MULTIPLE JURIES
Epistemic Democracy and Institutional Design Conference 25/05/2017-27/05/2017, Montreal
Abstract. In this paper, I present some arguments in favor of multiple (criminal) juries as an alternative to the current single jury system. I do this by arguing that two of the main epistemic justifications for jury trials – i.e. aggregation and deliberation – have good (and independent) reasons for favoring multiple juries over single juries. By multiple juries, I mean different groups of jurors that are selected in the same way, have the same size and simultaneously, though separately, decide the outcome of a trial. An institutional scheme with multiple juries is any scheme that counts two or more juries. There are three types of multiple juries systems explored in this paper: (1) provisionally multiple sub-juries that are part of a bigger jury; (2) multiple independent juries that follow the same decision-making procedure and whose verdicts are counted following a majority or super-majority rule (3) multiple and procedurally mixed (i.e. both deliberative and non- deliberative) juries whose verdicts are counted following a majority or super- majority rule. These three multiple juries systems are examined separately and rely on different kinds of considerations.
One standard justification for juries is epistemic, and holds that juries are truth- finders.
1There are two dominant strategies for substantiating this epistemic justification.
The first one is quantitative, and roughly holds that more heads are better the one, insofar as the sheer addition of different individual judgments tends to generate better collective judgments overall. Call this the aggregative epistemic view of jury justification. The second strategy is qualitative, and contends that juries are apt truth- finding institutions because of the process whereby individual jurors talk about, deliberate on and critically assess their private judgments about the evidence presented at trial. Call this the deliberative epistemic view of jury deliberation.
This paper argues that, on both the aggregative and the deliberative view, there are good reasons for favoring multiple juries over single juries. By multiple juries, I mean different groups of jurors that are selected in the same way, have the same size and simultaneously, though separately, decide the outcome of a trial. An institutional scheme with multiple juries is any scheme that counts two or more juries. For reasons that are going to become clear later on, it is sometimes preferable that the multiple number of juries be an even number.
1
This paper will focus on criminal juries and, in particular, on criminal juries in the US (unless
The aggregative case for multiple juries is a hypothetical one. By hypothetical, I mean in that it does not target the actual status quo of jury trials, but mostly the way in which the jury system would be organized under hypothetical conditions.
2The aggregative argument in favor of multiple juries will be premised on the contention that the relationship between jury size and its epistemic effectiveness is not linear, in the sense that increasing the size of the jury will not indefinitely increase its effectiveness in coming up with accurate decisions. This is because, passed a certain threshold, the epistemic benefits of increasing group size will be neutralized or, worse, potentially reversed by the jurors’ rational tendency to epistemically free-ride on the group’s judgment.
The deliberative case for multiple juries is, on the other hand, a real one. By real, I mean that it targets the current practice of jury trials and relies on the actual evidence we have about the reliability of jury deliberation in generating accurate verdicts. This evidence is sufficiently mixed so that one can plausibly assert that, when it comes to increasing the probability of accurate verdicts, jury deliberation is roughly epistemically equivalent to the jurors’ voting on the verdict without any prior deliberation.
There are different normative and institutional implications one can draw from the epistemic equivalence between deliberation and aggregation. One of the implications is that we are better off with procedurally distinct multiple juries, where some juries decide the verdict following a purely aggregative procedure and others decide it according to the prevailing deliberative procedure. Though such an institutional change is not necessarily ideal, it represents an improvement over the current status quo. Or so, at least, I will argue.
3Before moving on to my argument, a few remarks are in order. The first remark is that multiple juries are already used only in criminal cases multiple defendant as a way of dealing with inadmissible evidence. In such cases, there can be two or more juries, with each jury deciding the verdict of one of the multiple defendants and hearing only the evidence that is admissible for their defendant (Canty 2010). This is not the sense in which I will defend the multiple jury proposal. My argument is in favor of multiple
2
There is however an argument why the aggregative scheme might embrace a certain scheme of multiple juries even in relation to the current status quo. This is a point I will come to in Section I.
3
If Rawls (1971) is right in claiming that imperfect procedural justice is a structural feature of criminal
justice, it follows that all practical proposals in this domain will be imperfect in some respect. I claim
no different for the proposals that I’m outlining in this paper. More generally, if criminal justice is
procedurally imperfect, it makes little sense to criticize its rules and procedures or the proposals for
reforming them for not being perfect or for not spelling out a procedural ideal.
juries judging the case of a single defendant, and should not be confused with arguments in favor of multiple juries for multiple defendants. But, bearing this distinction in mind, the fact that multiple juries are already used today is important insofar as it shows that a multiple jury system would not face considerable practical (or constitutional) obstacles.
My second remark is that multiple groups are often employed for epistemic purposes outside of the criminal justice system. For example, focus groups are often resorted to when we want to corroborate or extend our findings about questions concerning policy perceptions or the effects of commercial goods on lay citizens. If we think that the logic of multiple groups is important for increasing the accuracy and exhaustivity in other domains, why not do the same in the criminal justice context?
The third remark is that, in examining the aggregative and deliberative arguments for a multiple juries system, I do not claim that these arguments work cumulatively or that they can always be combined in a non-contradictory way. Rather, my goal is to examine these arguments separately in order to see whether the two dominant justifications that capture the epistemic benefits of jury trials should favor a multiple juries system over a single jury one. Whether these reasons can be tracked by the same multiple jury system is a question that would have to considered on another occasion.
