Legislation, Planning and Deliberation

PA Wire/Press Association Images
Print Friendly, PDF & Email

The planning perspective is a democratic conception of legislation that relies on two set of normative obligations which may come into conflict. John A. Ferejohn examines ways of dealing with that conflict. This paper is part of an issue on Collective wisdom.

Introduction

Legislation directs and regulates coercive state powers and anyone subject to those powers is entitled to an explanation of why her interests are to be subordinated to those of others or the state1I owe special thanks to Lewis Kornhauser for extensive comments on this paper, not all of which I have been able to take account of, and to Seana Schiffrin for many comments on an earlier version of it. If you give me comments, you too may have a place in this footnote. These people are blameless for all the faults that remain. Early versions of some of these ideas have been presented at the Dworkin-Nagel Colloquium (I think in 2001), at a conference at the Australian National University in 2004, and at a lecture at the University of Minnesota Law School in 2005. I am going to keep at it til I get it right. . This justificatory burden requires a statute be seen as an action aimed at furthering some legitimate state interest.2 What counts as a valid or legitimate purpose is a complex topic that is not addressed here beyond saying that validity would include both procedural elements – that the statute was enacted according to established and fair procedures – and substantive ones as well (including constitutional and possibly other criteria as well).

A statute therefore has to be interpreted as an intentional action but people disagree as to how to do this. Some judges think that statutes should be interpreted as their drafters intended; others think they should be interpreted as the legislature which enacted them intended. There is also disagreement as to what evidence is relevant for making intentional attributions but all approaches share the idea that legislation is a kind of intentional action and therefore sound practices of interpretation ought to take account of this fact.

Legally the main issue has to do with whether the state’s purpose or interest is of the sort that could possibly justify state coercion and whether the coercive measures it authorizes are proportionate to that interest.3 This is a complex question that is resolved in various ways in different legal systems. But a prior question has to do with the best or more appropriate attribution of legislative intention. Is the statute aimed merely to further the particular aims of the individual legislators who happened to shape and support the legislation? If that were the case, the explanation offered for coercion would simply amount to telling the complainer that this or that majority simply happened to prevail on the crucial vote according the procedures of the legislature. That kind of theory seems incapable of playing a strong justificatory role. The alternative is to say that the legislation was intended to achieve some legitimate public purpose in a way that represents a reasonable balancing of your interests and that of the state. Presumably such an account is needed for justification.

Legislative intentions capable of playing this justificatory role must be relatively stable over time and not merely the (momentary) intentions of a particular legislature or any of its particular members. This is a controversial claim. Some would argue instead that the only significant (legal and moral) intentions are those of the enacting legislature because it was the legislature that enacted the statute. While the text enacted by the original legislature has a great deal of evidentiary authority in attributing the intentions and beliefs of the enacting legislature, the beliefs and expectations of the enacting legislature may not have been accurate guides to how it would apply over time. For this reason, notions of original intent that privilege the intentions of the enacting legislature give the wrong kind of guidance about the purposes to which the statue is directed. Later lawmakers will have had the chance to learn much more about the circumstances of application of the original statute. Having the benefit of experience in applying the law, their expectations and beliefs are likely to be more accurate than those of the original legislature.4 The text of the statute, which was produced by that legislature, is another matter. Any theory of legislation – normative or positive needs to take account of text as both something that is fought over in the legislature and as having constraining effects on what the law can be subsequently. And as they will have had many opportunities to update or change or even abolish the statute in light of these new developments, it seems to me plausible to think of them as continuing to intend to pursue the same aim, more or less, as the original legislature. So I want to think of the intent of the legislature as attributable to the legislature conceived as an entity that exists stably over time.

I want to distinguish between a realist notion of legislative intent, which sees intentions as playing a causal role in enacting statutes, and an instrumentalist notion which sees legislative intent as a “posited” entity, which provides a rationale or interpretation of the statute rather than a causal account of it. There are two kinds of realist theories of legislative intent: aggregative theories, which legislative intention as an aggregation of the actions and intentions of the individual members – perhaps those of the majority that prevailed in a particular vote or that of the “median voter” if one exists – and holistic theories, which see the legislative intentions as irreducible to individual actions or intentions.

According to an aggregative theory legislative intentions tend to be incomplete, incoherent and therefore irrational. Such a theory may be able to explain a statute, in the sense of showing why it was enacted or took the shape it did, but it generally cannot explain it as intelligible: as the action of an intentional agent. By contrast a holist theory sees the legislature as capable of having an intention of the kind that (rational) individuals do. Of course, unless there is some reason to believe that a “real” legislative intention exists played a role in shaping the legislation, it would be hard to see a holist account as realist.

Ontological doubts of this kind have led some to argue that the only plausible candidates for a holist idea of legislative intention are interpretive.5 One could also imagine an instrumentalist aggregative view. Such a view may have scientific value I suppose in permitting the attribution of intentions to members or coalitions. Interpretive which “construct” legislative intent as the best explanation of the legislative materials (text, history, applications, etc.) are often informative and sometimes indispensable for courts and agencies faced with applying a statute. But, this kind of interpretation lacks a certain kind of force unless it is claimed that that intention actually existed and was the reason the legislature enacted the statute. But such a claim is realist and not simply interpretive. I shall argue that there is a plausible realist construction that can play the appropriate role in shaping statutes as well as providing later interpretations of them. The idea is to see statutes as the plans of a continuing intentional agent aimed at coordinating future actions.

There are reasons – grounded in democratic norms — that notions of continuing legislative intentions of the kind discussed here may be resisted. Legislatures, for example, are not supposed to be able to bind their successors.6 Formal precedents that govern how a legislature conducts its business are often important of course. The procedural rules of both the United States House and Senate are often found in precedents. The idea that new legislation could be constrained by previous legislation ways seems undemocratic in interfering with the capacity of the present generation to govern itself. While I think there is something to this belief, it is often too crudely stated to be correct.7 In the areas of taxes and appropriations Congress frequently sees itself as bound by its previous commitments. This is also the case with government contracts. Legislative “entitlements” represent, perhaps, a weaker sense of bindingness since Congress does sometimes adjust entitlements. For a detailed treatment of this issue see Eric Posner and Adrian Vermeule, Legislative Entrenchment: A Reappraisal, Yale Law Journal , volume 111 (May 2002), 1665-1705. In any case, my construction sees the current legislature as having the intention so there is no sense that the current legislature is constrained. Indeed, legislative intentions do not generally bind by constraining a future Congress from taking some action. Rather, these intentions are a source of reasons for action, in just the way that a person’s intentions provide reasons for her to act in a certain ways.8 If you intend to drive to the store from your house, that intention gives you reasons to get in the car, back out of your driveway, turn left or right at various intersections. You could refused to do these things but then the chances are you will not fulfill your intention. The power of intentions to shape actions is normative in that sense.

Some idea of collective intent seems required to make sense of the idea that “we” – as a nation and perhaps each of us — have ongoing (normative) commitments to certain policies, and that these commitments arise (partly) from the fact that those policies were intentional policies adopted by “our” legislature. These commitments are formed in circumstances of conflict and disagreement and their content and scope are often politically and legally controversial.9 This is not to minimize the role of judges or agencies in construing statutes but only insists that they ought to see a statute as intentionally directed and try to construct or recover the relevant intentions when applying the legislation. The language of collective intent is familiar in the usages of the Supreme Court: someone writing an opinion typically refers to what we or this Court decided when referring to decisions a century old (and with which the current opinion writer may disagree). The idea is that those previous decisions establish a continuing policy of the Court itself that is (normatively) binding on its current members in any similar case.10 While the notion of the Court, as a multimember body, acting as an intentional agent may be familiar, it is no less mysterious, in the terms of this paper, than the idea of a legislature having an intent. For a court, this is not a mere manner of speaking, but a recognition by members of the current court that they have duties to present the court, as a continuing body, as a unified agent acting over time, capable of offering to possible litigants a coherent account of their legal rights and duties. The notion that horizontal precedent has binding force on the current court is a normative claim. It is an expression of the particular ethical notion of how a judge on a court operating over time ought to behave. I argue that similar commitments and concomitant supporting norms can be traced in the legislature as well.11 Kornhauser and Sager, in a recent article urge caution in comparing courts and legislatures in respects under discussion in this paper. Lewis Kornhauser and Laurence Sager, “The One and the Many: Integrity and Group Choice in Paradoxical Cases, “ Philosophy and Public Affairs, vol 32 (2004) 249-276.

