THE TRACKING ASSUMPTION

Empirical and normative theorists generally treat responsiveness as a core desideratum. For a democracy to earn its name, surely there must be some connection between the inputs—whether preferences, well-being, or beliefs of an electorate—and the policy outputs (Barry Reference Barry and Christiano2003). There is a diversity of ways that a property can be tracked, from the raw preferences to laundered interests of citizens (Goodin Reference Goodin1995). If tracking constitutes democratic institutions, the alleged tension with leadership comes into focus. Let's see how two classes of theories incorporate this assumption.

Delegation Theories. Say lawmakers conceive themselves as responding to the discoverable preferences of constituents. Terms of office are short, opinion polling is rife, and citizens have ready access to democratic mechanisms like referendum and recall. What emerges from this is a familiar agency relationship.Footnote ² Representative institutions attempt to simulate assembly democracy as closely as possible. Pioneering work on representation searched for overlap between the opinions of a constituency and the floor votes of lawmakers (Miller and Stokes Reference Miller and Stokes1963). It was natural to treat democracy, at its base, as the “continuing responsiveness of government to the preferences of its citizens” (Dahl Reference Dahl1971). Recent work on democracy and unequal representation works from this assumption (Bartels Reference Bartels2008; Gilens Reference Gilens2013). Hannah Pitkin first noticed that descriptive representation, in its purest form, leaves “no room within such a concept of political representation for leadership, initiative, or creative action.”Footnote ³ We can follow this point to its natural conclusion. Every instance of a citizen leading a set of citizens is probative evidence of the unequal exercise of power. Leadership starts to look like an incorrigible threat to democracy.

To avoid this result, let's shift the relevant time slice of delegation theories. Suppose lawmakers respond not to existing preferences of constituents, which may not exist in a usable form, but to those preferences that they can be expected to hold at the next election (Arnold Reference Arnold1990). This account relies upon a simple rational model. Politicians maximize their chances of reelection. But there is an ambiguity in this inferred strategy, since the idea of “anticipating future preferences” admits of two readings. For lawmakers who have no appreciable impact on an inattentive public, the task is predictive. Tailoring positions to track future constituent preferences remains a passive activity. But insofar as lawmakers can themselves play a role in fashioning the content of their constituents’ anticipated preferences, the theory leaves open the possibility of pushback. Is a legislator who significantly alters the future preferences of constituents merely “anticipating” them? We are left with representatives who look more like coach drivers than agenda setters.

Trust Theories. If the simple agency model leaves no room for leadership, it's natural to turn to accounts of democracy that give representatives greater discretion. Trust theories loosen the leash. They give lawmakers license to veer from the discoverable preferences of a constituency. Their unit of representation is best put as a counterfactual: “what the public would think, if it had a more adequate chance to think about the questions at issue” (Fishkin Reference Fishkin1991). How much time and cognitive energy is “adequate” to render a verdict on the most complicated public policy questions that modern democracies face? Notice that this imagined world can stray far from actual politics. Lawmakers can vote in ways that they believe their constituency would have wanted were they fundamentally different people, who make hefty investments of time and resources in informing themselves about public life.

It is tempting to conclude that trust theories can secure a stable home for leadership. Theorists have tended to reason in this way: “The ‘trustee’ advocates low responsiveness and decisive, independent leadership, while the ‘delegate’ paradigm champions minimal leadership and strong responsiveness” (Jacobs and Shapiro Reference Jacobs and Shapiro2000). The pairing is superficially appealing. It recasts a familiar taxonomy, porting the delegate/trustee distinction into a binary between responsive nonleaders and independent leaders. But the argument rests on a disanalogy. Its mistake is to unpack the trust relationship with insufficient care. Think of the quotidian trust-based relationships that we avail ourselves of each day. My will establishes a trust for my toddler. It instructs the designated trustee to attend to a minor's well-being. The same structure applies to the blind trust that the lawmaker establishes. Her financial consultant will serve as a discrete trustee, managing her assets while avoiding conflicts of interest. There is an intuitive difference between the activities of the benefactor—whether court-appointed or personally hired—and a leader. Trust-based benefactors can be licensed to act without our knowledge and without attempts to change our beliefs or desires. We should resist construing trustees as “independent leaders.” The formulation itself is revealing. It counts activities that aren't in any way jointly performed.

But democratic leadership cannot exhibit literal independence from the electorate. Missing is the relational structure that we intuitively think of as part of this activity. Leadership needs a treatment group. It can't be performed in the isolation of a clean room. Consider the lawmaker who takes himself as constitutively unresponsive to constituents. John Stuart Mill famously conditioned his intention to run for Parliament. He would remain a free agent, unattached to “local interests” and unwilling to waver on his political convictions (Thompson Reference Thompson1976, Reference Thompson, Urbinati and Zakaras2007). While he participated in lawmaking with moral seriousness, Mill devoted comparatively little time to identifying and engaging with the discoverable views of his constituents. At the same time, Mill engaged in occasional projects of leadership. His calls for women's suffrage secured a surprisingly large, if insufficient, number of votes in the House of Commons. For the purposes of argument, consider a more severely independent version of Mill. He squirrels away from politics, actively avoiding contact with constituents, who may pressure him to act against his best judgment. This introverted form of lawmaking may earn him a “profile in courage” for breaking with strongly held convictions of a constituency. But there's no attempt to alter the intentional structure of his constituents.

We have auditioned two modes of responsiveness. Both treat episodes of leadership as counterdemocratic. This follows from their shared view of democracy as reducible to a tracking relation. Delegate theories granted equal authority to the expressed preferences of the electorate. They give weight to the desires of individual citizens, regardless of their content. This bottom-up perspective has lawmakers track preferences as they are, not as they might be. Limiting democracy to responding to identifiable desires can seem artificial, even mechanical. In Charles Beitz's apt metaphor, it treats democracy as “a crude hydraulic device” (Beitz Reference Beitz1989). The same assumption runs through trust theories. Their metric is the aggregated interests of the constituency, as discerned by a well-meaning benefactor. From this top-down point of view, elites respond to a passive electorate by asking counterfactual questions: what would this constituency think were they better informed and politically engaged? But this tracking relation can't register our basic convictions about democracy's value. Occupying colonizers may be best positioned to track our informed interests. Their epistemic interventions—constant correction of our stated desires—are good for us, after all. Burkeans are tempted to embrace the picture of the lawmaker as an enlightened benefactor.Footnote ⁴ Legislators track the counterfactual judgments that a more deliberative electorate would render. No room is left for a two-way relationship, where representatives alter the intentional structure of constituents. Trust theories leave citizens wholly “untreated” by officials who act in their name.