Fourth, and finally, my different practical proposals in favor of multiple juries are meant to work in a context where jury trials are decided based on a majority or super- majority rule, and not on the basis of a unanimity rule. This is because the epistemic benefits of deliberation and aggregation risk being reduced or, worse, reversed in systems that rely on the unanimity rule. We know that, in the jury context, the unanimity rule risks increasing the chances of both false positive and false negative decisions (Feddersen and Pesendorfer 1998). More generally, as Melissa Schwartzberg shows, the unanimity rule is not only epistemically unwarranted, but also morally problematic. This is a limit of unanimity that was noted early on in the 18
thcentury, when philosophers interested in juries realized that the unanimity rule did not only not improve the likelihood of correct decisions, but increased the chances of individual jurors being coerced in supporting a majority judgment that did not believe in.
The paper is structured as follows. In Section I, I spell out the reasons why a
supporter of the aggregative view ought to push for a multiple jury scheme over a single
jury one. I do this by relying on the Condorcet Jury Theorem (CJT) and argue that a
with certain jury sizes. In Section II, I present the evidence on the epistemic effectiveness of deliberation. Based on those findings, I suggest some possible reforms, one of which is a procedurally mixed system of multiple juries. Finally, in Section III, I briefly discuss whether and, if so, how multiple juries matter for democracy.
I. Aggregation, Epistemic Free-Riders and Multiple Juries
The Condorcet Jury Theorem (CJT) provides a concise formulation of the aggregative case for the jury’s epistemic justification. The CJT states that, if the probability of each individual juror of reaching an accurate decision on a binary question is better than random, then the probability of the entire jury reaching an accurate decision following a majority decision rule is greater than the one of the individual juror and positively proportional to the size of the jury. This implies that the probability that the jury decision is accurate increases with its size.
Following the CJT, the probability of an accurate decision increases under any of the following three conditions: (1) there is an absolute increase in the size of the group majority, i.e. there is an increase in the difference between the size of the group and the size of the group minority or the added group minorities, (2) there is an increase in the probability of the average juror making an accurate decision, or (3) with group size being kept constant, there is an increase in the ordinal size of the group majority, i.e.
there is an increase in the relative size of the difference between the size of the group and the size of the group minority or the added group minorities. These three conditions are non-exclusive, in the sense that they can be satisfied simultaneously in a way that increases the probability of a correct group decision and the majority rule.
When referred to the actual practice of criminal juries, the CJT has been criticized for relying on two unrealistic assumptions: one concerning the individual juror’s epistemic competence and the other, the independence of the jurors’ judgments.
None of these criticisms is necessarily fatal to the CJT as applied to criminal juries. The first criticism can be tackled by pointing out that the CJT can be satisfied even in cases where not all individual jurors have an epistemic competence that is better than random.
The CJT can be realized even under conditions where only the average juror (as distinct from each individual juror) has an epistemic competence that is above the 50%
threshold and that the jurors’ competence is normally distributed around the average
(Grofman & Feld 1983; Owen et al. 1989; List 2001; Estlund 2008). Not every juror
has to be smart as long as, on average, jurors are smart enough and there are no sizeable minorities that are too dull.
4One way to deal with the second critique is to show that it rests on a conceptual confusion. As Estlund (2008) notes, the kind of independence that matters for satisfying the CJT is not causal, but statistical independence. Two individual judgments are statistically independent if the probability of either one being correct (or not) is not affected by the probability of the other being correct (or not). This means that the CJT can accommodate a process of deliberation whereby individual jurors can causally influence each other’s judgments. Thus, the CJT can be satisfied in a deliberative settings with large groups (Ladha 1992). Given their current size, this latter point is not directly relevant to juries, but it could be relevant to a jury system that, following the aggregative view, would require substantially bigger juries.
This brief digression into two of the main critiques of the CJT was meant to show that the aggregative view paints a plausible picture of the jury’s truth-finding properties. It should however be obvious that, with jury sizes as small as 12 or under 12, the aggregative view is far from being optimally satisfied. The current jury system does not take full advantage of the epistemic potential of aggregation. In the rest of this section, I want to examine how, if we adopt the aggregative view, a jury system would look like under conditions where a criminal justice system could count on sufficient resources. I will argue that, under such conditions, supporters of the aggregative view have reasons to prefer a multiple jury system to a single jury one. In addition, I will suggest that the aggregative view might be compatible with a modified scheme of multiple juries even if we keep the small juries that we have today.
The aggregative view, as formulated by the CJT, seems to require that a jury be as big as possible. If one believes that aggregating individual judgments whose average competence is better than random increases the probability of generating accurate decisions, there is no apparent principled reason why the number of jurors should be kept small. The aggregative view may thus guide us toward a jury system that would resemble the ancient Athenian one, where the size of juries ranged from approximately 200 jurors for minor cases to roughly 2,500 for serious offences (Hansen 1998;
Schwartzberg 2014). Note that such a substantial increase in jury size might also deal
4
Moreover, as Landemore (2013) underlines, epistemic pessimism about the epistemic abilities of lay
citizens is far from being obvious contention. This is because the assumption about individual
epistemic competence ‘does not require a leap of faith, but a simple belief in common sense, which has
indirectly with some of the judgment dependence problems that can be prompted by a deliberative setting. This would happen because deliberation is considerably more difficult (if not entirely impractical) in big groups as compared to smaller ones: 200 jurors cannot influence each other’s judgments in the same way that 12 jurors can.