No-one could plausibly defend an intentional theory of legislation unless statutes normally displayed some degree of coherence. I shall argue that there are some grounds for this belief: members have normative reasons to try to make their statutes coherent and, if they can resolve some difficult coordination problems, one can expect them to be relatively successful in doing this. Coherence may therefore be brought about in the course of deliberation. Deliberation will not, of course, always succeed in producing coherent legislation. Coherence is one value and is in conflict with others and any particular statute may be imperfect in this regard. But, it is an especially important value, and I expect that legislative deliberation will normally give it a great deal of weight. The theory of legislative intention developed here is in this respect a theory of democratic deliberation.

An Example

The issue may be illustrated by the recent debate over how to reform the social security program. The Social Security Act of 1935 arose out of a deliberate campaign by President Roosevelt to persuade Congress and the people to embrace a new program for economic security – one that would create a new kind of economic right or entitlement to income during retirement. He explained the need for the provision of a new economic right which he said were made necessary by structural changes in the economy undermining the capacity of families to support their older members. He established a commission made up mostly of economists and insurance experts to come up with a legislative plan which produced a blueprint for the new program. His congressional allies engaged in extensive congressional debates aimed at crafting and justifying the new legislation which was then enacted. Subsequently, Congress revisited the program several times (most notably in1939, 1950, 1972 and 1983) to amend and usually to expand it, as well as to make its promises more credible and secure. Each revisiting was predicated on the notion that the United States had undertaken a long term commitment, in the Social Security Act, to provide certain kinds of “guarantees” its citizens that would enable them to plan their lives, and that it (the Congress) recognized an obligation to ensure that the program was structured in a way to redeem the promises it made in the Act.

Changing external facts arising from demography and fiscal constraints, as well as from the automatic cost of living adjustments in benefits formula have made social security reform more pressing . Republicans, especially in the House of Representatives, have long been encouraging leaders to push for reforming the original program by introducing private accounts that would be “owned” and perhaps controlled by individuals. On its face, a system based on private accounts seems to depart from the social insurance principle of the traditional system. Should we therefore understand the position of the Republicans in this debate as simply a call to put an end to the social security program altogether and put some new program in its place? Or, should we see the Republicans as proposing private accounts as somehow fitting in with longstanding congressional intentions and merely proposing a more efficient set of means to achieve it? On the latter account, Republican proposals for private accounts would be offered as a better interpretation of an abstract congressional intention to provide retirement income as applied to modern conditions.

This posture – framing the argument as one about ongoing congressional intentions – offers powerful rhetorical leverage. It sees advocacy of a policy as an argument about the content of an ongoing commitment. Advocates of private account seek to persuade those who believe that congress is committed to social security by making arguments that they could embrace and accept without abandoning this commitment. Conversely, defenders of the current system can oppose this interpretation of collective intentions either on the grounds that it is insincere, or else that it fails to assure adequate guarantees of security and is therefore inconsistent with longstanding congressional intent. That is to say, they reject the notion that private accounts would be a better interpretation of the congressional commitment to provide adequate social security to everyone. Thus, for supporters as well as opponents of the proposed reform, the argument centers on congressional intentions or commitments.

Two Conceptions of Group Intentions

Like individuals, legislatures need to decide what to do. They are large and complex organizations which are distinctive in having little control over their membership, but they need to coordinate the actions of their diverse members through time in order to gather information, make decisions, and take action. The actions they take are not usually self executing but rather establish programs and policies that extend through time and which have the aim of coordinating the actions of millions of other people, among them judges and executive officials. Statutes, on this view, are plans for the future, and legislating amounts to deliberating about plans.

This conception of deliberation arises from seeing the legislature as a “planning agent.” According to Michael Bratman, deciding what to do – practical reasoning – generally ends in settling on an intention to take some course of action. To form an intention to do something is to commit to a plan of action that may need to be developed or “filled out” over time. According to Bratman to intend to do something is, in effect, to adopt a partial or skeletal plan that will eventuate in the intention being realized. As time goes on a planning agent acquires new information and, normally, tries to fill in partial plans to which she has committed in light of these new considerations. A planning agent may decide to modify her plans later on or even abandon a previous intention (and associated plan) because its demands turn out to be too arduous or because its fruits come to seem less attractive. But abandoning plans to which one is committed is a serious step not usually undertaken lightly.12 Michael Bratman, Intentions, Plans and Practical Reasoning, Cambridge: Harvard University Press, 1987.

Legislatures are different than ordinary planning agents in some respects: internal partisan divisions are both legitimate and sharp and are even institutionalized in various ways. Moreover, an electoral result can give the legislature a special reason to abandon plans. But, if anything, the coordination value of plans seems more important for a large multiperson institution than in is for an individual. Moreover, it is in the business of making norms binding others, the legislature has a special stake in making clear and stable rules capable of guiding behavior. For that reasons a legislature has especially strong reasons to make plans and keep to them not only as a way of organizing its further actions and deliberations, but also to coordinate the actions of others to whom it can direct commands (judges, state officers, ordinary citizens). A legislature that regularly gave up its previous commitments would not only be disorganized and ineffective at decision making but it would also command little obedience. So there is good reason think that legislative deliberation can be understood as planning activity, that statutes are (incomplete or partial) plans with an intentional aspect, and that subsequent legislators are obliged to see themselves as participants in filling out incomplete statutory plans.

To allege that legislatures are planning agents is to claim that it can act in some respects as an agent or person so we need to have an account of how this is possible. One approach, exemplified in work by Michael Bratman, develops a conception of shared or group intentions from the “ground” up in the sense that group intentions are constituted by appropriate configurations of personal intentions of their members. On his account group intentions are not regarded as actual mental states of a group but are configurations of mental states held by group members. Thus, while group intentions (like individual intentions) are propositional attitudes they are not themselves mental states of an agent; rather, they correspond to complex combinations of the mental states of the group’s members.13 “Shared intentions are intentions of the group….. what they consist in is a public, interlocking web of intentions of the individuals.” (Bratman, Faces of Intention, Cambridge: Cambridge University Press, 1999, p.43. Thus, as with the social choice theory account, Bratmans theory reduces group intentions to individual intentions but the kind of reduction is not aggregative in the sense of social choice theory at least not obviously so — but constitutive. This construction has the virtue of being “ontologically conservative” in that there are no new autonomous entities such as group minds. But, is not clear that his construction can work for large and complex groups such as legislatures.

A joint or shared or group intention (I shall use these expressions interchangeably) is different from a personal intention in the following way. We can understand personal intentions as intentions to do something. This formulation takes account of the fact that one can only intend to do something that one is able to accomplish at least in more or less normal circumstances (assuming you are not struck by lightening, etc.). But Bratman wishes to avoid introducing a wholly new entity for example, a group that can form intentions in just the same way that individuals can and so he connects group intentions to more basic individual intentions.14 Whether individual attitudes are rightly thought to be simpler or more primitive is a controversial matter. There some complications for group intentions: individual members cannot normally form ordinary intentions – intentions “to” X – on behalf of the group because individual members are not normally in control of a group’s actions in this way. But a member can intend that the group take some action, where having a “that” intention requires the individual to do her part in bringing about X. And, in Bratman’s account, if members have the right combination of “that” intentions then the group will have a shared or group intention.15 Shared intention, as I understand it, is not an attitude in any mind. It is not an attitude in the mind of some fused agent, for there is no such mind; and it is not an attitude in the mind of either or both participants. Rather, it is a state of affairs that consists primarily in attitudes … of the participants and interrelations between those attitudes. (Bratman, Faces of Intention, 122-3.

Intending that will involve the same kind of constraints as intending to does. These constraints, arguably, follow from features of planning agency. Thus the principle of agglomerativity implies that a group member cannot (knowingly) intend that the group achieve two contradictory results, and so the group itself cannot have a shared intention to achieve inconsistent objectives. Group members can of course want or wish that the group to do inconsistent things but intention cannot have this feature. Moreover, group members cannot have intentions that the group do a thing that the group cannot (normally) bring about and so shared intentions seem subject to further practical constraints in the same way that individual intentions are.