The conclusion of this section is limited. Democratic leadership seems impossible only if we accept a nested premise: that democracy is in the business of tracking citizen's preferences—whether as they are or as they might be. To arrive here, we’ve made use of the dilemma first identified by the philosopher-political scientist Arthur Bentley (Bentley Reference Bentley1908; Hardin Reference Hardin2004). If representatives are tethered to our preferences, they aren't meaningfully agents. But if they respond to an idealized version of ourselves, we aren't agents. To wrap up the argument in Lincoln's triad, the delegate does something by the people, the trustee does something for them, but the leader does something to them, and as I will argue, with them.

A FUNCTIONALIST ACCOUNT

Leadership isn't an accordion concept. While in everyday politics it is treated with notorious imprecision, its rough boundaries are clear. An agent gets another set of agents to do or believe something without overt coercion. This generic picture won't do if we want a conception of leadership that can survive in a democratic habitat. So let us try to understand the activity's function within actual political practice. What is leadership for in a system where decision-making power is shared among citizens? This methodological choice is significant, since it departs from a linguistic approach. We needn't canvass political discourse for word frequency. That would produce a positive theory. It would abscond from any appeal to normative considerations. In contrast, our strategy starts by picking out characteristic activities constitutive of leadership within democratic institutions. To fix ideas, consider two pictures of the concept's distinctive political role:

These remarks strike a significant overlap—the first from a political biographer, the second a political philosopher. It is not enough for leaders to select among a menu of prefabricated goals. Often they have to create a plan from materials found in a group of individuals. If the activity is, at its core, shared, what precisely is shared, and by whom? Start with a toy example. You are walking along Massachusetts Avenue with your friend. A stranger is also walking nearby—moving in the same direction and at the same pace. We want to say that only two of you are walking together (Gilbert Reference Gilbert2006). What is distinctive about your mental contents that make this the case? You and your friend share an intention to walk together.Footnote ⁵ The structure of your intention might look something like this:

How can we come to hold this joint intention? In the smallest scale cases, there is nothing alien about the idea. We may have decided to form this intention after a moment of eye contact, or a quick glance at our watches. But if the intention is genuinely shared, it's not clear how this is possible. How could I take myself as settling that we will go for a walk? Wasn't this also up to you? But who exactly settled it then. The problem seems to arise in the simplest actions: “How can I frame the intention that ‘we’ are going to act, if I simultaneously regard the matter as being partly up to you?” (Velleman Reference Velleman1997). Adding numbers to small-scale cases of joint agency can make the problem seem more intractable. How can a large group of persons come to share an intention to do anything? Without face-to-face communication, what would it mean for hundreds or even millions of people to arrive at a shared commitment to a political aim? It is hard to see how I can intend that we revise our political institutions. Assurance worries compound the problem. If the relevant “we” is a powerful enough subset of the population to alter our political institutions, how can I frame an intention about what we do, since my power over this collection is seriously diluted? These concerns have considerable force. We are looking for an account of shared activity that can simultaneously solve the problem in micro and macro cases—from the joint walk to the directed petition.

Let us vet two characteristic functions of democratic leadership. The first condition marks out the activity's product, while the second describes its mode:

These two conditions, along with the side constraints of the next section, make up the commitment theory of democratic leadership. Intentions are “stickier” than beliefs or desires, and commitments pick out a familiar and comparatively weighty genus of intentions (Calhoun Reference Calhoun2009). We use them every day in promising, vow-making, and the giving of assurances. They serve as relatively fixed points in our reasoning about what to do. Their persistence safeguards us against well-known psychological limitations.Footnote ⁶ The theory's core element is the creation and sustenance of joint intentional activity among followers. It isn't strictly incompatible with trait-based approaches, unless they insist upon the theoretical primacy of character traits. If anything, the theory offers a unifying explanation of the character traits of democratic leaders. Still, in my view, cataloging traits is a secondary task. What is fundamental is the interdependency between leaders and followers. The objects of leadership, call them F, come to share a commitment to political action, J. Put this way, leadership cannot be abstracted from the mental lives of followers. By working out the theory from the point of view of the democratic citizen, rather than the leader, we gain explanatory power. It's not enough for them to hope that political action J happens in an agent-neutral way. When each of these conditions is satisfied, followers hold a commitment that “that we J”—with J as a stand-in variable for some political action. Their shared aim makes essential reference to them as a group. On my theory, then, we can only explain the activity in agent-relative terms—it is about what treated citizens commit themselves to do.

Leadership's characteristic product has a first-person plural aim: that we will do something together. Consider Lincoln's recasting of the moral purpose of the Civil War. To put his effort in terms of our theory, he managed to reset the shared commitments of a substantial portion of the electorate, including a remarkable number of Union soldiers, who came to see the war as in the service of equality as much as union (Wills Reference Wills1992).Footnote ⁷ Despite Lincoln's claim “not to have controlled events, but confess plainly that they controlled [him],” he avidly gauged public opinion. In 1962 he received a letter from Carl Schurz, which stressed the relationship between public opinion and leadership: “Your personal influence upon public opinion is immense; you are perhaps not aware of the whole extent of your moral power” (Foner Reference Foner2010). In reinterpreting the meaning of the war efforts, Lincoln's achievement wasn't epistemic, but practical. He didn't just change the beliefs of citizens about the status of slavery, but their practical commitments towards ending it.