The reasons usually given for not having juries as big as the Athenian one point to their costliness. The argument here is that having juries that count hundreds or thousands of members would prove too expensive, both in terms of their actual organization and for the opportunity costs that they would impose on individual jurors.
Suppose, however, that there were sufficient resources both for organizing Athenian- sized juries and for compensating individuals for the opportunities they lose in performing their jury duty. Would supporters of the aggregative view still have reasons to oppose having Athenian-sized juries? I will suggest that they would.
Supporters of the aggregative view have good reasons not to prefer one jury that is as big as possible. Instead, they should prefer a system with multiple juries. The argument for this is that juries that are too big create rational incentives for individual jurors to free-ride on the judgment of the jury majority. The bigger the jury, the more reason a juror has to rely on whatever the majority decision turns out to be. Call this the problem of epistemic free-riding. The problem here is that the mere size of the jury might have a negative effect on the accuracy of the jury’s decision as a whole.
Epistemic free-riding will happen because, with the size of the jury increasing, each individual juror has fewer reasons either to give due epistemic attention to his decision or to stick to his own individual judgment if confronted with the majority judgment. A juror who doesn’t rely on his judgment behaves like an epistemic free- rider. When this happens, a juror’s judgment depends on the majority judgment and, when every or even when most jurors’ judgments depend on the majority judgment – i.e. when the individual jurors’ judgments are not independent – the CJT is no longer satisfied. The implication, then, is that, because of epistemic free-riding, jury size will have to be kept relatively small, thus preventing the aggregative view from fulfilling its full epistemic promise.
55
The problem of epistemic free-riding is different from the second critique sketched out at the
beginning of this section, which targets the violation of the independence condition on deliberative
grounds. The problem here is not deliberation per se, but rather that, passed a certain size, jurors will
lose their incentives to invest their epistemic competence in the decision-making process, thereby
altering the quality of the final verdict in a negative way.
One explanation as to why epistemic free-riding might occur is that the act of judging demands effort and that, when the jury is big enough, individual jurors will find it rational to invest less of their competence in forming an accurate judgment themselves and will prefer to defer this task to others. When such epistemic disinvestment occurs across all or most jurors, the jury’s decision might, under certain conditions, be less accurate for bigger juries than for smaller ones.
This explanation makes sense if we think about jury decision-making in terms of rational choice. The hypothesis here is that every juror shares the goal of reaching an accurate verdict, but rationally prefers to reach this goal while reducing the epistemic costs that come with it. The epistemic costs in this case refer to the degree of attention that jurors pay to the evidence presented at trial, as well as to the degree of care with which they examine the information provided by that evidence. Given their preference to reduce epistemic costs, jurors are going to tend to satisfy this preference more in larger groups, where they know that they can count on the aggregate competence of others. Larger juries can thus create an incentive for epistemic freeriding, whereby jurors will be inclined to align their judgment with the majority’s judgment. If everyone does this, the decisions taken by larger juries will be less accurate than those taken by smaller ones.
Mukhopadaya (1998) shows that, for certain realistic specifications of the precision of the evidence and jury size, small juries of 6 members can make more accurate decisions at the aggregate level than juries of 12 members. Similarly, McCannon and Walker (2016) argue that, because, within groups epistemic competence constitutes a public good, ‘adding more members to the group discourages investments in it’ (2).
Battacharya, Duffy and Kim (2017) have recently corroborated the free-riding
hypothesis in an experimental setting in cases where evidence is clear and precise and
group size is increased. Under such conditions, they have found that individual
epistemic investment drops significantly when group size increases from 3 to 7
members. More impressionistically, Williams (2005) notes that ‘the jury task seems
custom-made to elicit free-riding responses,’ since it is rational for jurors to suppose
that, even if they fail to pay proper attention to the evidence presented during the trial
proceedings, such a failure ‘will be of no great consequence, because the others’ efforts
will compensate for their own lapses’ (298).
A different explanation for epistemic free-riding is articulated by Pettit and List (2004). They argue that the aggregative view, as formulated by the CJT might be practically self-defeating in the following way: if every juror came to be convinced of the truth of the CJT, there would be realistic circumstances where such a generalized conviction might block or, worse, reverse the epistemic benefits of aggregation.
These circumstances would obtain when every juror is determined to make a correct decision, believes that others are expressing their judgment truthfully and the jury’s decision-making process is organized so that every juror can observe the majority’s judgment. When this happens and insofar as every juror takes the CJT to be true, it seems epistemically justified for jurors to think that they improve the accuracy of their individual decision by voting on the verdict like the majority does. The outcome in such cases is that the contribution each individual judgment makes to the final one becomes virtually null. Surrendering one’s judgment to the majority seems at the same time epistemically justified at the individual level, but turns out to be potentially disastrous at the collective level. This is because, in submitting to the majority judgment, every individual judgment will fail to add anything substantive to it and will put individual jurors in the position of epistemic free-riders. Unlike the previous explanation, this scenario shows how epistemic freeriding might rely on right reasons at the individual level.
List and Pettit (2004) argue that this form of epistemic freeriding has more chances of occurring when decision-making is organized transparently and sequentially.
This is because, in this latter scenario, ‘any majority that accidentally emerges among the first two or three jurors may grow further and further, suggesting, mistakenly, an increasing degree of support for the hypothesis that things are as this growing majority says they are’ (13). Because they want to keep jury deliberation in place, List and Pettit’s solution to avoid such free-riding cascades is for each individual juror to avoid sharing her exact verdict with others, while at the same time share information that she thinks will have been ignored by the other jurors and that does not have a conclusive implication for a specific verdict.