In addition there are vertical relations among shared and personal intention and actions. If I intend that we (our group) do x, I would, normally, have appropriate personal intentions to do my part in bringing x about. Or at least, I must be disposed to form such personal intentions along with others in order that the groups intention (my intention that) be realized. The part I am to play in the plan has to mesh with the plan. And you would have to have similar individual intentions to do your part.16 What my part and your part amount to is a complex issue. Our parts may be fully specified in the partial plan in some cases. In others there may be some background idea of fairness that would play a role in determining what our individual obligations are. I owe this observation to Seana Shiffrin. Liam Murphys work is relevant here. Meshing also requires that both of us will feel normative pressure to bring our other plans into line with our shared intention. Finally, in some way, we must each know that these considerations are satisfied.

Bratmans formal account of shared intention goes like this. 1. I intend that we X and you intend that we X. 2. I intend that we X by means of 1. (including meshing subplans of 1.). And, you intend that we X by means of 1. (including meshing subplans in 1.), and 3. All this is common knowledge between us. An important aspect of this definition is that it defines shared intention in terms of a complex interlocking set of individual attitudes (intentions and beliefs). Moreover, shared or group intentions require unanimous personal intentions that. This condition implies that properties of shared intentions to including coherence properties will be inherited from the properties of the individual attitudes.

Thus, Bratmans definition of a shared intention seems delicate precisely because it requires that an elaborate set of interlocking attitudes of beliefs and intentions must be held by the group’s members. In larger and more complex groups, it is difficult to believe that such conditions could be achieved very often or at all. So, some writers have attempted to relax these conditions somewhat while keeping the idea of shared or group intention. This may be a fruitful path. Scott Shapiro, for example, has sought to weaken both the unanimity condition and the common knowledge condition for shared intention. This may be a fruitful idea but one problem with it that the coherence of shared intentions will not follow from the coherence of the constituent personal attitudes. Moreover, insofar as his modifications stay near to Bratmans theory, it is not clear that Shapiro’s approach will permit the formation of very many shared intentions.

Other theorists think that shared or group intentions are not reducible to individual intentions in the way that Bratman others seek to do. Margaret Gilbert argues that shared intentions are like agreements, though these agreements are implicit in many circumstances. That shared intentions are virtual or implicit agreements introduces two new features. First, Gilbert thinks that the creation of a shared intention entails the creation of something she calls a plural subject, an entity that can have intentions in just the same way that individuals can. The intentional attitudes of this subject are somewhat independent of those of its constituent members. Second, in the relationship between group members of a plural subject, an additional normative element enters the picture; something like obligations to keep “agreements with others. I think that these two strands are separable but that each is compatible with a kind of “realism.” The ontological account envisions a new psychological entity which could, presumably enter into causal interactions with things in the world. The second, normative, account would permit a different kind of normative realism presenting reasons for action rather than causes. It is this second idea that I will explore here.

As I understand her argument, for a group to agree, in Gilberts sense, to form a plural subject, there must (already) be a convention to mark what counts as agreeing. Her story about how two people come to be walking together describes a kind of elaborate signaling ritual in which separate walkers converge (perhaps wordlessly) on a joint commitment to walk together and not merely in parallel. In the case of a two person interaction, such conventions operate implicitly and probably have the feature that forming a plural subject requires unanimous assent. We are not walking together unless both of us have somehow committed to walking together. Gilberts shared intentions, once formed, are autonomous from the mental attitudes of group members. Two people, walking together, constitute themselves as a plural subject, capable of having intentions and other mental states, and capable of responding to new circumstances as any rational agent would.

In larger and more complex organizations things are different. We cannot generally expect to have implicit background conventions that would permit all of us to recognize when we have acted as a plural subject. Moreover, the formation of an intention by a plural subject cannot plausibly require unanimous agreement or anything as strong as Bratman’s meshing condition. We should expect that, in the case of larger groups, more formalized norms will be relied on to describe the circumstances in which a plural subject has been acted. A large corporation for example can be committed to a course of action by acts of its CEO or of its Board of Directors. Legislatures, our example, can take action through their formal rules for proposing and accepting proposals which usually require the assent of only a part of the body. Still, Gilbert would think the actions and policies of a corporation or a legislature are the acts of plural subjects.

Bratman and Gilbert agree that joint intentions put normative pressures on the actions and intentions of the group members. Bratman thinks this pressure is weak in the sense that it is grounded in the same kind of stabilizing or coordinating force that exists for individuals as planning agents. If you and I have a shared intention (and have adopted a partial plan) to X, then that gives us reason to coordinate our actions and intentions on pain of being indecisive or inconstant, or perhaps of acting suboptimally in disregarding our cognitive limitations. Gilbert argues that additional and stronger kinds of interpersonal obligations come into play at the group level, so that failing to keep up our part of a joint intention essentially involves failure to fulfill the obligations that arise from the shared intention itself.17 One of Gilberts criteria for identifying a joint intention is the existence of obligations (or at least a sense of obligation) to pursue and help the others in the group to pursue the joint intention. These obligations have something of the same force as the obligation to keep promises. Even so, if such obligations conflict with individual intentions or other commitments, such obligations might be overridden, even at the price of having to suffer rebuke from other group members.

The plural subject view offers the prospect of a realist conception of legislative intent. Plural subjects, on Gilberts account, have genuine intentions and these intentions are autonomous from the intentions of group members. Gilbert’s conditions may trigger worries about ‘realism.’ Shared intentions are supposed to be mental states and such states are supposed to require minds, and perhaps minds require brains. Bratman’s solution is attractive precisely in that shared intentions are seen as (complex configurations of) individual mental states, and those (individual) states are linked to brains in the ordinary way. Gilbert’s view breaks the link between group and individual mental states and so it raises questions about the status of group mental attitudes and indeed about group minds. She could argue that positing group minds is objectionable only if group minds are considered to be very much like individual human minds. But group minds may simply lack much of the power of human minds and simply be places where quite sparse and primitive mental attitudes are housed.18 For an exploration of these issues see Philip Pettit, ‘Groups with Minds of their Own’ in Frederick Schmitt, ed., Socializing Metaphysics, Rowan and Littlefield, pp. 167-93.

In any case, the intentions of a plural subject are supposed to respond rationally to other group attitudes, desires and beliefs, in more or less the ways that individual intentions: we expect a plural subject’s intentions to be rationally related to its beliefs and desires. But, whatever it is that regulates the relations among group mental attitudes may be very different than the processes that work for individuals. That regulation may work through normatively guided social processes rather than neurophysiological ones. If that is right, Gilbert’s group intentions lack some key features of individual mental attitudes. They are not realized any kind of neural machinery and they may lack what might be called “subjectivity” in the sense that the group as such may not have any privileged access to them in the way that ordinary individuals might be thought to have special access to their own attitudes. More importantly, the way that group and individual intentions interact may be completely different from the way individual intentional attitudes do:” normatively rather than chemically.

Because Gilbert’s group intentions are not reducible, as Bratman’s are, to complex sets of individual mental attitudes, and are built instead on the public actions (virtual “agreements”) of the members her theory seems to offer a plausible basis for a realist theory of legislative intent. Her “agreements” could be realized both through institutionalized rituals such as votes in a legislature as well as in the less formal ways she discusses. Moreover, her idea that group intentions are the intentions of an autonomous subject forces us to ask how it is that group mental attitudes are made coherent. Specifically it requires us to focus directly on the normative (rather than constitutive) relationships among group members and the intentions of their plural subject. That is, it permits us to ask directly about the normative obligations group members may have to help the group pursue its purposes. These features make the notion of a plural subject a plausible grounding for a realist notion of legislative intent.

A digression on collective rationality

There are reasons to be skeptical about group or social rationality: the idea that groups could have intentions that are similar to those we think individuals’ have. The results of social choice theory are usually taken to demonstrate precisely this: that decision procedures commonly used in legislatures and electorates (and multi-member courts) cannot be counted on exhibit coherent patterns of decisions in all circumstances in which they are supposed to operate. Empirically, generations of political observers have argued that public policy is best understood not as intentional action but as a record of the outcomes of sequences of group struggles. Coherent intention on such views would only be expected if very unusual conditions happened to obtain.