Theoretical reasoning alone isn't sufficient. Suppose a lawmaker holds a town meeting with her constituents on the science of anthropogenic climate change. She knows from her internal polling that there is considerable skepticism in this group. In her address she only presents arguments about climate science. She doesn't propose any joint cooperative effort. No appeal is made to seeking carbon-offsets or funding alternative energy. Suppose that her presentation succeeds in changing many of her constituent's beliefs about climate change. No matter how influential her words are, on our theory, this is not an instance of democratic leadership. Educative efforts needn't be wholly detached from recognizable leadership. We can leave open the possibility that future commitment-framing and setting will turn these changes into belief. Epistemic improvement isn't sufficient, unless it is accompanied by a change in the practical attitudes in prospective followers.

The Commitment Theory avoids the tracking assumption in two ways. First, it departs from the view that leadership is constituted by unresponsiveness, or more strongly, counter-representation. Acting against the grain of public opinion is neither necessary nor sufficient to mark out the practice. A representative that incurs risks by voting in a countermajoritarian way isn't leading any set of agents. There is a form of risk-bearing that my approach can register. At least some of the better-known legislative “profiles in courage” are, on my view, genuine instances of leadership. What unifies this subset of leaders isn't their self-sacrifice, or their willingness to act against popular opinion. It lies in the commitments they come to share with constituents. Second, the Commitment-Setting Condition resists attempts to reduce leadership to a special form of responsiveness. Leaders can act creatively in drawing consequences from preexisting beliefs of their targeted audiences or showing how ends demand particular means. They can still draw on source material from the beliefs and desires of the electorate. Indeed, the mental contents of one's constituents can play a necessary role in intention-setting in the first-personal plural. This explains the characteristic product of the activity: a commitment that we J.

We can now see two joint activities that the tracking assumption obscured. The first, brought out by the Commitment-Setting Condition, is generative. To get a grip, it relies upon a modest form of value pluralism in the political domain. We should not assume that any set of preferences is susceptible to harmonious arrangement. The electorates’ preferences may be so inchoate that there isn't enough material to attempt any kind of coherence. There's no assumption that the citizenry's preferences or judgments can be reconciled. If disagreement is a permanent feature of political life, this opens up the need for leaders who make up their mind—and if successful, help make up some nontrivial constituency's mind—about what goals are worthy of pursuit. On my model, citizens can share a political commitment without accepting each other's particular reasons for participating as followers. In the simplest kind of cases, we can build a cabinet together without co-holding each other's values and goals all the way down. It's enough that we jointly share a plan to act in some way.

Representatives act for groups that are constitutively disorganized. The second joint activity that leadership makes possible is coordinative. Suppose that a polity has not just perfect obligations, but imperfect ones as well. This gives it considerable discretion in the time, place and matter of discharging this genus of obligation. We may, for instance, be collectively obligated to fund the arts in a way that honors equality of opportunity. But if this obligation exists, surely it is best seen as imperfect in character. No theory of distributive justice, however regimented, will issue fine-grained instructions at this level. There will be plenty of elbow room in how we can satisfy an obligation of this discretionary character. A related coordinative function is a response to the indeterminacy of political values. Even if an electorate had clearly discoverable policy commitments, the decision of what policy to pursue may be seriously underdetermined. This holds true however precisely ranked they might be.

Suppose an electorate's preferences are genuinely torn between two aims: the development of a space program to Mars and a program to make the arts a ubiquitous presence in public life. Suppose further that neither goal is more desirable from an impersonal point of view. If there is no way of appealing to political values of distributive justice or legitimate rule, it is rationally deadlocked. No further inspection into rival theories will inform this decision. What is needed is not another round of philosophical debate, but a leader who selects one of these goals—breaking the standoff between two policies. The commitment-setting condition not only explains this activity, but it provides an account of its valuable function. The analogy here is with Buridan's mule. It starves when equidistant between two bales of hay, since it can find no rational way to choose among them. Our theory predicts that an electorate that shuns leadership will fail to resolve chronic indeterminacies among political values.

We can extend this line of argument once we notice that it is an instance of a more general relationship between reasons and individuality. Joseph Raz has argued that our individual distinctiveness as persons is a function of the underdetermination of reasons. As individuals, we often face choices that cannot be neatly resolved by the balance of reasons. That we still take ourselves to make meaningful choices under conditions of this kind plays a central role in making “us into who we are” (Raz Reference Raz and Chang1997). I want to suggest that there is an analogy with the democratic electorate. If we accept the intuitive view above that there are many worthwhile things people can do together, an exercise in democratic leadership can settle this indeterminacy. Such a process may help shape the personality of a people. To be sure, behind this claim there is a set of metaphysical views about the character of groups that I cannot address here. I will leave it open whether this appeal to supra-agents is a fruitful direction.

If one accepts the possibility of group agents who can exhibit consistent commitments across time, there is nothing puzzling about understanding their personality as something that is crucially forged by democratic leaders. Nozick concludes the earlier passage with a rare non sequitur: “Only under very special conditions can a society avoid the need for leadership of some sort” (Nozick Reference Nozick1989). It is not clear how this follows from his analysis, since we are left without any attempt to work out “very special conditions.” Perhaps he is supposing that there will be rare cases of spontaneous social homogeneity, where policy goals align with little or no commitment-shaping by leaders. Exercises of democratic leadership needn't clash with a sensitivity to the identifiable preferences of an electorate.

To capture leadership's coordinative function, we need a second condition. The base material of this account—the shared political commitment—is a notoriously fragile attitude. It is vulnerable to demagoguery and deceptive framing effects. Its persistence over time relies upon the reassurance that one's co-citizens will continue their resolve in the face of pressure to abandon the intention. So a second condition characterizes the means by which the first condition is satisfied:

Democratic leadership organizes political activity in the service of a collective commitment. It serves to connect up the ends of leadership with the means of political organization. Setting a political commitment has value beyond its informational role. It will also structure the kinds of actions that citizens ought to take in support of an end. The leader may settle the commitment that we pursue a political end, or the means by which we pursue it. It is instructive to compare this function with a proposed division of labor between citizens and representatives. Thomas Christiano argues that citizens should set the broad ends of policy, while representatives should select the appropriate means (Christiano Reference Christiano1996). Like most works in democratic theory, his account does not make explicit the role of leadership. We can, however, infer its place from his ends-means distinction. Since democratic responsiveness is limited to ends, the account appears to create room for leadership in the process of means selection.

