Abstract
Much liberal-democratic thought has concerned itself primarily – even exclusively – with coercive interference in citizens’ lives. But political actors do things – they engage in influential speech, they offer incentives, they mislead other actors, they disrupt the expected functioning of decision-making mechanisms etc. – that fall short of coercion, yet may nonetheless call for normative evaluation and public justification, precisely because they serve to purposively alter citizens’ beliefs, intentions and behaviour. With this article, I explicate a conception of political manipulation to capture this sort of interference, and to distinguish individual manipulation from the manipulation of nonindividual agents like committees, institutions and states. The account, beyond being necessary for further work on the ethics of political manipulation, should prove useful to both normative thinkers interested in power, justice and the ethics of democratic decision-making, and empirical scholars in search of a conceptual apparatus to sharpen their investigations into the exercise of subtle forms of political power.
Introduction
Consider the following short cases. A candidate misleads a voter about the record of an opponent. A lobbying organization prints marketing materials that closely resemble utility bills to increase recipient attention. Minority voters are targeted with calls falsely informing them of polling place closures. A campaign operative pursues an undercover operation to bait an opponent into recorded corruption. A legislator undermines the alliance of a pair of colleagues, knowing each will need to offer partnership with him instead. A voter submits an inaccurate ordering of her preferences with an aim to better secure her true preferences in the social ordering produced through a ranked ballot election. A committee draws a new electoral map isolating the opposition party’s support in fewer districts than the previous map.
My aim is to systematize the intuition that these cases are of a kind, that they share relevant features, that our thinking on liberty and democracy is improved by conceptualizing cases like them together as illustrating a distinct species of interference. Readers may not share normative intuitions about the rightness, legitimacy, etc. of these examples. Indeed, they should separate us across divisive principled commitments, just as a range of examples on state coercion would. But I argue that, like examples of coercion, they share characteristics uniting them conceptually within a single phenomenon I will call political manipulation.
I argue this against those like Cass Sunstein, who doubt that manipulation is a coherent single concept for which necessary and sufficient conditions may be given (Sunstein, 2016: 81), 1 and against those like Robert Goodin (1980) and Anne Barnhill (2014), who give analytic definitions that I will show conflict with considered judgments about the relevant conceptual issues, or are not concerned with political contexts, respectively. But we need such a conception in the first place because much liberal-democratic political thought has concerned itself primarily – even exclusively – with coercive interference in citizens’ lives. Following Locke and Mill, liberals have developed principles of non-interference to capture the wrongfulness of undermining others’ agency by forcing them to act according to some external-to-them idea of normative action. On this broad set of accounts, justification is owed for coercive policies precisely because they are coercive (Hardin, 1990: 79; Nagel, 2003: 218; Nozick, 1974: ix; Rawls, 1993: 216–217). 2 None of the cases above involved coercion, yet they are certainly exercises of (sometimes significant) power for political ends. We need to think carefully about whether and how such exercises might call for justification.
The stakes of clarifying manipulation are high. We need normative theories that parse the ethics of a variety of forms of interference, coercive or otherwise. Such theories also serve to assess the justice of political institutions that determine the extent of manipulation within a system, and the degree of citizen and institutional resilience against inevitable acts of manipulation. Without a thorough conceptual understanding of the nature and varieties of manipulation, we cannot hope to begin thinking normatively about its presence in our politics. It is towards that wider normative project that this article’s conceptual work aims. In what follows, I argue for a conception of political manipulation covering a wide range of cases not well described as coercion, persuasion, interference or alternative takes on manipulation.
The conceptual approach I take has three methodological features worth noting. First, it is non-moralized. Following several other thinkers (e.g. Baron, 2003: 39–40; Coons and Weber, 2014: 5–6), I focus on the conceptual account as a prior condition for exploring a practice’s normative status. Expanding on the liberal tradition, I take it that there are forms of non-coercive interference that call for justification in much the same way that coercion does. Good liberals should understand manipulation similarly to the way they understand coercion: as a non-moralized concept that is nonetheless prima facie wrongful, such that justice requires reasons of some sort for its authorization. 3 Manipulation carries a strongly negative moral valence for good reason. Indeed, we might expect that manipulation will be all-things-considered wrongful in many cases. But a non-moralized account leaves open the possibility that even the forms which are most often wrongful can in principle be permissible depending on circumstance and a wider theory of justice. Defining manipulation as ‘wrongful covert influence’, for example, begs agreement on what constitutes wrongful – it builds the full normative assessment into the concept itself. This article instead seeks to develop a conception of manipulation that can then be employed according to the dictates of a variety of normative theories. It is in that way a tool of wide application, unwedded to any particular normative theory, standing as a conceptual tool that should both highlight the places where theories of justice come together or apart on cases, and serve as a common language in describing a set of practices absent joint moral commitments.