I think the results of social choice theory are less subversive of the notion of legislative intention than the above remarks suggest. Indeed, the theory provides some guidance as to how coherent collective outcomes could be brought about within, for example, an institution that made its decisions by majority rule. Suppose the members of the collectivity place a value on coherence and were willing to sacrifice achievement of other values to bring it about. In that case processes of deliberation and compromise may induce people to modify their initial preferences in ways that reduce the heterogeneity of post deliberation preferences. This reduction in heterogeneity may be sufficient to permit a decision rule to produce coherent social judgments. I shall explore a version of this idea below. But there are other ways too: legislators might, again as a result of deliberative consideration, decide to require supermajorities or even unanimity to take action (even though the formal rules permit action by mere majorities) as, for many purposes, the United State Senate does in conducting its ordinary legislative business. This too can produce coherent collective preferences according to results in social choice theory. There are other ways as well. The point is that social choice theory ought not to be understood as undermining the possibility of coherence and therefore collective action. Indeed, that theory shows how it is that such coherence may be achieved.

Plans and Planning

We, says Michael Bratman, are planning agents. Bratman thinks the Aristotelian account of action – in which actions are chosen to satisfy desires in light of beliefs — is inadequate to describe how we deliberate about what to do. Rather, he thinks that we formulate intentions, and as part of forming an intention, we commit to a course of action or plan, or, even more usually, on a partially specified course of action: a partial plan. Thus, Bratman thinks not only that there is something that comes in between beliefs and desires on the one side, and action, on the other, but that this something can be described either as an intention or as a partial plan.19Bratman partly grounds the idea of planning on limited cognitive capacities. By adopting a partial plan early in the deliberative process, the complexity of the decision problem is reduced. And, because this simplification is valuable, an agent will have reason to hold onto a plan even if there is some evidence that the plan is suboptimal. Bratman recognizes other reasons for sticking with a plan rather than jettisoning it at the first sign that it is not the best possible course of action.

Thus, at some (early) point in practical reasoning we adopt or posit an intention to do something. We might intend, in one of Bratman’s examples, to present a paper at a professional meeting in Boston in October. In adopting this intention, we have already (at the same time) adopted a partial plan to arrive, somehow, at the meeting in Boston on the appointed date. This plan may be very incomplete: it may not for example involve a commitment, yet, to go by plane or train, to buy the appropriate tickets, to arrange for a trip to the airport or train station, etc. Still, the intention itself already involves a part of plan: to arrive in Boston on the appointed day, with a paper to present. As we go forward in time, assuming that we retain the intention, we are moved, somehow, to fill in the partial plan by planning on doing the required things to make the intended state of affairs come about.

So, Bratman sees planning as central to deliberating over what to do and we do it because it helps us to coordinate our activities. Partial plans represent commitments to courses of action and they provide reasons that enter into or choice of actions over time. Plans are relatively stable in the sense that we do not constantly reconsider them as our beliefs and temporal location changes; if we did so, plans would not help us coordinate our actions and intentions over time. And because they are relatively stable, plans help us to structure our further deliberations by permitting us also to consider a manageable set of options (train, plane, boat) to decide among.20 On the cognitive limitations view, failure to adopt plans would be a failure of at least second best rationality. It is not an option to act as though we could plan synoptically, taking every conceivable option into account, so a refusal to plan would tend to lead us to worse outcomes than those we could achieve if we took account of our limitations. They also make our future actions more predictable both to ourselves and to others. Knowing that I intend to be in Boston at the professional meeting, helps me to make other decisions about how to spend my time (writing the paper, etc), and permits others who may want to see me to form their own intentions. Plans help, in these ways, to coordinate or organize our deliberations over time. Importantly, plans exert normative or critical pressure on our actions and on our other intentions and actions: if I plan to go to Boston, there is reason for me to decide about mode of transport, to see to getting the tickets, etc. I can, of course, abandon my plan so this normative pressure is defeasible but as long as my plan remains in place, it (along with other considerations) generates reasons for action. I want to call this idea weak normativity to distinguish it from such things as obligations to others.21 The authority of our plans over our further deliberations takes the form of exclusionary reasons in Razs sense. That is, if I am considering taking an action and I have adopted a plan that excludes taking that action, I have a reason to take the plans constraint as a reason for not doing the action in question. I do not look behind the plan for the reasons that led to adopting the plan in the first place; rather, the plan itself provides reasons for action or restraint.

Plans are, in some respects, like recipes but they also have propositional content, in the sense that one plans to do something. Planning do something entails intending to act in a certain way. I can have a recipe for making bread, and follow its steps mechanically with no intention to make bread at all. I might simply find it calming to mix the ingredients and knead the dough and enjoy the smells that emanate from over at certain stages in the process. The bread would result as an unintended (but forseen) consequence of my following the steps in the recipe. But, if that is my orientation to the steps, I am not following a recipe for making bread. The recipe is directed to directed to making bread and it is this “directness” that is central. Recipes are (partial) plans in this respect. If I plan to make bread I will have to go through some series of actions that can be described as a recipe but now I am following the steps with a certain attitude in mind. Having a plan to do something also seems to entail having a kind of critical attitude toward my activities and to the plan itself. If I plan to make bread, I should not follow a recipe blindly, regardless of the circumstances but, if necessary, make adjustments for humidity and altitude and, perhaps, deficiencies in my oven. And, if conditions are sufficiently unfavorable I should decide to skip the bread making altogether.

In any case, it seems plausible that groups can have plans in very much the same way that individuals do; indeed, insofar as interpersonal coordination problems a more difficult than those faced by individuals, groups have stronger reasons to plan. But group plans must be connected to shared or group intentions. And there are some new twists that arise for groups. An important aspect of having a joint or shared intention is that such intentions, or at least their associated plans, are supposed to fit with the personal intentions or plans of the group members. This is not the same as Bratman’s meshing condition but is, rather, a normative requirement that the plans of a group and its members have to adjust to each other in some ways. If you and I intend that we paint the house, then as time goes on, at least one of us had better get some paint and brushes and so make room in our personal subplan to ensure that this takes place. Moreover, if part of our shared intention is that we paint the house the same color, then our personal plans (which correspond to intentions to) had better be made to mesh or agree so that this occurs. We might sensibly have a joint intention to paint the house yellow, or blue, but have no shared intention to paint it yellow and blue.

Bratman thinks that once we have formed a partial plan to paint the house, that plan then structures a further negotiation over a choice of color. He doesnt say much about how this process might proceed. I might simply buy some blue paint and get started and you, while preferring a yellow house to a blue one, still decide to go along with my choice rather than painting over my blue with yellow. Or the process could involve more explicit bargaining and agreement. The point is that joint intentions involve coherence constraints of two kinds. Joint plans are constrained in the same way that personal plans are (to be coherent, etc) and, additionally, to fit or mesh with personal plans of group members (and with other joint plans involving group members). How this coherence or harmonization is brought about is not specified.

This idea, that partial plans provide a basis for further negotiating, bargaining, and deliberation, seems especially important for large and heterogeneous groups. Must group intentions achieve full coherence (horizontal consistency) or must group and personal intentions fully mesh (vertical consistency)? It would seem to me that in heterogeneous groups there might be a sense in which coherence remains a goal or aspiration in the sense that members might willingly forgo full coherence or consistency in order to pursue other aims. For example, an ethnically diverse group might be willing to tolerate some degree of inconsistency in order to recognize values important to various subgroups. Or, to take a more pedestrian example: we might have a policy of subsidizing tobacco farmers while at the same time discouraging or even forbidding smoking. Such a policy recognizes both the value of a traditional agrarian lifestyle for the farmers whose land might be suitable only for tobacco cultivation, and a public concern for the malign effects and addictive qualities of tobacco consumption. Obviously such policy combinations are vulnerable to criticism from a coherence viewpoint. But it is not clear that it will always be best, all things considered, to forbid such policy choices.22 These examples are supposed to illustrate aspects of value pluralism circumstances of tragic choice that may prevent either groups or individuals from acting with full coherence.

