Joint intentions are plans held in common by members of a group to structure their joint action to their shared end. The group’s intention is the rational order that explains and structures its action, just as an individual person’s intention defines and explains his action. The intentions of a rational person
follow from and are explained by the individual agent’s response to reasons, which there is good reason to expect to be coherent and at least somewhat stable over time. I consider now whether there is good reason to expect the intentions of groups to be coherent and consistent and to follow from the group’s response to reasons. The plans of any group arise in some way from the intentional action of its members. When the group acts on related propositions, there is a standing risk that integrating the rational intentions of individual members of the group into the joint intentions of the group will result in collectively irrational action. That is, the group may form an incoherent, inconsistent or unreasoned plan that is an unfit means to the ends for which the group acts. I maintain that a group may avoid this outcome if its members attend to the relations amongst propositions and aim to act jointly like a rational agent.
The problem of collective irrationality is explored with care in List and Pettit’s important work on group agency. They begin by discussing groups that use majority voting procedures, but their argument is, as they make clear, of more general application.45 Their discussion proceeds by way of a
reflection on the so-called doctrinal paradox in appellate court decision-making.46 The paradox may arise when a court of three or more judges has to decide an issue that turns on two or more premises, say, the issue of the defendant’s liability which turns on (i) whether he caused the loss and (ii)
whether a duty of care existed. The court decides by each judge taking a view on the premises
(causation and duty of care) and then voting on the conclusion (liability). There is a doctrinal paradox if the pattern of votes is as follows:
A majority of judges conclude that the defendant was not liable. That conclusion is not supported by the view of a majority of judges on the relevant premises. Instead, a majority of judges believed that there was causation and another majority believed that there was a duty of care. Despite a majority of judges believing that there was causation and a duty of care, the aggregation of the rational votes of each judge yields the conclusion that there is no liability. Yet if the court had decided by voting on the relevant premises and taking the premises to dictate the conclusion, it would have reached the
opposite decision. Hence, the choice of decision-making procedure can be very important, for a premise-based procedure leads to one outcome, and a conclusion-based procedure to another.
This example is important and illuminating, but also somewhat obscure. In reflecting on the doctrinal paradox, List and Pettit argue that:
Relative to the given legal doctrine—that obligation and action are jointly necessary and sufficient for liability—the set of propositions accepted by a majority, namely that the defendant had an obligation, did the action, and yet is not liable, is
inconsistent. If adherence to legal doctrine is considered a requirement of rationality in a court, then majority voting will prevent the court from meeting this requirement in the present case.47
The point of the court is to decide the dispute, which is to reach a conclusion on the appellant’s liability. The court uses majority vote to reach this conclusion. List and Pettit take the court to form judgments on each premise and on the conclusion. However, the court need not use majority vote to reach a position on each premise: the judges may each form views on the premises and then vote on the conclusion. Hence, the pattern of votes—or rather of views on premises and then the vote on the conclusion—does not entail that the court’s decision is inconsistent. The decision would be
inconsistent if, as List and Pettit present the example, the court were to decide on the premises by way of majority vote and then on the conclusion in the same way, for then the court would decide ‘yes, causation and duty of care, but yet no liability’, which is flatly inconsistent. Inviting the court to vote on premises and conclusion would be to invite just such inconsistency, because if the court is to take a position qua court on the premises, this will entail a consistent conclusion. The subsequent vote on the conclusion would then offer only an opportunity to affirm the entailed conclusion or to decide inconsistently with the premises that support a particular conclusion. Importantly, however, the court need not decide on premises and on conclusion. Each judge must decide on all three propositions, but the court may settle the dispute by asserting just that the defendant is not liable. The court’s decision is then not collectively irrational because inconsistent. However, it is irrational to the extent that it is incomplete. That is, the court’s decision does not follow from reasons that the court itself adopts and takes to justify its decision.
