would like to thank Jeffrey Bellin, Chris Bottomley, Lara Buchak, Simon Chin, Brian Cross, Bob Goodin, John Hawthorne, Sarah Moss, Philip Pettit, Mark Satta, Arijeet Sensharma, Ripley Stroud, Daniel Wodak, Phil Yaure, and three anonymous reviewers for Political Philosophy for helpful conversations or written feedback. An early version was presented at Princeton University and was greatly improved by feedback from the graduate students.
COMPETING INTERESTS
The authors declare that they have no competing interests.
REFERENCES
Alschuler, A., & A. Deiss. 1994. A brief history of the criminal jury in the United States. University of Chicago Law Review, 61(3): 867–928. https://chicagounbound.uchicago.edu/uclrev/vol61/iss3/2
Amar, Akhil Reed. 1998. The Bill of Rights: Creation and Reconstruction. New Haven, CT: Yale University Press.
Anderson, Charity, and John Hawthorne. 2018. Pragmatic encroachment and closure. Ch. 7 in Pragmatic Encroachment in Epistemology, ed. Brian Kim and Matthew McGrath. Abingdon: Routledge. Accessed February 15, 2024. https://philarchive.org/rec/ANDPEA-5
Anderson, Charity, and John Hawthorne. 2019. Knowledge, practical adequacy and stakes. Ch. 10 in Oxford Studies in Epistemology: Volume 6, ed. Tamar Szabó Gendler and John Hawthorne. Oxford: Oxford University Press.
Basu, Rima. 2020. The specter of normative conflict: does fairness require inaccuracy? Pp. 191–210 in An Introduction to Implicit Bias: Knowledge, Justice, and the Social Mind, ed. Erin Beeghly and Alex Madva. Abingdon: Routledge.
Bellin, Jeffrey. 2010. Is punishment relevant after all? A prescription for informing juries of the consequences of conviction. Boston University Law Review, 90: 2223–2265. https://www.bu.edu/law/journals-archive/bulr/documents/bellin.pdf
Biebel, Nathan. 2023. Pragmatic encroachment and justified group belief. Synthese, 202 (2): 50. http://doi.org/10.1007/s11229-023-04271-7
Bronsteen, John; Christopher Buccafusco; and Jonathan Masur. 2009. Happiness and punishment. University of Chicago Law Review 76 (3): 1037–1081. https://chicagounbound.uchicago.edu/uclrev/vol76/iss3/2
Brown, J. 2008. Subject-sensitive invariantism and the knowledge norm for practical reasoning. Noûs, 42 (2): 167–189. http://doi.org/10.1111/j.1468-0068.2008.00677.x
Crewe, Bianca, and Jonathan Jenkins Ichikawa. 2021. Rape culture and epistemology. Pp. 253–82 in Applied Epistemology, ed. Jennifer Lackey. Oxford: Oxford University Press.
Enoch, David; Talia Fisher; and Levi Spectre. 2021. Does legal epistemology rest on a mistake? On fetishism, two-tier system design, and conscientious fact-finding. Philosophical Issues, 31 (1): 85–103. http://doi.org/10.1111/phis.12203
Epps, D., and W. Ortman. 2023. The informed jury. Vanderbilt Law Review, 75(3), 823–90.
Fantl, Jeremy, and Matthew McGrath. 2002. Evidence, pragmatics, and justification. Philosophical Review, 111 (1): 67–94.
Fantl, Jeremy, and Matthew McGrath. 2007. On pragmatic encroachment in epistemology. Philosophy and Phenomenological Research, 75 (3): 558–589.
Federal Farmer. 1788. Letter XV, Jan. 18, 1788. Reprinted in vol. 20, pp. 1043 ff., The Documentary History of the Ratification of the Constitution, ed. John P. Kaminski et al. Madison: Wisconsin Historical Society Press, 2004. http://archive.org/details/documentaryhisto0000jens
Gardiner, Georgi, 2019. The reasonable and the relevant: legal standards of proof. Philosophy and Public Affairs, 47 (3): 288–318. http://doi.org/10.1111/papa.12149
Gordon-Smith, Eleanor. Forthcoming. The ethics of disbelief. Australasian Journal of Philosophy.
Hawthorne, J., and J. Stanley. 2008. Knowledge and action. Journal of Philosophy, 105 (10): 571–590.