A related issue arises when a policy can be explained or justified in terms of conflicting underlying principles. Here the policy itself, as a partial plan, might be seen as coherent when viewed in either in two ways: as pursuing goal A or goal B. But these goals are themselves inconsistent in the sense that each would fill in the partial plan in different ways. The policy can be seen as a part of a plan to pursue A, or as part of a plan to pursue B. The adherents of each these interpretations will argue that their principle gives the better explanation or justification of the policy and therefore, gives the better reasons for how the plan should be completed. As long as the underlying issue remains unresolved, the authority that the plan exerts on future actions will remain controversial.

What then is the force of coherence on shared intention? Bratmans house painters suggest two different kinds of force or pressure. First, if no-one undertakes to get the paint and brushes if no one forms the personal intentions to do her part — there may be a sense in which we do not have a joint intention that we paint the house. We each may wish that we paint the house but nothing is set in motion that leads to the implementation of this wish. Lets call that a constitutive constraint. A second source of constraint is perhaps more interesting in that it arises from the desires or values of the group members. We could, sensibly, paint the house both blue and yellow, but if each of us prefers either a blue or yellow house to a multiply colored one, there is a reason for us to bargain or negotiate over the color so that we can decide on a single color plan. We can call this a regulatory or normative constraint.

I dont think that it is necessary to assume that the direction of these normative forces is fixed. A normative constraint arising between a shared intention and the personal intention of a group member requires only that one or the other intention give way at least a little. Which intention is to be altered or abandoned is another question. If a joint intention requires, for example, that one of the group members act immorally, the pressure would probably be to reinterpret or abandon the joint intention.

Legislative Intentions

We return now to the main object of this paper: legislation. A legislature is a group authorized to produce coercive rules that regulate behavior. When we turn away from Bratman’s two-person problem and consider the circumstance of legislation we need to confront several new issues. Members of a legislature are not required to agree with each other, even after long argument, in order to decide what to do. A legislature may decide, or adopt a plan, even if only a bare majority of its members vote for it. Is there, in this case, a plausible sense in which we could say that the legislature itself has adopted a plan? Or, is the most we can assert is that the majority – those who have voted for the successful proposal – has adopted a partial plan and has, as well, a shared intention? If only this last claim is true, then must we say that the members in the minority may have no reason to see the statute as having any normative claims on them.

Bratman’s view of shared intentions – because it sees those intentions as identical to a complex of individual intentions – seems vulnerable to this worry. It is hard to see, specifically how anything like his “meshing” condition can be expected to hold in a way that involves the minorities intentions.23This may simply be a failure of my imagination: for a meshing condition to hold in a majoritarian body would seem to require that the interlocking individual mental states be a combinatorially complex disjunction of individual intentions, where each element of the disjunction is a legislative majority. And, if that condition fails for the legislature, it is hard to see how members of the minority have reason to see a statute as in any sense “their” plan. For this reason, I am inclined to reject the meshing condition, as a requirement for the formation of legislative intention and instead pursue Margaret Gilbert’s notion that shared intentions may be formed under weaker conditions; perhaps by the occurrence of certain formal actions, such as a procedurally legitimate vote to enact a statute, rather than requiring the existence of a complex set of individual intentions.

It is important to see that the close connection between plans and intentions is maintained for groups: when the legislature agrees to a statute, that statute constitutes a partial or incomplete plan of what the government is to do, and that partial plan corresponds with a shared intention. This shared intention is an intention the legislature itself, in virtue of its rule for making binding decisions, and not that of a mere majority. And as an intention of the legislature, the statute imposes normative pressures on all members whether or not they supported the statute initially. I do not mean to argue that any member of the legislature has reason simply to support any particular statute enacted by the legislature or to resist limiting or even overturning it. Under some conditions legislators are right to reject or radically alter previously enacted laws. Rather, I want to make weaker claims that arise from the arguments for coherence introduced previously: that the existence of conflicting patterns of legislative action – such as statutes mandating X and not X or, in the present case, a statute which cannot be given a coherent justification – gives legislators reasons to make adjustments to reduce or eliminate those conflicts. These reasons follow from the principle that people confronted with coercive commands are entitled to be shown how those commands are reasonable: specifically that they are the commands of an intentional agent. The adjustments required are to be a matter for legislative deliberation.

Legislative Deliberation

In this section I shall explore the deliberative implications of planning and specifically of a legislature taking the demands of coherence seriously. I do so by exploring a simple example, chosen to illustrate a circumstance in which the achievement of coherent legislation is costly to the members in an important sense. The members wish to present the legislature as having a coherent policy of the kind that a unified agent could rationally adopt, but achieving coherence entails giving up other things of value. It is meant to be a model of planning: the legislature is operating with an existing plan and is deciding whether to alter or perhaps to abolish it. The members disagree as to whether the existing plan is viable so at least some members are open to taking radical steps. But others disagree. The example is very simple: there are only three members in the legislature and a restricted number of issues and alternatives. It is, however, possible to extend the reasoning to more complex situations.24 For a more developed argument see John Ferejohn, “Conversibility and Deliberation”

In the example – a stylized picture of the current debates about social security – there is a conflict between the members acting as rational agents and group acting in that way. Such circumstances may seem quite special but they allow us to see if the planning perspective helps understand what kinds of reason are available for guiding legislators when such choices must be made. I chose the social security debate partly for its contemporary familiarity, partly for the recognition on all sides of the importance of taking a long term view, and partly because, leaving aside the eligibility issues, it is nearly a purely legislative program. Judges have had no role in setting benefit levels or tax rates or any other important aspects of the program. The Supreme Court was asked whether the program was constitution and they agreed that it was. Except for that, everything has been up to Congress. If coherence or integrity are to play any role at all, they must be applied legislatively rather than judicially.

Consider the following hypothetical pattern of judgments about social security reform.25 This arrangement of the debate into an issue structure is admittedly not “given” by the record but represents a simplified version of the deliberation. Member A thinks the current system is bankrupt but doubts whether private accounts will work.26 The private accounts proposal must be packaged with associated benefit reductions and tax increases. Such accounts may be said to work if they are fiscally feasible and if they provide “adequate” guarantees of financial security. In other words “work” in this context is interpretive in that it includes some notion of what such adequate guarantees would require. She may think that they will be too expensive or will result in unacceptable risks to some. Member B doubts that social security is in fact bankrupt, but is optimistic about private accounts, thinking them somehow a feasible way of providing adequate guarantees of financial security. And C thinks both that the current system is bankrupt and that the proposed new system is workable in the sense given in the previous sentence.27 Preference distributions of this kind have been called a “doctrinal paradox,” by Lewis Kornhauser and Larry Sager. Philip Pettit terms them instances of a “discursive dilemma.” I assume, for the sake of argument, that a member would favor adopting private accounts only if she accepted both premises, as C does.28 This assumption is certainly challengeable: a member could vote for the reform if she thinks it workable even while thinking that the current system is not bankrupt. I ignore that possibility here. I suppose one could claim that this exclusion could be justified on the basis of a kind of political rationality: that proposing to replace the current system that appears to be working well enough and on which people have relied is so politically costly that no representative would try it.

Table One:

SS “bankrupt”

Private accounts work

Adopt Reform

A

yes

no

no

B

no

yes

no

C

yes

yes

yes

How might deliberation proceed in the circumstance described in Table One? One possibility is that the members simply vote on whether or not to adopt the reform. Presumably, member C would vote for the reform, while neither A nor B would; both are skeptical about one of reform’s supporting premises. We call this way of deciding “results oriented” voting. Note that if the information in Table One were to be revealed during the debate, or if the legislature were actually to take public votes on the premises, the majority outcome (rejection of the reform) would be unsupported by majority views on each of the premises – there are majorities in favor of the proposition that social security is bankrupt and that private accounts are workable. Indeed, we can imagine another way of proceeding: the legislature could take votes on the premises and then, decide on the result in the way consistent with the vote on the premises. Philip Pettit calls this procedure “premise oriented” and notes that it has the virtue of picking an outcome that is supported by majority endorsed premises. Indeed, if its members were concerned to produce this kind of coherence, one could even imagine the legislature committing self by formal rules to doing business in the premise oriented way.29 Nothing in my argument depends on the preference based procedure but only the broader class of procedures that force the pattern of collective choices to be coherent: in this case, deciding on any two issues would determine the outcome of the third.