This separation between one’s verdict and one’s non-conclusive views about a
case is epistemically unrealistic (given the interconnected nature of different pieces of
evidence presented at trial), practically undesirable (insofar as it risks pushing each
juror into a problematic game of second-guessing the other jurors’ judgment) and
psychologically implausible (to the extent that we might have reasonable doubts about
individual being able to systematically avoid revealing their position on the case even if they are motivated to do so).
For the sake of the argument, I grant that renouncing deliberation might be ‘an unacceptably high price for the avoidance of free-riding’ (152). But I disagree with List and Pettit’s institutional proposal. Instead, my suggestion is that free-riding in List and Pettit’s sense can be avoided or at least alleviated by provisionally dividing the jury into smaller (and therefore multiple) sub-juries that can deliberate separately on the merits and verdict of the case before engaging in deliberation with all the others. This is to say that a scheme of provisional multiple sub-juries might help with fre-riding in List and Pettit’s sense. The reason why this might happen is that, if every sub-jury is kept small enough, the reasons for deferring to the majority judgment can be significantly weakened, at least for the time that is needed for each individual juror to make an independent contribution to deciding the case at hand. A scheme with multiple provisional sub-juries would be one where jurors first deliberate in smaller groups and then unite at the jury level to deliberate again based on the decisions that they made at the sub-jury level.
There are several ways in which the provisional division of the jury into sub- juries can be organized. As a general rule, it would be desirable that the sub-juries be small enough and of equal size. This is both because we want the decision-making processes to be comparable across the different multiple sub-juries and because a sub- jury that is too big in relation to another one – say, a sub-jury of 6 as compared to a one of 3 – might face the free-riding problems that we want to avoid in the first place.
Moreover, because sub-juries are only provisional, groups of unequal size might have less deliberative weight when groups reunite to decide the final verdict.
As far as the composition of the sub-juries is concerned, one could imagine a system where jurors first vote secretly on the verdict and, insofar as individual verdicts turn out to be heterogeneous, the jurors are then allocated to different sub-juries following a procedure that maximizes the number of heterogeneous sub-juries. This means that the size of the sub-juries might vary from one case to another within the bounds that plausibly avoid free-riding problems and satisfy the equality requirement.
The reason for preferring heterogeneous to homogenous sub-juries is that we know that
genuine epistemic disagreement can lead to more accurate decisions than agreement or
contrived disagreement (Dunn, forthcoming).
However, when the voting patterns turn out to be too homogenous but not homogenous enough to decide a verdict, one can resort to techniques of contrived disagreement that might improve the prospects of accurate decisions at the sub-jury level. These techniques include mechanisms like devil’s advocate or red-teaming and their fundamental rationale is to inject ‘some cognitive diversity where it is lacking or stifled’ (Landemore 2013: 122).
Alternatively, jurors can be randomly allocated to sub-juries. This procedure might not maximize the number of heterogeneous sub-juries, but it would nonetheless be an ex ante fair way of dividing the jurors into different groups.
6One advantage of lotteries is that they are better placed in designing sub-juries that are cognitively and epistemically more diverse than the juries that would result from a patterned, heterogeneity-maximizing allocation procedure. Another advantage is that, because random allocation does not require jurors to first vote on the verdict, each juror can take more time to think about the case than she could in a situation where we follow a first- vote-then-allocate procedure.
The benefit of a provisional division of the jury into sub-juries is that, given the small size of each sub-jury and the fact that jurors ignore other sub-juries’ verdicts, each juror has strong initial reasons to make an independent contribution to deciding the verdict. Moreover, if there is agreement on the verdict at the sub-jury level, each sub- jury might be less vulnerable to epistemic pressure from the other sub-juries than any individual juror is vulnerable in relation to the other jurors within a single jury. But it bears repeating that, even in cases where there is no agreement at the sub-jury level, individual jurors would still have had more opportunity to make an independent contribution to the overall verdict.
I will now briefly turn to the kind of epistemic freeriding identified by the rational choice model, where, unlike in List and Pettit’s scenario, jurors do not defer to the majority for the right reasons, but rather because they selfishly prefer to minimize epistemic costs. On this reading, the tendency to engage in epistemic freeriding increases with the size of the group, and raises a worry that very big groups would actually turn out to be less accurate than smaller groups. If that is the case, then it is not epistemically desirable to have Athenian-sized juries.
6
On ex ante fairness in the criminal justice context, see Chiao (2012).
However, reducing the size of a jury to avoid epistemic freeriding of the selfish kind does not necessarily imply that we should also reduce the number of jurors who decide a case under the hypothetical indefinite budget scenario. There is an institutional se-up that might allow us to keep the epistemic benefits of aggregation, while avoiding the risk of epistemic freeriding. This might be done by introducing a system of multiple juries that are big enough to produce some of the benefits of aggregation and small enough to avoid the negative effects that come with freeriding.
Determining the exact size of such juries is an empirical issue that should be explored experimentally before being set-up institutionally. Given the financial and time resources that such a research program would demand, I am aware that a multiple jury system of this second kind will not be instituted any time soon. Note, however, that my case for preferring a multiple jury system is meant to be highly hypothetical. The point of this form of the multiple jury argument is not that it should be introduced here and now, but that this is the kind of jury system that we should prefer under circumstances where we want to track the epistemic properties of aggregation.