This means-ends view crowds out leadership in two ways. First, limiting leadership to means-setting appears to invert our most basic assumptions about the activity. That a leader can identify shared aims is an intuitively plausible piece of democratic thought. This fault in the means-ends view should now be familiar. It signs on to a qualified version of the tracking assumption that we saw earlier. If lawmakers track the selected ends of the electorate, they don't seem capable of engaging in some of the most garden-variety moments of leadership—at least not without undermining their democratic credentials. Second, it is often very difficult to carve the joints along means and ends. Consider cases where there is a tight-knit relation between shared policy intentions and their required means. The centerpiece of Bill Clinton's Violent Crime Control and Law Enforcement Act was an assault weapons ban and the addition of 100,000 police officers on the streets. The Bill contains a diverse menu of means to tackle a worthy policy end. But when there are many policy ends competing for our allegiance, it can seem implausible to fence off questions of means from democratic accountability.

It is worth remarking on two features of the mobilization function. First, it avoids any artificial separation between the ends and means of political reform. By foregrounding intention formation it avoids the narrower scoped concept of “agenda-setting” (Kingdon Reference Kingdon2003). This idea has a large literature of its own, the natural focus of which is on the list of ends to be pursued. To be sure, episodes of agenda-setting will be candidates for satisfying the two functions outlined here. But it is crucial to admit reforms that aim at altering our means of pursuing a policy as sites for the exercise of leadership. As a candidate failure of leadership, consider the Obama Administration's inaction on reforming the habeas treatment of enemy combatants. Critics argued that the administration's incremental reform have been unaccompanied by any attempts at public rallying. For instance, the administration dropped the official designation of “enemy combatants” without the public urging of reforms (Perez and Bravin Reference Perez and Bravin2009).

Second, the mobilization function captures the thought that the successful instances of democratic leadership translate changes in political commitments to substantive policy outcomes. On my account, leaders leverage their successful intention-setting—forming and sustaining shared commitments—to coordinate political action. To take an example, consider Senator Richard Russell, Jr., who was a supremely skilled lawmaker known for his “ability to persuade men to cooperate and unite behind his aims” (Caro Reference Caro2009). Russell's skill at mobilizing individual votes and organizational support around these shared ends was remarkable, so much so that he was said to “lead without one's consciousness of his leadership” (Caro Reference Caro2009). On my view, followers co-hold a new or modified political commitment, but they needn't possess a self-conscious awareness that they are following a discrete agent. They may not be able to cite the relevant passages of a speech or press conference that induced them to adopt this commitment.

Let's close this section by taking an inventory of the Commitment Theory. Start with the complete universe of political activity. You already intuitively hold that some subset of this activity consists in democratic leadership. Rather than hunting for character traits that leaders tend to possess, we’ve taken a different methodological route. Our point of view has been the citizen’s. What must happen in the mental just life of democratic citizens for this activity to be authentically in play? I claimed that the activity comes into focus when we see it as essentially shared. Leadership is a doubly relational property. First, it involves a dyadic relationship between the intention-setter and the agent whose shared political commitments are set (or reset). Seen this way, leadership is a distinctive kind of “treatment.” Its success condition is a commitment—a relatively weighty intention, shared with the leader, to pursue political action. Second, the product of leadership is a commitment shared horizontally among some nontrivial subset of citizens. It was necessary for us to add the Mobilization Condition because these jointly held commitments cannot be inert. Leadership bears not only on how citizens come to have their commitments shaped and even reset, but also how they together pursue the agglomerated commitment.

DEMOCRATIC SIDE CONSTRAINTS

We have in place two necessary, but not sufficient, conditions. To earn the democratic predicate, commitment-setting and mobilizing cannot offend the ideal of political equality. In this sense, the commitment theory is modestly moralized. First, it does not aim to pick out acts of leadership that are morally justified, all things considered. Authentic instances of democratic leadership, on my view, will offend our political morality. This is a virtue of the approach. It's implausible to think that our conception of democracy should do all the moral work—so that democracy and general morality are perfectly aligned. The theory's selective appeal to normative considerations reflects the idea that democracy is not an all-encompassing political value. Second, the ambition of theory is quite unlike all-purpose theories of leadership. Well-known accounts start with premises that are unmoralized. For instance, they acknowledge that both transactional and transformational leadership can rely upon noxious means (Bass Reference Bass1985; Blondel Reference Blondel1987). This inclines them to err on the side of overinclusiveness. They authenticate exercises of power accomplished by arm twisting and other forms of duress. Whether this casts doubt on the fruitfulness of the “unified theory” approach to leadership, I flag it to distinguish my view from two rival approaches.

Our functional approach already blocks certain forms of undemocratic leadership. If you hold unaccountable authority over me, it is hardly obvious that you can satisfy the two conditions. The Mafioso who tells his rival that “we are going to New York City” and proceeds to throw him in the trunk may create what looks like a shared intention. After all, the rival's life may depend on forming that intention. But there's no sense in which their two intentions interlock. His intention doesn't hold that they travel by way of his rival’s intention (Bratman Reference Bratman2014). Here the commitment-setting condition simply isn't satisfied. The same can hold true on the political scene. A menacing politician can build up avowed support for a pet policy without creating a joint intention, for implicit threats can make joint activity impossible. Followership, on this view, isn't mere compliance. An electorate can even come to hold a political commitment that no more authentic than the artificially “shared” intention of the person in the trunk and the person in the driver's seat.