I preface the analysis with a few remarks. First, examples of this sort of “dilemma” depend on the use of a nonunanimous decision procedure such as majority rule of some version of it. Further, the only kind of dilemma that can arise is one in which the results vote is negative, while the premise based votes are all positive. One interpretation of this is that under results voting the “status quo” (no change in the program) is privileged relative to the premised based outcome. Third, the status quo “bias” would be even greater had the affirmation of more premises been required for an affirmative premise based vote. And finally, this implies that one way of resolving the dilemma — increasing the required majority for assent to the premises – would further increase the status quo bias.

Another feature of the pattern of judgments in Table One is that each member is “pivotal” to the pattern of outcomes: there is a way she could change her votes so that the pattern of outcomes would become coherent. If member A were convinced that social security is not bankrupt, the vote on that premise would then switch to negative, as would the collective judgment on the premise, which would support the negative outcome. Similarly, if either of the other members had a different view of one premise the dilemma would disappear. Thus, in this example, each member is “pivotal” to the dilemma. This suggests a sense in which each member can be thought of bearing some kind of responsibility for the overall outcome in the sense that if she could be persuaded to change her views on one issue, a coherent result would be produced.

I distinguish two models of deliberation: pure and impure. In pure deliberation, the group deliberates about all of its judgments in a symmetric way, giving no priority to judgments about one issue over another, nor to the final result over the issue. The aim is to produce a coherent and acceptable pattern of outcomes – to get things to fit together — and, in this process, each of the group’s judgments is on the table for reconsideration. This is a process of reflective equilibration. There may be substantive reasons why it is more difficult to adjust some judgments than others – some beliefs for example may be tethered closely to empirical evidence for example — but such considerations are invisible at the formal level. In any case, each of the judgments that must be made in these examples embody important normative judgments as well as factual ones. The judgment that the current system is bankrupt is offered as a reason to do something to correct or reform it. The judgment that private accounts are workable implies that such a system is financially viable and that it would constitute an adequate substitute for the current system of providing social security.

Impure deliberation proceeds sequentially, first fixing its decision on one issue first and then taking up further issues in sequence without reconsidering the initial decision. In Table One, the legislature might decide to fix its collective judgment, irrevocably, on the first question on whether social security is bankrupt, and then deliberate about private accounts without revisiting the previous decision. Proceeding in the impure way corresponds roughly to an “institutional” resolution of discursive dilemmas: the structure of the procedure guarantees a coherent pattern. It seems clear enough why such a deliberative procedure would be called impure. There is a sense in which there has been an arbitrary closing of debate on one issue, and a sense that the discursive dilemma has been short circuited rather than resolved.

Impure procedures may be adopted for reasons of practicality or perhaps for some other reason relating to the formation of group intention. Members of a group deciding over time may feel bound by precedent for example. Or, a group may parcel out its decision tasks to subgroups and each subgroup may feel bound to honor the prior determinations of other subgroups. The current example doesn’t exhibit this idea, but one could imagine an appellate court stipulating facts established at an earlier trial as an example of this sort. So, there is a variety of impure procedures and there are sometimes good practical reasons to adopt them.

Impure Deliberation

Suppose that the legislature proceeds by using a premise based procedure – first voting on bankruptcy issue, and then on the workability private accounts, and then determining the result in view of the premise judgments. But now, rather than taking the judgments in Table One as immutable, members are free to reconsider them in light of their effects on the overall outcome. Recall, as these judgments contain normative as well as empirical considerations, it makes some sense that members would be willing to reconsider their normative views in light of new evidence or in light of competing normative considerations such as coherence. How would we expect the members of the legislature to deliberate and vote in light of the sequential procedure specified above?

One way to proceed is to assume the member are sequentially rational in expressing their personal judgments, which means that at every point in which a member needs to make a decision, that decision must be best for her for that point on. Members are free to alter their judgments or results in light of earlier outcomes, so we need to specify the ways in which they may do this. For a start, let’s assume that the legislators are more committed to their views as to the final outcome – whether the private account system ought to be adopted — than they are to their judgments about the premises. In this case, the final decision (on the workability of private accounts) is to be taken after the issue of bankruptcy is resolved. And so each member would see that if the legislature has decided affirmatively on the bankruptcy issue, the overall outcome will turn on how the vote goes on whether private accounts are thought to be workable. On this issue, members B and C are pivotal in that if either is be persuaded to changer her views, the collective judgment on the premises would be negative. And as B thinks the reform is unwise, she has a strong reason to reconsider her judgment on this issue. Of course, had the initial vote (on bankruptcy) been negative no-one would be pivotal to the outcome and one would not expect much pressure to reassess views on the issue of private accounts.

So, if the legislature adopts the sequential premise based procedure and its members are result oriented and sequentially rational in the way assumed here, the outcome would be to reject the reform and preserve the current system, and there would be a coherent and public explanation for that result. Note that had the votes occurred in the other order, the result would have been the same in this respect though the voting pattern – and therefore the public justification for it – would have been different.

In this example there is no need to consider the question of coherence independently of the decision itself – the sequential premise based procedure itself guarantees that the pattern of votes will cohere. But, the outcome is the opposite of what Pettit hypothesizes under premise-based voting. This is not surprising in that we assumed that members guide their voting on premises in order to try to reach final outcome each thinks most justified. Had they been more confident in their views about the premises than they were about the final outcome, perhaps none would have been willing to change her views on premises and the reform would be adopted.

While I don’t have space to worry the argument out in detail, there is a theorem here that would have a statement roughly like this: in the conjunctive case, if the pattern of initial judgments produce a discursive dilemma, sequentially rational results oriented players (in the sense given above) will always produce premise judgments supporting a negative result, irrespective of the order of voting. No discursive dilemma will be observed except possibly at early stages of the deliberative process. The argument would go as follows: assume that there are k premises and n voters, that the underlying or prior judgments form a discursive dilemma, and that the premises are voted on first. Then, at a final node following a sequence of positive premise votes, for each person that prefers a negative result voting against accepting the last premise weakly dominates voting for it and, by assumption, this set of people forms a majority.30 A strategy, x, weakly dominates another strategy, y, for a player, if the result of playing x always leaves the player at least as well off as playing y, and better off in some circumstance. Thus the outcome is negative and there is a coherent justification for it.31 I think something like this could be established for indefinite procedures of the following kind: voting terminates whenever a result and coherent pattern of premise votes is reached. Otherwise, it continues in a fixed order. The resulting extensive form is infinite but has a simple markovian structure. As long as the voters are (at least a little) impatient, I believe that there will be an equilibrium in which the game terminates after one round of voting, that it will not “reveal” a discursive dilemma, and that in conjunctive case the result will be negative. If the result itself was voted as one of the first k votes, it would fail as there is no premise based reason not to vote initial outcome preferences in the first k votes. And on every one of the k-1 premise votes the judges are free to vote their initial judgments. Only the last premise judgment is to be implicitly fixed in a way that supports previous votes. What if we assume that the members are premise rather than result oriented, in the sense of being more confident in their premise judgments than their views about the outcome? A parallel argument suggests that members should not alter their initial premise judgments and result will be that the reform would be adopted.

It may seem that a particular idea of sequential rationality is playing too determinative a role in these examples. Sequential rationality makes the choice of voting order inordinately important. And, in any case, the deliberative practices described here may be troubling if one thinks that the members are voting deceptively: concealing their own best judgments (on the premises) in order to produce an appearance of coherence. Perhaps this kind of voting is unacceptable in itself even if the legislator has no invidious reason for voting in this way. Perhaps that is reason to adopt an alternative maxim: that legislators ought to stick with their initial premise judgments, and adopt the result that is required to vindicate those judgments. But why should they privilege their initial judgments in this way, independently of how deliberations proceed? Perhaps it is better simply to abandon fixed sequences and try to see how pure deliberation might proceed.