Note also that the multiple juries designed to deal with the selfish version of epistemic freeriding are different from the multiple provisional sub-juries discussed in relation to List and Pettit’s version of freeriding. The difference is that, unlike provisional juries, multiple juries in this case are permanently separate and have to form an independent verdict on the case at hand. Each jury verdict would be decisive in that it could not be revised when added to the other verdicts. Thus, each jury’s verdict would count as a single vote in favor or against the final overall verdict.
There are two epistemically attractive schemes whereby multiple juries could be instituted. Following a first scheme, there would be an even number of multiple juries.
In this case, if we settle for a scheme with two juries, both juries would have to agree on the verdict for it to be conclusive. In cases where the juries disagree, the judge would have to declare a mistrial for the same kind of reasons that mistrials are currently declared when individual jurors fail to agree on the defendant’s guilt or innocence.
In cases where we have 4 or any other superior (and even) number of juries, the
verdict could be decided following either a majority or a super-majority rule. The
choice between these two formulas depends on whether we decide to keep our bias in
favor of false negatives or not. If we do, then it is preferable to opt for the super-
majority rule (Schwartzberg 2013). Absent a super-majority of juries in favor of
Following a second scheme, we could go with an odd number of multiple juries.
For reasons that are the same as those indicated in the scheme with an even number of juries, verdicts would be decided following either a majority or a super-majority rule.
For practical reasons that have to do with minimizing the probability of deadlock and delivering justice in a timely way, we might prefer to go with this latter arrangement (thus, we would prefer 3 to 4 juries, 5 to 6, and so on).
As I see it, there would be two main epistemic advantages that a multiple jury system would have over a hypothetical big single jury system. First, a multiple jury system would prevent the potentially pervasive free-riding problems that would occur in a big single jury system. This would happen mechanically, i.e. simply in virtue of keeping the jury size small enough to neutralize the incentives for freeriding. More speculatively, the incentive for any particular jury to free-ride on the other juries might be out-weighed by the jury’s incentive to have its verdict counted as one of the verdicts forming the putatively correct (super-)majority verdict. A system with multiple juries would thus work as a platform that fosters a sense of epistemic competition among different juries. There is no analogue for such an incentive under a single jury system.
Second, a verdict on which there is agreement among a majority or super- majority of juries is a verdict that warrants more epistemic confidence. This is because, under the hypothetical scenario where the assumptions of the CJT are granted, the chances that any particular jury will make an accurate decision are bigger than the chances of individual jurors within a single jury making an accurate decision. Because every jury’s competence is better than random and duly insulated from the risk of epistemic free-riding, the decisions produced by a multiple jury system are comparatively better than the decisions produced either by the current small single jury system or by a hypothetical alternative big single jury system.
My proposal for multiple independent juries might come across as particularly unrealistic. This is because we have neither the resources nor the willingness to introduce such a system. I am the first to recognize the practical limits of my proposal.
However, the point of this section was not to push for immediate reform of the current jury system. Rather, the point was to provide an outline for how a jury system should look like if we accept the aggregative view of the jury’s epistemic justification and we dispose of an indefinite budget that would allow us to realize it. In granting the possibility of epistemic free-riding, I wanted to keep in line with Rousseau’s dictum of
‘taking men as they are and laws as they can be.’ In the following section, I will present
a case for multiple juries that is less hypothetical and that, to my mind, targets a real moral and epistemic problem that our current jury system confronts.
II. Deliberation, Epistemic Equivalence and Multiple Juries
The focus in this section will be on the epistemic record of jury deliberation and on the institutional implications of the empirical evidence we have about it. Jury deliberation is currently the mandatory and unique procedure that juries rely on to form their verdicts. ‘Mandatory’ means that jurors are strictly required to deliberate and that failure to do so could, if, provide reasons for a mistrial.
The argument that I am going to present in this section is twofold. First, I am going to argue that, given what we know about the epistemic effectiveness of jury deliberation, we lack good reasons for keeping deliberation as the unique method for deciding the outcome of trials by jury. Second, I will provide some reasons as to why a multiple jury system whereby trials are decided following a mixed procedure that combines deliberation with pure aggregation is preferable to the status quo. I will start with the evidence and then move to the institutional implications, one of which is a mixed multiple jury system.
Jury Deliberation: The Evidence
The findings about the epistemic effectiveness of jury deliberation are mixed.
7This means that, on the whole, deliberation doesn’t have a record that guarantees it may work as the single most reliable procedure available for structuring the jury’s decision- making process. It is important to underline from the start that the epistemic assessment of deliberation presented by the empirical evidence is a narrow one. It offers a comparison between deliberative and non-deliberative verdicts that result from a purely aggregative (i.e. non-deliberative voting) procedure.
Based on this comparison with non-deliberative voting, the current evidence indicates that deliberation is, on the whole, both epistemically good and bad, and persistently so.
8I will argue that the mixed character of the evidence that we have about deliberation warrants an attitude of uncertainty about its epistemic effectiveness.
Capitalizing on this uncertainty, I will also argue that deliberation should not have the
7
In what follows, I adopt a reliability-based conception of epistemic justification. Though this is not
the only way of thinking about knowledge and epistemic justification, it has the merit of fitting the
logic of the criminal trial, which is both a process and an evidence-based one.
special procedural position it enjoys today. In particular, I will contend that we are better off with an institutional set-up that either radically reshapes the structure of deliberation or, more realistically, that we should favor an arrangement where deliberation is on a par with non-deliberative voting.
I identify five sources of uncertainty about the effectiveness of jury deliberation, although I admit that there might be more and that some of the ones I examine here might be overlapping.