Suppose democracy's value lies in the relationships of political equality. In a system of self-rule, we simultaneously relate to each another as co-authors and co-subjects of our political structures (Beerbohm Reference Beerbohm2012; Christiano Reference Christiano1996). This invites two kinds of side constraints. The first depends on the character of the relationship between leaders and followers:

Undemocratic relationships come in grades. This constraint embeds the intuition that instances of leadership can be more or less democratic. At some point—possibly one that is irreducibly vague—a practice will lose its democratic license altogether. When strong-armed threats are used to coordinate collective activity, we don't normally think of this as an instance of us engaging in shared intentional action. Then we are not aptly described as cooperating together. You’ve ordered me to do something. There's a lightweight sense in which we may perform it together. I put my hands behind my back, and you fasten the handcuffs. Given the background threat conditions, it isn't untrue to say that our intentions are interlocking in a particular way. But the presence of this background condition snuffs out the possibility of the kind of joint intentional action that I’ve been appealing to in this article. We aren't co-engaged in my arrest in the same way that we went on a walk earlier.

Suppose a prison warden gives a rousing speech to the inmates in the mess hall. He is very convincing. They alter their everyday commitments. But given that he is not ultimately answerable to them, it would be a category mistake to claim that his leadership is democratic in kind. These examples may strike some readers as overly stylized, since it involves such an overwhelming asymmetry of power relations. Consider, then, the conductor of a symphony. Her authoritative position to make interpretative decisions may disqualify her influence from counting as democratic. Notice that there is nothing in this condition that limits the exercise of democratic leadership to elected or appointed officials. In contrast, consider a nonpolitical setting where democratic leadership is not only possible, but appears to be a routine occurrence. The Orpheus Chamber Orchestra prides itself on embodying democratic principles, jettisoning a conductor for what it describes as a more collaborative musical experience.

How does leadership, democratically exercised, bear on the authority of leaders? If unaccountable asymmetric authority relations make democratic leadership impossible, what follows for the obligations of citizens as prospective followers? To make sense of the place of authority in our conception of leadership, consider a thought experiment far outside ordinary politics. Joe survives a plane crash. A flight attendant is coordinating rescue efforts, and issues an explicit order to him in this process. David Estlund diagnoses the life-and-death case in this way:

We are to assume that the flight attendant has the expertise to craft the most effective exit plan. But experts are fallible—famously so. Suppose the passengers are told to “do as I say” but given a substantively mistaken order. Estlund thinks that the obligation to comply retains its force. For him, it's not enough to think that Joe has an obligation to treat her as a leader. This gets at the right verdict for the wrong reasons. Joe should follow her precisely because she is a practical authority. Her authority derives from the function of her leadership—setting shared ends for the injured passengers to follow and then organizing a process of rescue. We can remain agnostic about the source of Joe's obligations. But the theory should help explain how democratic leadership can confer authority, if legitimate political authority can be wielded. It's common to assume that authority and leadership stand in is some close-knit relationship.Footnote ⁸ For those who think that legal officials are capable of creating genuine obligations, the view can shed light on how exercises in leadership can enhance the authority of binding pronouncements. A representative may, on my view, earn the power to act in our name. But the theory needn't take a position on whether legal obligations exist, or whether democratic leadership can make the claims of an authority weightier. It can and should be modular, working with a range of views about political authority.

The relational constraint needs a complementary principle. It guards against coercion, but leaves other moral offenses available that aren't compatible with the value of democracy. Even if democratically authorized officials manage to create an authentic shared commitment in a group of individuals, they can deploy profoundly undemocratic means. The second constraint limits the ways an electorate can be brought to hold policy commitments:

There's no general prohibition on forms of rhetoric or oratory. Used properly, these tools of communication are perfectly capable of avoiding wrongful deception. The constraint flags methods of influence that are deceptive, breaching trust with its targets. It is broad enough to encompass deception by falsely asserting, misleading or even omitting information. This rider is necessary to trim leading accounts of persuasion. Take, for instance, Bryan Garsten's claim that to “truly persuade people is to induce them to change their own beliefs and desires in light of what has been said” (Garsten Reference Garsten2006). Taken on its face, this condition cannot suffice. We know well the politician who wrongly deceives without lying. If he leads citizens to hold false beliefs, and willfully refrains from revealing facts that would lead to correction of these false beliefs, he violates this background condition. I doubt we’d wish to say that any speech act en route to an epistemic change is an instance of persuasion, let alone democratic persuasion. For our purposes, we can leave open the complete explanation of the wrongs of deception and manipulation. If democracy just is a series of answerability relationships, it is clear what is motivating this constraint. It's undemocratic to for agents to bypassing the rational faculties of their principals. It deprives us of the ability to hold political actors to answer for their statements and actions.Footnote ⁹

The relationship between leadership and democracy is not an impossible one. But it can be mutually destructive. Our theory shouldn't explain away the potential for normative tension. It owes us an account of why democratic theorists have been tempted to banish leadership from their ideal polity. We now have the resources for such an error theory. There's nothing off about thinking of leadership as a “treatment.” After all, by design, leadership treats individuals as a means. Whether it also treats them as ends will make all the difference. Joint intentional activity serves as the primal scene of our theory. It's not enough for leaders to impact the intentional structure of followers. The resulting commitment is something shared by F and L. This reveals a comparative advantage of the view over theories that rely on currencies like beliefs, desires, or even goals. It's not enough to correct the beliefs of a group, or to get them to endorse a valuable end without the resolution to act. Democratic leaders act with, not principally on, their followers, precisely because they create jointly held political commitments, and come to share them.

With a working theory in hand, the next two sections are first-pass attempts to see if the view serves the usual theoretical ambitions—parsimony, explanatory power, and economy. Its plausibility will turn on its ability to illuminate actual and imagined cases. I aim to show how the theory can intuitively bring out intuitive features of agenda-setting in a democracy and can sort out a range of cases of political influence. I will do this by arguing that my account can avoid two core classes of challenges—or, at least, sustains less damage in the face of these objections than rival accounts.

THE OVERCOUNT OBJECTION

My argument has been indirect. I have given an account of a puzzling political phenomenon of large-scale action by appealing to a puzzle of small-scale action. How can we come to share an intention in ordinary life? Now let's subject the Commitment Theory to two kinds of stress testing. Does it overcount the phenomenon, delivering false positives? Or does it fail to account for cases that are intuitive candidates? Let me suggest three points of contrast to show the comparative advantages of my approach. In each case, more permissive views engage in what I will argue is overcounting.