Hedden, B., and M. Colyvan. 2019. Legal probabilism: a qualified defence. Journal of Political Philosophy, 27 (4): 448–468. http://doi.org/10.1111/jopp.12180
Ichikawa, J.J. 2012. Knowledge norms and acting well. Thought: A Journal of Philosophy, 1 (1), 49–55. http://doi.org/10.1002/tht3.7
Jackson, Elizabeth, and James Fritz. 2021. Belief, credence, and moral encroachment. Synthese, 199 (1–2): 1387–1408. http://doi.org/10.1007/s11229-020-02799-6
Kvanvig, J.L. 2011. Against pragmatic encroachment. Logos and Episteme, 2 (1): 77–85. http://doi.org/10.5840/logos-episteme20112150
Lackey, Jennifer. 2021. Norms of criminal conviction. Philosophical Issues, 31 (1): 188–209. http://doi.org/10.1111/phis.12199
Laudan, Larry. 2011. The rules of trial, political morality, and the costs of error: or, is proof beyond a reasonable doubt doing more harm than good? In Oxford Studies in Philosophy of Law: Volume 1, ed. Leslie Green and Brian Leiter. Oxford: Oxford University Press. http://doi.org/10.1093/acprof:oso/9780199606443.003.0006
Levin, Sam. 2016. Dozens of letters urge leniency for Brock Turner in Stanford sexual assault case. The Guardian, June 8, 2016, sec. US news. https://www.theguardian.com/us-news/2016/jun/07/stanford-sexual-assault-letters-brock-turner-judge
Littlejohn, Clayton. 2020. Truth, knowledge, and the standard of proof in criminal law. Synthese, 197 (12): 5253–5286. http://doi.org/10.1007/s11229-017-1608-4
McCauliff, C.M.A. 1982. Burdens of proof: degrees of belief, quanta of evidence, or constitutional guarantees. Vanderbilt Law Review, 35: 1293–1335. https://scholarship.law.vanderbilt.edu/vlr/vol35/iss6/1/
Moss, Sarah. 2018. Moral encroachment. Proceedings of the Aristotelian Society, 118 (2): 177–205. http://doi.org/10.1093/arisoc/aoy007
Moss, Sarah. 2021. Pragmatic encroachment and legal proof. Philosophical Issues, 31 (1): 258–279. http://doi.org/10.1111/phis.12207
Moss, Sarah. 2022. Knowledge and legal proof. In Oxford Studies in Epistemology Volume 7, ed. Tamar Szabó Gendler, John Hawthorne, and Julianne Chung. Oxford: Oxford University Press. http://doi.org/10.1093/oso/9780192868978.003.0006
Pettit, Philip. 2023. The State. Princeton, NJ: Princeton University Press.
Ross, Jacob, and Mark Schroeder. 2014. Belief, credence, and pragmatic encroachment. Philosophy and Phenomenological Research, 88 (2): 259–288. http://doi.org/10.1111/j.1933-1592.2011.00552.x
Schroeder, Mark. 2012. Stakes, withholding, and pragmatic encroachment on knowledge. Philosophical Studies, 160 (2): 265–85. http://doi.org/10.1007/s11098-011-9718-1
Schwartz, D., and E. Sober. 2017. The conjunction problem and the logic of jury findings. William & Mary Law Review, 59 (2): 619–692. https://scholarship.law.wm.edu/wmlr/vol59/iss2/5/
Smartt, Timothy. 2022. Reconsidering the rule of consideration: probabilistic knowledge and legal proof. Episteme, 19 (2): 303–18. https://researchonline.nd.edu.au/ies_article/5
Smith, Martin. 2018. When does evidence suffice for conviction? Mind, 127 (508): 1193–1218. http://doi.org/10.1093/mind/fzx026
Sood, A.M. 2021. What’s so special about general verdicts? Questioning the preferred verdict format in American criminal jury trials. Theoretical Inquiries in Law, 22 (2): 55–84. http://doi.org/10.1515/til-2021-0017
Sood, A.M. 2023. Reaching a verdict: empirical evidence of the crumbling conventional wisdom on criminal verdict format. New York University Law Review, 98 (4): 1265–1357.
Stanley, Jason. 2005. Knowledge and Practical Interests. Oxford: Oxford University Press.
Thayer, James Bradley. 1892. The jury and its development: III. Harvard Law Review, 5 (8): 357–388. http://doi.org/10.2307/1321147
Thomson, Judith Jarvis. 1986. Liability and individualized evidence. Law and Contemporary Problems, 49 (3): 199–219. http://doi.org/10.2307/1191633
Tocqueville, Alexis de. 1840. Democracy in America, vol. 2, trans. George Lawrence, ed. J.P. Mayer and Max Lerner. New York: Harper & Row, 1966.
von Klemperer, Caroline. 2023. Moral encroachment, symmetry, and believing against the evidence. Philosophical Studies, 180: 2175–2190. http://doi.org/10.1007/s11098-023-01969-9
Worsnip, Alex. 2020. Can pragmatists be moderate? Philosophy and Phenomenological Research, 102 (3): 531–558. http://doi.org/10.1111/phpr.12673
Zweber, Adam. 2016. Fallibilism, closure, and pragmatic encroachment. Philosophical Studies, 173: 2745–2757. http://doi.org/10.1007/s11098-016-0631-5