Second, it is concerned with interference over agents qua decision-makers, extending our focus beyond individuals to group decision-makers. One of the central conceptual advances of this article is that it defends an account that captures both individual manipulation and instances where the target of manipulation, the relevant decision-maker, is not an individual but a group of individuals working in concert to produce a social choice.
Finally, it is a political, not a broadly ethical conception. While the conception is analytically general in that it covers a range of cases involving varying agents, techniques and relationships, I modify it here with the adjective political to focus on cases involving the organized use and distribution of political power typified by the state. This follows liberals’ traditional focus on state coercion as the main justificatory gauntlet to be taken up by non-anarchist thinkers. It expands that focus by taking democratic values seriously alongside liberal ones, allowing us to take manipulated group agents seriously.
The article is divided into two main parts. The second section is concerned with crafting and defending an analytic definition of political manipulation. That conception aims to cover a range of political cases of covert interference both directly and indirectly applied to the subject. Subsections elaborate arguments for indirect manipulation that does not see the manipulator acting upon the subject, the role of covertness and the extension of the idea to group agents of various kinds.
The third section concludes by anticipating a critical methodological reply, namely that the account is insufficient if its aim is to allow us to empirically assess cases as either manipulative or not. I argue that this is an unavoidable feature in describing intentionally covert forms of influence – they are designed by their orchestrators to be empirically untractable.
Conceptualizing political manipulation
To fix terms, the conception I argue for can be given as:
An act of manipulation is any intentional attempt by an agent (A) to cause another agent (B) to will/prefer/intend/act 4 other than what A takes B’s will, preference or intention to be, where A does so utilizing methods that obscure and render deniable A’s intentions vis à vis B.
Manipulation on this conception sits on a spectrum of social interaction bounded on one hand by persuasion, which treats others as reasonable, and on the other by coercion, which goes beyond manipulation into threat of force (Rudinow, 1978: 338; cf. Goodin, 1980: 8). 5 Persuasion is typified by the communication of information, analysis and/or reasons meant to produce assent to some proposition. The term persuasion indicates the kind of influence that engages and respects another’s autonomous decision-making processes, rather than influence attempting to subvert those processes, thereby alienating the target from their agency. 6 Keith Dowding (2016, 2018) argues that it is the persuader’s relationship to the reasons offered that distinguishes persuasion from manipulative discourse. He identifies the former as communication causal in the target’s assent to some proposition, where the interaction is characterized by (1) common reasons for assenting to the propositional target of the persuasion, meaning that the reasons that the persuaded person assents to the proposition must constitute the reasons the persuader has for her own assent; and (2) the persuader’s intentions must be that she and the target reach the truth, not that the target assent by whatever rhetorical means necessary. Dowding rightly highlights some core features distinguishing persuasion from interference, most notably the attention to reasons and reasonableness evinced by his two conditions. An action pursued as the result of persuasion, while counterfactual to the subject’s original intent, nonetheless is consistent with the subject’s full autonomy. 7
Coercion is an attempt by A to influence B to conform to A’s will by threatening some eventuality that B has good reason to wish to avoid. 8 I will flesh out a sharper division between manipulation and coercion below, but for now a crucial difference is that coercion makes no effort to hide itself from B: B should understand the threat, explicit or implicit, in A’s actions, and see its motivating force towards A’s ends, knowing they are A’s ends. Manipulation, whether or not it involves something threat-like, seeks to (at least plausibly) hide A’s intentions.
A more complete normative analysis of manipulation, its harms, wrongs and (potential) justification, is a subject for other work. However, it is worth highlighting an intuitive conceptual feature that would bear significantly on any such analysis as well as motivating this one. Manipulation disrupts autonomy. There are many ways to describe autonomy disruption, external control, reason alienation 9 and related concepts, each corresponding to myriad schools of thought on the wider idea, but for our purposes both coercion and manipulation disrupt autonomy by altering the locus of decision-making. Each disrupts B’s authorship of a decision, transferring it to A in whole or in part. When A employs coercion to force B into some action, it is A’s preferences which guide B’s hand. Similarly, when A manipulates B through selecting the options which B must choose between, it is A’s preferences and not B’s which are predictably secured. A thorough treatment of the value of autonomy and the concomitant wrong of its violation through manipulation is not the topic here. It is enough for our purposes to speak of A authoring decisions that absent A’s intervention would be authored by B. This is the thinner idea used throughout.