Second- vs. Third-Personal Justification. Viewed together, the two functions help us notice a distinctive property of democratic leadership. It is neither addressed to a single person, nor can it abstract away from interpersonal address altogether. The identity of the addressee and the addresser play an irreducible role in our understanding of them. On my account, democratic leadership supplies directed reasons that are distinctly second-personal plural in kind. We can use this feature to register G. A. Cohen's distinction between two kinds of normative justification (Cohen Reference Cohen and Darwall1995). Third-personal justification provides the materials for a policy conclusion, but it is identity-neutral in two ways. The identities of the agent doing the justifying and the agents on the receiving end of justification are irrelevant. To see third-personal justification in practice, imagine a lawmaker who notes the reasons for a policy on the committee room's chalkboard. She frames these as desirable states of affairs that the candidate policy will tend to bring out. Even if her board work leads to a shared intention to enact the policy, the form of her justification excludes it from satisfying the desiderata outlined above.

On my account, a democratic leader succeeds in agenda-setting and mobilization by making normative claims on her addressees. This presupposes some kind of accountability relation between the addresser and addressee. When she makes a claim or demand on her audience, it takes on an essentially directed form. The claim that she makes on them depends on her possessing a kind of standing. The kind of authority implicated in democratic leadership derives from the institutional positions of lawmakers and citizens, who stand in a distinctive kind of bilateral accountability relationship. This can help answer the concern that the stock model of leadership and “followership” has no comfortable home in a representative democracy. How can elected officials who answer to us credibly make demands on us? The demands that they are in the position to make on us are of a different kind than the commands that we are in the position to deliver to them. They can hold us to account for our political convictions, challenging us to rethink means-ends relationship between our principles and policy. They can convince us to act against our short-term interests. But it's not enough for them to strong-arm elected officials into agreement. For an alternative view, Samuel Kernell's account of leadership has presidents make direct appeals to the public to induce legislators to vote for their preferred positions (Kernell Reference Kernel2007). The success condition of this view is the change of view of lawmakers in response to this kind of political leveraging.

This mislocates the site of leadership. To be sure, if a politician is able to reshape the policy commitments of a constituency and mobilize them around that commitment, this may have the effect of putting pressure on her representative. My approach is about to register certain instances of “going public” as authentic acts of leadership. But without an account of the proper wiring between a leader's address to a subset of citizens and her response, the view delivers a string of false positives. Running an exclusively negative campaign can be an effective way of forcing a lawmaker to revise her position on a vote. But inducing votes in this way—shorn of mobilization of political intentions—is not counted by our theory. And I take that to be a virtue of a less permissive picture of use of the bully pulpit. Genuine bullying, on my view, should not be conceived of as leadership.Footnote ¹⁰

Plans vs. Goals. The idea of shared political commitments has been put broadly, encompassing general institutional aims and particular choices of public policy. Still, the concept serves as an essential constraint in the theory. Consider the distinction between having a goal and having a plan (Bratman Reference Bratman1987). You have a lot of goals; they needn't all be co-possible. Your bucket list is too ambitious. So you could aim for each, letting the world decide which of these goals is met. Plans aren't like this, if we take them to involve commitments to act. They impose rational pressures on us to avoid accepting inconsistent projects. Here I have taken political commitments to be subject to the rational pressures of authentic planning. Consider the lawmaker who addresses a constituency with a major policy speech. She convinces a significant number of them to endorse a set of mutually incompatible policy goals. Accounts of leadership organized around goals aren't in the position to resist this as a token of leadership (Howell and Brent Reference Howell and Brent2013; Kane and Patapan Reference Kane and Patapan2012; Keohane Reference Keohane2012).

By taking shared commitments as its unit, our theory avoids overcounting. The rational prohibition of holding inconsistent intentions is no less important for the individual than the collective. If anything, it's more troubling when a large set of citizens are induced to hold mutually incompatible policy intentions by an enterprising politician. Lawmakers who propose plans that are not co-possible fail to satisfy the theory's two conditions. If the goal itself is too vague to count as a plan, they will fail to meet the commitment-setting condition. Or if the shared commitments induced by a leader aren't jointly actionable by a polity, the mobilization condition will go unsatisfied. Institutional policies like party planks, written in sufficient detail, can help lawmakers propose co-possible goals. They make possible policy itineraries, a set of political aims that are co-realizable. Ruling out incompatible goal-setting offers a second advantage. It helps us explain what is troublingly undemocratic about certain kinds of issue framing. John Zaller explains how framing can appeal to incoherent plans:

There is a natural worry about this theory of mind of the citizen. If we are so highly susceptible to frame manipulation, it's hard to see how nondeceptive means of persuasion are available. An electorate may be so benighted that authentic leadership becomes impossible. We still have to face the less dire problem. When politicians ply with “top of the head” considerations, is there a sense in which they can be said to lead? We know that salience-raising is capable of impacting aggregate opinion, even when the bare policy preferences of individuals hold constant. President George H. W. Bush was able to raise the profile of the “war on drugs.” By the end of his term, a record 64% of respondents claimed that illegal drugs were the most important problem before the nation. Route attention to the issue drove the issue's salience, but it isn't clear that this campaign affected policy commitments of the electorate. Greater salience of the issue didn't foster a practical commitment to support policy reforms. If so, this isn't an instance of leadership on my account, since no new shared intention was forged or sustained.

Opinion Convergence vs. Joint Commitments. Consider two political speeches that are plausible-sounding candidates for democratic leadership. The first generated a 20-point spike in support for universal health insurance. The second produced an equally significant boost in support for the Second Gulf War overnight. Bill Clinton's speech to a Joint Meeting of Congress is sometimes cited as a paradigmatic moment of leadership: a compelling script coupled with a specific veto threat. In another plenary legislative session, George W. Bush promised the Iraqi people that the “the day of your liberation is near.” It is tempting to treat such strong shifts in public opinion as probative evidence of democratic leadership. But this kind of preference shifting, on my view, isn't sufficient. This logic applies with equal force to William Riker's category of heresthetics, where a political agent manipulates the order of a vote in order to change policy without any underlying attitudinal shift. Reducing leadership to a metric of policy impact will generate false positives (Riker Reference Riker1986).