9The first one concerns the impact that deliberation has on jurors’
reasoning skills. McCoy, Nunez & Dammeyer (1990) have found that, after a period of deliberation on a murder case, jurors proved more nuanced in their interpretation of the facts and were better at evaluating the conflicting evidence they were confronted with.
This is however hard to reconcile with the fact that group deliberation has also been shown to stifle the exercise of jurors’ individual reasoning. This latter phenomenon is typically known as the reputational cascade effect, whereby individuals rely on or repeat what other seemingly more competent jurors think (Sunstein 2006: 92). These findings give us non-trivial reasons to think that the perfecting effect of deliberation on jurors’ inferential competence may, at times, be more apparent than real.
The second source of uncertainty questions whether deliberation can strengthen the jurors’ capacity to identify accurate evidence and detect deceit. Salerno & Mccauley (2009) found that group deliberation improved jurors’ ability to detect flawed scientific expertise by increasing the truth-discriminating virtues of cross-examination during trial.
In particular, they found that deliberation had a more substantial impact on individuals who initially lacked the motivation to process experts’ testimony. The explanation for this post-deliberative upgrade might be that “jurors who processed the cross- examination deeply might force other jurors who either distrusted, did not understand, or did not pay attention to cross-examination to incorporate it into their judgments”
(Salerno, Mccauley 2009: 5). But the same authors admit that it remains unclear whether deliberation is bolstering one’s motivation for detecting inaccurate expertise or whether jurors with poor cognitive motivation are merely giving in to the pressures of the more cognitively active ones. The logic of these latter findings is also at odds with
9
The disagreement over the reliability of jury deliberation is genuine, in that it is not correlated to any
specific jury deliberation method. In particular, there is no correlation between verdict-based
deliberation and negative findings or evidence-based deliberation and positive findings about its
reliability. Neither is there a relation between the positive or negative findings about the epistemic
record of deliberation and absence or presence of the unanimity rule. This means that group
deliberation, as currently practiced by juries, performs well or badly independently of the (bad) effects
unanimity might have on it. For an epistemic critique of the unanimity rule, see Schwartzberg (2013).
Kraus and Lee’s earlier study (2003), showing that, in the context of a capital sentencing trial, deliberation had (at best) a very weak impact on identifying accurate expertise.
The third level of uncertainty pertains to the relation between deliberation and its role in dismissing inadmissible evidence in general, not only expertise. In a simulated case of armed robbery, Kerwin & Shaffer (1994) found that deliberating juries were more effective than non-deliberating ones at following judicial instructions in discarding inadmissible evidence. London & Nunez (2000) obtained similar findings in a mock sexual assault case, which prompted them to conclude that “the impact of inadmissible evidence was lessened following jury deliberations” (937). More generally, Steblay et al.’s (2006) meta-study found some evidence that deliberation can “diminish the influence of otherwise damaging inadmissible information” (486). However, this contradicts Steblay et al.’s (1999) previous meta-study concluding that deliberation could not eliminate the influence of pre-trial publicity on the jury’s final decision. It is also difficult to reconcile with Thompson, Fong and Rosenhan’s (1981) or Carretta &
Moreland’s (1983) findings that deliberation helped discard inadmissible evidence presented by the prosecution, but was unable to do so when the evidence came from the defense.
Fourth, jury deliberation plays an uncertain role in offsetting any biases and stereotypes that may enter the jury’s decision-making process. For example, Izzett &
Leginski (1974) found that deliberation lowered the influence of a defendant’s unattractiveness on the verdict. Kaplan & Miller (1978) showed that deliberation is effective at countering situational biases, like the ones induced by an obnoxious lawyer during trial. These findings were nonetheless later nuanced by Kerr, Niedermeier &
Kepler (1999), who found that deliberation counters biases only in extreme cases, in which the probability of conviction is either very high or very low, but increases their influence in moderate cases. Similarly, in an earlier study, Kramer, Kerr & Caroll (1990) found that the impact of negative pre-trial publicity on the verdict was increased through deliberation, a phenomenon they attributed to less confident jurors being more easily persuaded to change their views at the end of group deliberations.
More recently, Haegerich, Salerno & Bottoms (2013) produced evidence that
pre-trial stereotypes held by individual jurors were minimized by jury deliberation. But
this study also showed that when stereotypes were activated during trial proceedings,
authors to conclude that “the impact of jurors’ stereotypes on case decisions could be
‘maximized’ or ‘minimized’ by the deliberation process” (81).
The picture becomes blurrier when we take into account Takada & Murata’s (2014) study showing that group deliberation can increase the impact of framing effects on the jury’s verdict or Robert MacCoun’s (1990) earlier experiment showing that a defendant’s physical attractiveness influenced the jurors only after group deliberation.
Still more confusingly, this latter study was contradicted by Patry’s (2008) more recent finding that “deliberation seemed to eliminate a predeliberation bias in favor of the attractive defendant” (731).
Fifth, and finally, it is uncertain whether deliberation improves the mnemonic quality of jury decisions. Hastie, Penrod & Pennington (1983: 78) found that deliberation substantially improves the deliberators’ recall reliability. Only 59% of individual jurors were accurate in their memory of the trial testimony, whereas the jury’s post-deliberative accuracy rate was 93%. Pritchard & Keenan (2002) confirmed this difference, although their results were considerably more modest. Jury deliberation improved the accuracy of recall by 1% for central information and by 6% for peripheral – that is, outcome-irrelevant – information. They conjectured that such minor effects happen because “highly confident individuals, who were slightly more likely to be controlling jury deliberation, were not always the most accurate” (600).