The court is able to attain its end, the fair adjudication of disputes, without itself adopting reasons. Instead, the dispute may be settled by the intersection of the rational decisions of each judge, each of whom takes a consistent view on premises and conclusion. The court’s decision is complete and follows from reasons that the court itself adopts when the majority’s conclusions follow from the same chain of reasoning. It is not unfair for a dispute to be settled by the conclusions of judges, each of whom reasons carefully, notwithstanding that the court itself does not act for reasons. Still, there are good reasons to think this adjudicative act less than satisfactory—if the court itself acts on
reasons, it is clearer that the dispute was settled rationally and the merits of the settlement are open to scrutiny. Further, the secondary point of adjudicative decisions in the common law world, to restate legal doctrine and perhaps to change it in relevant part, cannot be secured if in articulating and applying the law the court itself fails to act on a complete set of reasons. I do not mean that an incomplete decision unsettles the content of the law, although it may, but rather that it fails to constitute what the common law method of legal change requires, which is reasoned judgment.
The doctrinal paradox, List and Pettit argue, is just one instance of a more general discursive dilemma, which is the standing possibility that ‘majority voting on interconnected propositions may lead to inconsistent group judgments even when individual judgments are fully consistent’.48 In earlier work, Pettit argued that this possibility poses a dilemma because groups could be structured to reach decisions either by a conclusion-centred approach (voting on the conclusion) or a premise-centred approach (letting the vote on the relevant premises be decisive), and each approach has its
advantages and disadvantages.49 In his recent and most sophisticated work with List, he argues that groups should resolve the dilemma by forming rational agents, although this need not call for a strict premise-based approach.
To form plans of action by way of majority vote on a series of related propositions is to risk adopting plans that are inconsistent. This means that the group may form a plan that is an unsound means to the ends for which the group acts. To illustrate the point, consider another example of Pettit’s, a political party formulating policy.50
Pettit stipulates that if defence spending were increased, other spending could not be increased without raising taxes (there is no deficit borrowing and/or the point of the policy is to balance the budget). Therefore it would be irrational (as impossible) for anyone to set out to increase defence spending and other spending without also raising taxes. Yet the pattern of votes above would lead the party to adopt this policy. The joint plan is inconsistent, such that it is not fit to be chosen by any rational person. The plan is not an intelligible response to the relevant reasons for action. It is collectively irrational because it is an inept, incoherent means to the group’s end—developing
credible policy proposals, which may serve as the platform on which to contest and win elections and then to govern well.
Interestingly, Pettit imagines each of these questions being considered at different points in time. Assume that per the table above the members of the party have at T1 voted not to raise taxes and then at T2 voted to increase defence spending. He argues that a discursive dilemma arises at T3, when the question is whether to increase other spending. The decisions at T1 and T2 provide a reason for the group not to increase other spending. However, if the party reaches its decision at T3 on the basis of ordinary majority vote, without adopting procedures to take T1 and T2 into account (which may include repudiating either decision reached at T1 or T2, or letting the issue at T3 follow from the prior two decisions), then, as the pattern above indicates, the party will adopt an irrational policy platform. The individual members of the party, Pettit asserts, have voted rationally, but their votes yield a collectively irrational decision.51 (I argue below that the votes at T3 are not reasonable.)
This discussion makes clear that there are two dimensions to collective irrationality— incompleteness and inconsistency—and that the importance of joint action being consistent and
reasoned turns on the point of the group. The group that maintains collective rationality is thus a group structured to adopt plans that constitute intelligible responses to reason, and which are consistent, coherent, and complete.
The use of majority vote to form complex plans risks collective irrationality. This risk generalizes from majority vote to other procedures, such as supermajority or unanimity rules. Indeed, List and Pettit establish that it is impossible for any group acting in response to some set of propositions to maintain collective rationality if the procedure it adopts meets three other conditions.52 The first is universal domain, which is to admit all possible consistent and complete individual attitudes. The second is anonymity, which is to treat all individuals equally in forming group attitudes. The third is systematicity, which is to provide that:
The group attitude on each proposition depends only on the individuals’ attitudes towards it, not on their attitudes towards other propositions, and the pattern of dependence between individual and collective attitudes is the same for all propositions.53
The upshot of this impossibility theorem is that one cannot square collective rationality with universal domain, anonymity, and systematicity.