This suggests a burden of argument for any action-theoretic approach. If democratic leadership is about the forging of shared commitments, we need a rough and ready way to indicate what kind of attitude citizens are expressing. Are they testifying to a belief that has come, in John Zaller's parlance, to the “top of their heads,” a conviction that coherently connects up with their political views, or an attitude stable and action-directing to count as an intention? The best available measure separates “directive” and “permissive” opinion (Key Reference Key1961). A directive opinion is practically oriented. It involved a commitment that the polity act in a particular way. In contrast, a permissive opinion is not rationally limited in this way. One can hold permissible opinions that do not hang together, and that are not co-possible. After voicing their opinion about a policy choice, citizens are asked whether they would care if the government acted against their view. Those with permissive opinions answer in the negative: on their self-reporting, “they don't care one way or the other” (Moore Reference Moore2008). To call this posture towards one's beliefs about war and peace, or life-saving medical care “permissive” has the potential to mislead.

Permissive beliefs, on inspection, aren't beliefs at all. If the expression of beliefs involves truth assertion, then what is being voiced isn't a belief at all, but a coy language game between citizens and pollsters. The problem is that leadership is not properly measured in the currency of what one pollster calls “superficially held views which may be discarded at the next moment” (Moore Reference Moore2008). The numbers following Bush's speech bear out this disturbing attitude towards one's own views about political choices. When the 71% support for invading Iraq is broken down into directive and permissive opinion, the numbers change strikingly. Of those 71%—most of whom had come to announce support for the war in the past three months—41% admitted (disturbingly) that they did not care whether the United States went to war, while only 29% claimed they would be upset if the invasion did not occur. A slightly greater number of Americans (30%) announced that they would be upset if their government went to war. To be sure, directive opinions are an imperfect proxy for shared policy intentions. They indicate stronger attitudes than off-hand beliefs, coupled with negative reactive attitudes if one's intention failed to be enacted. They are, in short, the closest available measure of collective commitments available to democratic theorists.

Even if we avoid counting attitudinal changes that avoid any commitment, there remains a powerful line of objection to my view. Suppose a leader resets the joint commitment of a set of citizens around a trivial problem in order to distract them from a morally urgent problem. Isn't this a damning counterexample, since it forces the theory to “count” a reset intention without making a value judgment as to its moral significance? The challenge takes advantage of the modest normative assumptions that we packed into the view. We didn't want the concept of democratic leadership to be a vague term of praise for any political action that we approve of, all things considered. This makes the account vulnerable to the charge that it counts political activity as leadership in cases where we take it to be a counterparadigm of leadership. This kind of redirection seems to be a failure of leadership. To blunt this worry, does my approach need to import more robust normative content? Should we insist that the political commitments that result from leadership be morally worthy?

Consider two lines of response. First, the leader who attempts to use a distracting issue may well show their hand. They may find it difficult to produce political commitments toward issues that are wholly symbolic. One way of making sense of the idea of a purely symbolic voting issue is that it defies attempts to organize plans around it. It misdirects precisely by calling attention to goals that are abstracted from practice or, taken together, are not co-achievable. Here the distinction between mere epistemic changes in citizens and inducing practical commitments will help ruling out attempts at distracting a polity. So political distractions may avoid generating false positives for our theory, since they are transparently not about what we should do as a democratic people, but a vague idea of what we should be outraged at. Here our theory's base ingredient, joint commitments as an especially weighty from of intention, is crucial. It allows us to avoid registering any change in the intentional structure of followers as genuine leadership.

A second response is concessive. There's nothing built in my account that can block the full force of this objection. If an agent satisfies the theory's two conditions, and this creates or resets a shared intention towards a middling political issue, we have to acknowledge this as an instance of leadership. Nothing in the view demands that we delight in the satisfaction of the theory's conditions. But scrubbing the concept of abuses of leadership would create even more trouble for the theory. In epistemic environments that are toxic, where the citizenry has come to accept a wide range of false empirical beliefs about their shared political institutions, we shouldn't be surprised that leadership, even in its democratic mode, can be a tool of injustice. It has the capability to put political institutions into a standstill, or produce results that are collectively irrational. This verdict counts in favor of the theory, however unwelcome in practice.

THE UNDERCOUNT OBJECTION

The other worry is false negatives. Does the commitment theory fail to count the practice when we strongly suspect it to be occurring? A driving concern is that the normative assumptions of the theory are too thick. Demanding a species of jointly shared intention can seem too strong of a condition. The previous trio of objections focused our attention on the “supply side” of leadership. We considered how my view constrains the means by which a leader can get others to share a policy intention. Now we can consider the “demand side.” What are the means by which a citizen can come to accept the proposed policy intention of a candidate leader and act in ways with fellow citizens to bring about this joint intention?

Blind Following or Cue-Taking? Seen from the point of view of the citizen, does the presence of authority rule out the possibility of leadership? From the start, our account filtered cases of coercion. But subtler forms of authority may pose a challenge for my account. When we reflect on examples of leadership, we tend to imagine a certain pedagogical mechanism behind it. Citizens’ policy intentions are reset when they are exposed to the epistemic effects of campaigns. In support of this picture, Andrew Gelman and Gary King found that campaigns serve as critical treatments. When voters learn party positions, they can switch their support in line with their political convictions. Candidates can elicit support by making clear policy positions and attracting voters who arrive with a working ideology “in-hand” (German and King Reference Gelman and King1993). But this picture of campaigns may well get the causality backwards. There's impressive evidence that “instead of leading, citizens are just following” (Lenz Reference Lenz2012). Democrats who learn that their party is politically liberal may adjust their underlying ideology to map onto this piece of information. Citizens are inclined to accept policy positions that are held by their party and preferred candidates.

This is a broader idea of leadership and its cognates than what we are working with here. There's a brand of followership where policy positions and even certain political principles seem to be handed down from lawmakers to citizens without independent deliberation. For instance, when supporters of a politician learned that he opposed expanding a children's health care program (SCHIP), they took on this opposition in their own views. This effect holds true even when policy positions are ascribed to parties that cut against their usual ideology (Cohen Reference Cohen2003).