More generally, a growing body of findings shows that group collaboration sometimes diminishes recall performance (Basden et al. 1997; Weldon, Blair, Huebsch 2000; Maki et. al. 2008). For example, Hirst, Coman & Stone (2012: 164) suggested that group interaction is likely to improve the group’s collective memory, but only for groups whose members know each other and who are connected through stable relations. Furthermore, they suggest that, because of one or a few persons leading the group’s interactions (a phenomenon they call “the dominant narrator effect”), the other members’ individual memories might simply be excluded from the deliberative process.
The authors conclude that “there is no reason to believe that 12 people working together to remember the testimony from a trial will remember it more accurately than 1 person remembering on his or her own” (178).
Less speculatively, Nunez, McCrea & Culhane (2011) worry that, for various
reasons – like a failure to communicate one’s opinion properly, the disorganized pace
of group discussions or social loafing – increasing the size of groups could reduce “the
perceived responsibility any one individual feels for contributing to the discussion, and
so vital information may not enter into the discussion” (445). It is not clear whether this means that accuracy might be improved by reducing the size of juries. What is clear is that such a claim is at odds with the more widely shared belief that the main advantage of deliberation resides in the diversity of views that it brings together.
Before moving to the next section, a couple of remarks are in order. The first one is that the currently available evidence attests that jury deliberation is not bound to be reliable and that, as a result, we cannot argue ex ante that it is more epistemically better than non-deliberative voting. These studies show that the uncertainty surrounding jury deliberation is relative, not absolute: its mixed credentials are narrowly assessed in relation to a simple voting procedure, whereby individual jurors judge the case themselves and then make a non-deliberative decision on the verdict. This observation is important because it indicates that our uncertainty about jury deliberation does not cut sufficiently deep to warrant an attitude of wholesale skepticism. For all we know, deliberation remains epistemically better than a significant number of alternative decision-making procedures. It is surely better than deciding the outcome of the trial randomly – say, by rolling a die – or through some patently dubious procedure like the ordeal. Saying that jury deliberation is uncertain therefore strictly means that it is uncertain compared to the process of jurors voting on the verdict individually.
The second remark concerns the possible explanation for why deliberation as currently practiced is not, on the whole, epistemically better than the aggregation of individual verdicts. The explanation lies, I think, in the fact that there might be some mechanisms of the deliberative situation that increase the probability of an inaccurate verdict. These mechanisms may be activated in cases where deliberative interactions lead to strengthening certain biases the jurors might be already sensitive to or when it inhibits their memories of factually relevant considerations argued during the trial proceedings. My speculative contention is that it is the direct, face-to-face nature of the interactions between jurors that plays a central role in explaining both why some memories are inhibited or some biases are triggered and end up infecting the content of the jury’s decision.
One plausible way to make sense of the counter-intuitive finding that, on the
whole, deliberation performs as good as aggregation is to think about it in terms of
certain decision-making processes that might be activated during the deliberative
process which are absent in the aggregative scenario. Whatever the specific
confrontational, person-to-person dimension of jury deliberation. This dimension is absent in the case of pure aggregation.
Or it might be that, more generally, deliberation itself is incapable of filtering out epistemically illegitimate considerations, like biases or inflated epistemic claims, and that in communicating them to all the jurors it sometimes worsens the quality of their judgment. This would mean that deliberation is somewhat epistemically neutral, in the sense that it can result in outcomes whose content largely depends on the input of the deliberative process. Resorting to a musical analogy, deliberation might be like a type of instrument that can be played both well and bad, depending on who plays it and on the particular token of the instrument that is being played.
But no matter what the explanation for the epistemic equivalence between deliberation and aggregation is, I think it undermines the argument for imposing the former as the unique decision-making procedure in trials by jury. There is something arbitrary about making a method mandatory when one knows that, by and large, there is another method that works just as well (or just as bad). In making deliberation mandatory, we treat it as the best available procedure for deciding the result of a jury trial. By doing so, we fail to recognize and act on all the evidence we have about the equivalence between deliberation and the aggregation of the jurors’ individual verdicts.
This raises a question about the way in which the criminal justice institutions should respond to the current evidence about deliberation. In what follows, I outline three forms that such a response could take. The first one is the farthest away from the status quo and, as such, might not represent the most readily applicable reform. The other two proposals are comparatively more feasible and could work as alternatives to the status quo that could adequately play a transitional role until a restructuring of the deliberative process warrants its unique mandatory status. After having outlined these three proposals, I will point to some of the ways in which they – and, in particular, the multiple juries proposal – connect to democracy.
Making Jury Deliberation Virtual
One solution for addressing the epistemic pitfalls of jury deliberation is to
restructure it in a way that eliminates its face-to-face communication aspect. This will
be successful in improving deliberation only if the problem with deliberation lies in its
direct interactional dimension. Eliminating this dimension could be done by resorting to
virtualizing deliberative practices, i.e. by replacing the current jury room where jurors
deliberate in each other’s presence with a mediated, virtual environment where jurors have no direct mutual contact and perceive others as anonymized avatars.
In addition, as Hedden (2017) recently suggested, the virtualization of the deliberation process could be accompanied by standardizing it following the DELPHI method. This method works in a (potentially iterative) sequence of three stages: (1) the expression and collection of anonymous individual judgments, followed by (2) a summary presentation of these judgments to the group members and (3) a new round of expression and collection of individual judgments. The merits of such a method is that it addresses some of the biases and distortions that accompany the social pressure and informational cascades present in the face-to-face format of jury deliberation.