Pure Deliberation

In pure deliberation all the issues – premise judgments as well as the final result – need to be decided together. I imagine that the legislature would, in effect, take repeated straw votes on each of the issues and that individual legislators would be invited to reconsider their judgments in light of the pattern of collective judgments. The pure procedure seems less institutionally constrained than impure procedures in that each of the issues are treated symmetrically. The principle of pivotality, introduced earlier, would help each legislator to see the consequences of changing her vote in view of how others are voting and would thereby help to focus deliberation. As before I assume that each member places great value on producing a coherent pattern of results as well as on getting a good result — they want their premise judgments to support their final decision — and that they will see a failure of coherence as a reason to continue deliberating.

Is there anything objectionable about the idea of members seeing all of their judgments, on premises as well as on results, as being in play during the deliberative process? Each of these judgments are normative judgments as to what the group ought to do – they are not private regarding in the way that preferences are thought to be — and, from a formal perspective, they are more or less symmetrical. And insofar as normative judgments conflict, there is reason to be open to reconsideration. Moreover, as in the example, judgments may have empirical contents too: legislators may learn new information during deliberation that may cause them to reassess their own empirical views. And beyond these considerations, there may be substantive reasons to insist that one judgment has priority over the others. One of them might, for example, be dictated by considerations of morality. These are substantive issues, and perhaps controversial ones, that would themselves be the subject of deliberation.

We can start by partitioning the set, X, of outcomes into the following subsets: Let YC be the set of voting patterns that would produce a yes on the result and a coherent set of premise judgments, NC is a negative result with coherent preference judgments, YI a yes vote with an incoherent set of premise judgments, and NI a no vote with incoherent judgments. So NI is the set of initial discursive dilemma preferences. We have already remarked that in the case described in Table One, YI is empty so we need consider only three subsets. We may suppose that the members have initial rankings of the elements in X, and that these rankings may be revised or adjusted in deliberation.

We shall say that an outcome xεX is deliberatively unstable if someone may alter her judgments in such a way as to achieve a preferred outcome or a coherent voting pattern supporting the same outcome. The first thing to notice is that the pattern of initial judgments in Table One is deliberatively unstable. Member A could alter her judgment on bankruptcy from Y to N and thereby produce a coherent pattern of judgments in support of the negative result. Or, she could change her views about private accounts from N to Y (and therefore her judgment on the final result from N to Y) to produce a coherent judgment pattern in support of a positive result.32 Indeed, this seems to be a general result for a three person group with any number of premises: every judgment pattern in NI is unstable. The argument would go like this: for any x ε NI, there is at least one individual who is voting Y on a premise and N on the outcome and whose vote is pivotal on the premise. If she changes from Y to N on the premise the resulting pattern is in NC which is the better result for her. But the case of three voters special in that for any profile in NI, there is always at least one player who is pivotal. If each of the rows in Table One were to represent three judges, none would be pivotal and the pattern in Table Two would be stable. For this reason it makes sense to require each judge to express her judgment on a question on the assumption that she is pivotal to deciding it. The presumption of pivotality is, in this setting, a device for focusing responsibility and organizing deliberation. It does not say what to do, but it says that individuals, judges in this case, ought to formulate their votes on the assumption that it makes a difference how they do so. And, in particular it makes a different to how the court is able to present itself to those subject to its decisions.

It is easy to see that there will always be many deliberatively stable profiles and so, in a sense, the legislators must solve a “coordination” problem in order to settle on one rather than another. Each of the different stable profiles reflect a different way in which the members may adjust their views on the premises and outcomes. And, as the investigation here remains at a formal or nonsubstantive level, there seems no natural way to narrow this set of possible deliberative results. In specific cases there may well be substantive reasons for the legislature to choose one deliberatively stable profile over another. And, we have said nothing about the dynamics of the deliberative process that might lead to one particular outcome over others.

The point of thinking through these models of deliberation is to illustrate how members in a legislative body might try to take considerations of coherence into account. I claimed that they would be and ought to be disposed to reconsider their initial judgments in light of the consequences of exercising these judgments. If they do that, depending other assumptions, legislators will sometimes have reason to modify their “expressed” judgments in the course of deliberation. I don’t have the space to investigate the general effects of such deliberations but it seems likely that its effects would be to diminish the heterogeneity of expressed judgments relative to initial judgments. In the language of social choice theory, deliberation produces a restriction on domain of preference configurations.

More to the point, deliberation, if it proceeds in the way outlined here, will tend to produce a coherent plan and therefore something like a legislative intention. It is impossible to say in advance what partial plan will emerge from a more or less pure deliberative process. There are generally many deliberatively stable voting profiles and which emerges depends on how competing values are weighed. In this simple example, some profiles support continuing the existing social security program even if a majority thinks it is unsustainable over time because of bankruptcy. This is at best a partial or incomplete plan: there is reason for someone to come up with a proposal that could ameliorate the fiscal problem by raising taxes or reducing benefits. And the members would be expected to entertain such a proposal sympathetically in view of the incompleteness of the partial plan they have adopted. Indeed, it is the incompleteness of the partial plan that gives the legislature, and its members, reasons to continue.

Discussion

The planning perspective is a democratic conception of legislation in two senses. First, as in traditional theories of representation, legislators receive their rule making authority in democratic elections and are obliged to represent the peoples’ interests. Second, making rules that bind others gives rise to a normative obligation to show that those rules further valid social purposes and infringe on peoples’ rights only in ways that are necessary to accomplish those purposes. And this explanatory duty is owed to everyone equally.

These two sets of normative obligations, or at least some interpretations of them, may come into conflict: a narrow understanding of the duty to represent might direct legislators to pursue only those views expressed by their constituents and be unwilling to compromise in this pursuit. But, such an uncompromising posture will require that legislation is produced by aggregation and this will often lead to incoherent legislation for which there is no coherent public account. This is a reason to reject such a narrow understanding of representation and to say, instead, that the legislator’s duty is to integrate the interests of his constituents with those of the members of the polity generally. This requires that legislation be directed to further some conception of common purpose so that each person (inside or outside the constituency) can be given an explanation for laws that she can, in principle, embrace as her own.

Therefore, the planning view of legislation sees statutes as rationally directed to furthering common purposes or legislative intentions. These purposes remain, in most cases, matters of continuing conflict and dispute. The planning view regards a statute as a partial and incomplete expression of those intentions – partly because their content is not completely agreed upon – and sees the statute as needing of further development and specification, as applied to specific circumstances. It presumes that individual legislators regard statutes as imposing further normative pressures on their personal actions and intentions – reasons to take or restrain their intentions and actions — as well as on the actions and intentions of many outside the legislature. And, there is pressure in the other direction as well – pressure to refine or redirect the legislative intention. Central to these normative pressures is the notion that those involved in with the statute ought to seek to present it as rationally directed to further a shared intention or policy of the statute and as coherent or exhibiting integrity. These normative pressures will normally be manifested in other legislative deliberations – certainly in deciding how much funding ought to be provided and in considering amendments that may be proposed in light of experience – but also in administrative and judicial proceedings as well. I think that they will also be articulated in elections when candidates compete for positions in the legislature on the basis of pledges to pursue certain policies or to alter or reject them.

In these ways, the planning perspective illustrates that legislative intentions are subject to contestation and argument. This is as one would expect. There are many ways to fill in or extend a partial plan and people can be expected to fight over how these steps should be taken. Moreover, even popular legislative programs are opposed by some who would prefer to repeal the statute rather than extend it. But unless they can find ways to do that that can win majority support from people who have already made some commitments to the statutory scheme, opponents may need to find ways to work within the legislative regime to divert and reshape legislative intentions gradually.

Finally, it is important to remark that as important as coherence is, it is one value to be weighed and considered against others. Sometimes, full coherence or integrity is not worth achieving because it would entail sacrificing other values of greater importance. This is true for individuals too: the writings on tragic choice or incommensurability are relevant in this connection. No one thinks that such examples undermine the possibility of human integrity or agency; indeed some might argue that such cases provide particularly good examples of what integrated moral agency actually requires. So, I am not sure that failures of complete coherence undercut group agency claims at least if the failures are of the right kind.

John A. Ferejohn est professeur de droit et de science politique à New York University. Il étudie en particulier le développement de la théorie politique positive et notamment son application à l'étude des institutions et des comportements juridiques et politiques.