Our dilemma is this. If my approach refuses to register this kind of “followership,” it may be seriously undercounting the pervasiveness of leadership in contemporary democracies. But if it includes this “inverted” species of leadership, it is liable to the overcounting objection. We can dissolve this dilemma by distinguishing different ways by which citizens come to share and organize around a shared policy intention. If citizens are truly flying blind, coming to hold a policy intention “on the authority” of a candidate or organizing policy, my view doesn't have the means to avoid the implication that this is an instance of leadership. What it reveals is that democratic leadership can be overdone. We can explain the worry about this “inverted” model of democracy without jettisoning our working idea of leadership. Taken together, our political morality calls for some kind of balancing between the “pulls” and “pushes” of democratic practice—where citizens demand policies from representatives on some issues, while representatives lead on other issues, pushing back and reshaping the commitments of their constituents. This back and forth, seen together, may roughly approximate a certain kind of democratic ideal. What is troubling about this emerging literature is that it suggests that there is, in one sense, too much followership. The citizenry is akin to a pliable mold upon which leaders press policy positions. What is worrying about a surplus of leadership and deficit of representation? Our theory helps explain what is democratically troubling about this imbalance. When Lincoln defended a kind of leadership when constituents had “gappy” views about policy, he implied that limited followership was acceptable. But his account of representation also treated responsiveness as functioning as a constraint on political action. If citizens are overly susceptible to attempts to set and reset their political commitments, they can seem flighty, subject to Alexander Hamilton's pejorative force, “the little arts of popularity.”

In contrast, consider a species of followership that need not concern us, but gets unduly attacked by normative and empirical theorists who hold that in an ideal democracy “the people lead.” This slogan implies a level of political entrepreneurship on the part of ordinary citizens that is wildly demanding. Citizens who rely on party cues sent by candidate and lawmakers need not be blindly following, “taking on authority” the beliefs of another agent. The worry of extreme deference can be exaggerated. A candidate or elected official who balances the need to be responsive with the obligation to lead can, on my view, earn the authority to speak and act for a group. If they build up a track record on policy issues, it can be sensible for citizens to take seriously their policy cues. They can think of their legislator's policy positions as a rough and ready guide for what they would think—were they to devote themselves to policy analysis. Now this kind of limited epistemic deference has its risks. In a system where policy influence can be bought and sold, it makes citizens vulnerable to supporting policies not on the merits, but on the mistaken assumption that their shared commitments are being honored.

Frame Manipulation. If we think of democratic leadership as abstaining from coercive or deceptive means, are we likely to undercount it? We now know how momentous alternative ways of posing an issue can be for eliciting different levels of public support. It's tempting to think that my account will refuse to count changes in political commitments that are the result of framing effects. Does deliberately altering a frame violate the nondeceptive constraint? When lawmakers use words that they know have taken on a strongly negative valence—like “welfare”—this may be plausibly seen as incompatible with democratic leadership. But to wish away framing is to make a simple mistake. It implies there is a relatively simple “neutral frame” available. It also puts a heavy constraint on the use of language in persuasive argument. If political agents must search out language that balances out competing frames, this would make for an onerous limitation on public argument. But there may well be a subclass of frames that fail the nondeception test. Strong candidates include frames that are deceptive or misleading, e.g., the “death tax” or “ground zero mosque.” The systematic use of frames that violate this condition will disqualify an exercise of influence consistent with the functional criteria.

Political Risk-Taking. We tend to associate leadership, at least in a democratic habitat, with self-sacrifice and its relative, courage. The “profiles in courage” model that we examined earlier attempted to build risk into the idea of democratic leadership. Here I’ve resisted making this move, since sticking one's neck out is not a necessary or sufficient condition for the exercise of leadership. Still, my account should explain why leadership has been perceived as a risk-bearing activity. James Read and Ian Shapiro have stressed this inherent risk in their account of “strategically hopeful leadership.” On their view, leaders “initiate new, tentatively cooperative approaches more readily than can whole communities, but in the process they accept a different degree of risk than the community does” (Read and Shapiro Reference Read and Shaprio2014). This mode of activity—coordinating others’ plans in a way that incurs risk—isn't incompatible with my approach here. My approach can explain why it's natural to think that democratic leadership carries with it significant political risks. The origins of joint activities, writ small and large, are typically risky. When you and I are able to go on our first base jump together, peering down off the edge of the cliff, each of us is inclined to suggest that the other “goes first.” Or, if we are attempting to synchronize the jump, we’ll decide to count down from 5. There's always that possibility that one of us will fail to act on a joint intention, whether to take the leap or to join the movement. That kind of breakability explains why risk-bearing has been seen as a central feature of paradigmatic moments of leadership.

Consider the panderer. Representatives allergic to political risk will accumulate a record of purely defensive voting, speaking, and position-taking. They treat constituent preferences as sacrosanct, even when they aren't co-achievable. What's troubling about this form of pandering? Surely the use of “acoustic separation”—saying different things to different audiences—can involve deception (Dan-Cohen Reference Dan-Cohen1984). Our theory suggests that the core wrong at stake is different. The panderer's fault lies in the all-out refusal to engage in the fragile activity of commitment setting and resetting with constituents. It doesn't follow that all anticipatory voting is worrisome. When an elected official is responsive to the preferences she anticipates her constituents holding on the next Election Day, is her voting record suggestive of leadership or followership? There is no general way to answer to this question. Learning effects can take considerable time and repeated exposure to reach less than fully attentive publics. What looks like under-the-radar priming effects may be explained with greater parsimony by understanding campaign as a pedagogical mechanism (Lenz Reference Lenz2009). There is nothing objectionable about showing political sensitivity to the views of a future constituency. That's basic to politics. The fault in pandering lies in treating the electorate's policy commitments as fixed points in a political strategy. To borrow Rousseau's maxim, pandering takes citizens’ preferences as they are, not as they might be. It eschews the risks inherent in democratic leadership.