The proposal to move from real to virtual juries draws on Benforado’s (2016) recent suggestion that we should replace real trial proceedings with virtual ones. The argument for this reform is that it might help eliminate some of the unwarranted biases that distort trial impartiality in its current set-up. ‘The simulated space of each trial – including the jury’s vantage point on the witnesses, the color of the courtroom walls, the amount of light from windows, and the height of the judge’s bench – would be uniform, so that all participants would know exactly what to expect’ (268). The idea, then, is to extend Benforado’s proposal from the courtroom to the jury room, in a way that virtualizes the latter following a set-up that is similar to the proposed virtualization of the former.
Note, however, that there is no guarantee that the virtualization of deliberation will make it fool-proof to the risk of biases or epistemic pressure. This is because the virtualization of jury deliberations will be able to get rid only of some unwarranted information and because the filtering out of this information will sometimes be itself influenced by the biases of whoever does the selection or is in charge of designing the selection algorithm.
Moreover, even if coupled with the DELPHI method, virtual jury deliberation
might have its epistemic prospects improved in a way that makes it decisively better
than pure aggregation. Though the DELPHI method might work effectively in
neutralizing the effects of social and informational pressure, this is not bound to happen
in all cases, and there is some evidence showing that it might not happen significantly
more if we resort to the DELPHI method as compared to face-to-face interactional
methods (Gustafson et al. 2003).
Lastly, turning toward virtual jury deliberation is not an institutional reform that we can expect any time soon. This is both for cultural reasons that have to do with judicial and social inertia and for empirical reasons that have to do with the scarce evidence we have about the effects and implications that virtualization would have on the quality of the deliberative process.
Randomizing Aggregation and Deliberation
A second, and readily available proposal is to introduce a system whereby the choice of the decision-making procedure is randomized between a purely aggregative and a deliberative procedure. This would avoid the current situation where deliberation is imposed in an arbitrary way. Instead, we would leave it to chance whether the outcome of a trial is decided following the aggregation of non-deliberative individual verdicts or according to a standard (or duly modified) deliberation procedure.
Randomization, in this case, provides us with a mechanism for not acting on inconclusive reasons. Acting on inconclusive reasons is, in some significant sense, acting on bad reasons. Peter Stone (2007) argues that, when our reasons for making a decision are relevantly underdetermined – that is, when we do not have comparatively conclusive reasons for preferring a course of action A to a course of action B –, but we are compelled to take the decision anyway, randomization provides an impartial (and, as such, fair) mechanism for making that decision.
Similarly, Vincent Chiao (2012) has argued for randomization as a mechanism for selecting the criminal sentences that are going to be effectively enforced, given the fact that, because of limited resources, the criminal justice system is unable to enforce all of them. Chiao contends that randomization is justified on grounds of ex ante fairness, that is, on grounds that there are no good reasons before the actual imposition of the criminal sentences on the basis of which the convicted defendants whose sentences are going to be enforced can object to their being submitted to those sentences. More relevantly, lotteries would eliminate the biases that often stand behind the decisions about which defendants actually end up serving their sentences.
There is a sense in which the decision concerning the procedure for deciding
criminal jury trials has the same underdetermined structure as the one identified in
deciding which of the guilty offenders are going to have their sentences enforced. The
analogy is that, as indicated above, we have no good ex ante reasons for preferring
aggregation to deliberation as the privileged procedure for deciding the outcomes of
trials by jury. To be sure, there are good reasons for preferring either one of these two procedures over any other procedure that is clearly random (rolling a die) or worse than random (the ordeal). But, given the evidence about their epistemic record, there are no good reasons for opting for aggregation or deliberation in particular. If one has to choose one of the two, it seems that randomizing that choice is the best way of choosing.
Note that defendants could not reasonably object that their case is decided on the basis of aggregation instead of deliberation or the other way around. This is because, for all that we know, the two procedures are epistemically equivalent. This is also because defendants could not point to other procedures that would be both available and better placed – that is, comparatively more accurate – than either aggregation or deliberation in deciding their verdict.
Note also that there is a sense in which deliberation is not completely absent from the aggregative procedure. If we adopt Robert Goodin’s (2003) argument about
‘private deliberation,’ individual judgments can be considered as quasi-deliberative. In examining their judgment, it is plausible to see jurors as engaging in counterfactual reasoning among different potential perspectives and contending views about the case at hand. There is thus a sense in which, because it involves a process whereby distinct reasons are weighed and challenged, the act of judgment is also one of ‘deliberation within’ (Goodin & Niemeyer 2003).
10Finally, a system whereby aggregation and deliberation are randomized would have the advantage of creating the conditions for a natural experiment that would allow us to better compare the epistemic merits of deliberation and aggregation. Currently, the evidence we have about the epistemic record of these two procedures is purely experimental and is based on simulated situations. Moving from the laboratory to the actual jury room would present the advantage of testing the validity of the evidence that we have about the two procedures.
A jury system that randomizes the use of aggregation and deliberation is a multiple jury system in the sense that there are two kinds of procedural design on whose basis jury trials are decided. Multiplicity here would be a property of the jury trial system considered as a whole and not one of each particular jury trial. In the following section, I provide an outline of how a system of multiple juries would work when individuated at the level of each trial by jury.
10