Notes

Notes
1 I owe special thanks to Lewis Kornhauser for extensive comments on this paper, not all of which I have been able to take account of, and to Seana Schiffrin for many comments on an earlier version of it. If you give me comments, you too may have a place in this footnote. These people are blameless for all the faults that remain. Early versions of some of these ideas have been presented at the Dworkin-Nagel Colloquium (I think in 2001), at a conference at the Australian National University in 2004, and at a lecture at the University of Minnesota Law School in 2005. I am going to keep at it til I get it right.
2 What counts as a valid or legitimate purpose is a complex topic that is not addressed here beyond saying that validity would include both procedural elements – that the statute was enacted according to established and fair procedures – and substantive ones as well (including constitutional and possibly other criteria as well).
3 This is a complex question that is resolved in various ways in different legal systems.
4 The text of the statute, which was produced by that legislature, is another matter. Any theory of legislation – normative or positive needs to take account of text as both something that is fought over in the legislature and as having constraining effects on what the law can be subsequently.
5 One could also imagine an instrumentalist aggregative view. Such a view may have scientific value I suppose in permitting the attribution of intentions to members or coalitions.
6 Formal precedents that govern how a legislature conducts its business are often important of course. The procedural rules of both the United States House and Senate are often found in precedents.
7 In the areas of taxes and appropriations Congress frequently sees itself as bound by its previous commitments. This is also the case with government contracts. Legislative “entitlements” represent, perhaps, a weaker sense of bindingness since Congress does sometimes adjust entitlements. For a detailed treatment of this issue see Eric Posner and Adrian Vermeule, Legislative Entrenchment: A Reappraisal, Yale Law Journal , volume 111 (May 2002), 1665-1705.
8 If you intend to drive to the store from your house, that intention gives you reasons to get in the car, back out of your driveway, turn left or right at various intersections. You could refused to do these things but then the chances are you will not fulfill your intention.
9 This is not to minimize the role of judges or agencies in construing statutes but only insists that they ought to see a statute as intentionally directed and try to construct or recover the relevant intentions when applying the legislation.
10 While the notion of the Court, as a multimember body, acting as an intentional agent may be familiar, it is no less mysterious, in the terms of this paper, than the idea of a legislature having an intent.
11 Kornhauser and Sager, in a recent article urge caution in comparing courts and legislatures in respects under discussion in this paper. Lewis Kornhauser and Laurence Sager, “The One and the Many: Integrity and Group Choice in Paradoxical Cases, “ Philosophy and Public Affairs, vol 32 (2004) 249-276.
12 Michael Bratman, Intentions, Plans and Practical Reasoning, Cambridge: Harvard University Press, 1987.
13 “Shared intentions are intentions of the group….. what they consist in is a public, interlocking web of intentions of the individuals.” (Bratman, Faces of Intention, Cambridge: Cambridge University Press, 1999, p.43. Thus, as with the social choice theory account, Bratmans theory reduces group intentions to individual intentions but the kind of reduction is not aggregative in the sense of social choice theory at least not obviously so — but constitutive.
14 Whether individual attitudes are rightly thought to be simpler or more primitive is a controversial matter.
15 Shared intention, as I understand it, is not an attitude in any mind. It is not an attitude in the mind of some fused agent, for there is no such mind; and it is not an attitude in the mind of either or both participants. Rather, it is a state of affairs that consists primarily in attitudes … of the participants and interrelations between those attitudes. (Bratman, Faces of Intention, 122-3.
16 What my part and your part amount to is a complex issue. Our parts may be fully specified in the partial plan in some cases. In others there may be some background idea of fairness that would play a role in determining what our individual obligations are. I owe this observation to Seana Shiffrin. Liam Murphys work is relevant here.
17 One of Gilberts criteria for identifying a joint intention is the existence of obligations (or at least a sense of obligation) to pursue and help the others in the group to pursue the joint intention. These obligations have something of the same force as the obligation to keep promises. Even so, if such obligations conflict with individual intentions or other commitments, such obligations might be overridden, even at the price of having to suffer rebuke from other group members.
18 For an exploration of these issues see Philip Pettit, ‘Groups with Minds of their Own’ in Frederick Schmitt, ed., Socializing Metaphysics, Rowan and Littlefield, pp. 167-93.
19 Bratman partly grounds the idea of planning on limited cognitive capacities. By adopting a partial plan early in the deliberative process, the complexity of the decision problem is reduced. And, because this simplification is valuable, an agent will have reason to hold onto a plan even if there is some evidence that the plan is suboptimal. Bratman recognizes other reasons for sticking with a plan rather than jettisoning it at the first sign that it is not the best possible course of action.
20 On the cognitive limitations view, failure to adopt plans would be a failure of at least second best rationality. It is not an option to act as though we could plan synoptically, taking every conceivable option into account, so a refusal to plan would tend to lead us to worse outcomes than those we could achieve if we took account of our limitations.
21 The authority of our plans over our further deliberations takes the form of exclusionary reasons in Razs sense. That is, if I am considering taking an action and I have adopted a plan that excludes taking that action, I have a reason to take the plans constraint as a reason for not doing the action in question. I do not look behind the plan for the reasons that led to adopting the plan in the first place; rather, the plan itself provides reasons for action or restraint.
22 These examples are supposed to illustrate aspects of value pluralism circumstances of tragic choice that may prevent either groups or individuals from acting with full coherence.
23 This may simply be a failure of my imagination: for a meshing condition to hold in a majoritarian body would seem to require that the interlocking individual mental states be a combinatorially complex disjunction of individual intentions, where each element of the disjunction is a legislative majority.
24 For a more developed argument see John Ferejohn, “Conversibility and Deliberation”
25 This arrangement of the debate into an issue structure is admittedly not “given” by the record but represents a simplified version of the deliberation.
26 The private accounts proposal must be packaged with associated benefit reductions and tax increases. Such accounts may be said to work if they are fiscally feasible and if they provide “adequate” guarantees of financial security. In other words “work” in this context is interpretive in that it includes some notion of what such adequate guarantees would require.
27 Preference distributions of this kind have been called a “doctrinal paradox,” by Lewis Kornhauser and Larry Sager. Philip Pettit terms them instances of a “discursive dilemma.”
28 This assumption is certainly challengeable: a member could vote for the reform if she thinks it workable even while thinking that the current system is not bankrupt. I ignore that possibility here. I suppose one could claim that this exclusion could be justified on the basis of a kind of political rationality: that proposing to replace the current system that appears to be working well enough and on which people have relied is so politically costly that no representative would try it.
29 Nothing in my argument depends on the preference based procedure but only the broader class of procedures that force the pattern of collective choices to be coherent: in this case, deciding on any two issues would determine the outcome of the third.
30 A strategy, x, weakly dominates another strategy, y, for a player, if the result of playing x always leaves the player at least as well off as playing y, and better off in some circumstance.
31 I think something like this could be established for indefinite procedures of the following kind: voting terminates whenever a result and coherent pattern of premise votes is reached. Otherwise, it continues in a fixed order. The resulting extensive form is infinite but has a simple markovian structure. As long as the voters are (at least a little) impatient, I believe that there will be an equilibrium in which the game terminates after one round of voting, that it will not “reveal” a discursive dilemma, and that in conjunctive case the result will be negative.
32 Indeed, this seems to be a general result for a three person group with any number of premises: every judgment pattern in NI is unstable. The argument would go like this: for any x ε NI, there is at least one individual who is voting Y on a premise and N on the outcome and whose vote is pivotal on the premise. If she changes from Y to N on the premise the resulting pattern is in NC which is the better result for her. But the case of three voters special in that for any profile in NI, there is always at least one player who is pivotal. If each of the rows in Table One were to represent three judges, none would be pivotal and the pattern in Table Two would be stable. For this reason it makes sense to require each judge to express her judgment on a question on the assumption that she is pivotal to deciding it. The presumption of pivotality is, in this setting, a device for focusing responsibility and organizing deliberation. It does not say what to do, but it says that individuals, judges in this case, ought to formulate their votes on the assumption that it makes a difference how they do so. And, in particular it makes a different to how the court is able to present itself to those subject to its decisions.