Individual manipulation
My first task is to account for the manipulation of individuals. The second is to expand that account to cover political group agents of various kinds. The second is particularly important because existing work (e.g. Barnhill, 2014; Baron, 2003; Dowding, 2016; Noggle, 1996) focuses exclusively on individual agents, or the collected individual decisions of populations (e.g. Sunstein and Thaler, 2008). 10
Consider as a starting point two definitions that have been offered by scholars. The first is an older working definition from psychology: ‘A (agent) attempts to manipulate S (subject) if A attempts to influence S’s behaviour by means of deception’ (Bursten, 1973: 11; cf. Rudinow, 1978: 339). The second comes from Eric Beerbohm, who writes: ‘Manipulation occurs whenever external agents exercise … influence in a way that gives the … [target] legitimate grounds for complaint’ (Beerbohm, forthcoming).
These definitions seem to me to fail on a number of dimensions. Surely (perhaps especially in political contexts) there are indirect varieties of manipulation, as well as techniques for manipulation that involve things other than deception, beliefs, desires or emotions. 11 Indeed, one of the central types of manipulation we will be concerned with below involves altering either incentives or the structure an agent acts within in order to secure a preferred outcome.
Two simple examples show the deficiencies of both accounts in this regard. Imagine first a case in which A undermines B’s alliance with C by giving C some negative falsehood about B knowing this will ensure B cooperates instead with A. And for a second case, change nothing but the deception by instead imagining the information about B is true and not false. In the first case none of B’s beliefs, desires or emotions have been altered: B has not been directly acted upon at all. But we should nonetheless describe this as a case of indirect manipulation on B (in addition to it standing as direct manipulation of C).
Now we might square this example with both definitions by following Robert Goodin, who in defence of Rudinow’s conception suggests that the counter-example unhelpfully obscures ‘who or what it is that is being manipulated’ (Goodin, 1980: 11). The deceived subject is not A, but C. Indeed, according to Goodin, it is only because C is deceived that this remains a case of manipulation at all. This response, in one sense, seems right: in the first counter-example, C is being made to act through A’s deception; her breaking the alliance with B may very well be the result of A’s manipulation. However, that says nothing at all about why B’s resulting need for an alliance with A is not also the direct result of A’s interference. A knows how B will react to changes in circumstance, and ensures that the circumstance that eventuates is one in which B acts according to his wishes. Insofar as any such A has the power to change B’s circumstances in such a way, it is thus A, and not B, who chooses B’s actions.
So the directness requirement seems suspect, but we should also doubt Goodin’s insistence on the necessity of deception. To see why, consider the second alliance case. Here no deception has occurred. A reveals some true fact about B that results predictably in C breaking the B/C alliance and B needing to cooperate instead with A. That the fact is true is immaterial to the thought that A has effectively chosen for B by acting on C. The change from the first to the second case is then only whether A manipulates C on his way to manipulating B. In the first, A deceptively influences C to abandon B in favour of A, and so manipulates both, while in the second he deceives neither but manipulates both. In either case A, with plausible deniability about his intentions, is causal in decisions made by both A and B.
To draw this point still further, imagine that B and C are not (contextually) autonomous agents, but instead are rule-bound actors (government agencies and grant applicants, for example). B must operate under an alliance with some agent, and C must never treat with others who have broken alliances in the past. A reveals to C that B has broken a past alliance and presents himself to B as the only alliance partner thus available. We might commonly say that A has ‘gamed the system’ in using the rules governing B and C to ensure his preferred outcome. I want to say, contra Goodin and Rudinow, that this is an example of manipulation, as his knowledge of the agents has enabled him to cause them to act on his will rather than their own, and he has done so at minimum plausible deniability about his intentions. Whether the agents are rule-bound or wilful is not a difference in kind but a difference in the degree of how legible/predictable their reactions are to A. So long as B and C are sufficiently predictable reactors to A’s possible interventions, A can select their ultimate actions at his whim. And note here that deceit has not entered into it.
What I argue is distinctive is that the subject of the manipulation is redirected in their will towards ends chosen by the manipulator while remaining unsure about the manipulator’s intentional causation of their choice. That may be accomplished through deceit, but covertness about one’s influence need not necessarily employ deceitful communication. This covertness condition distinguishes manipulation from cases of where we describe A ‘forcing the hand’ of B. These are cases where A simply has sufficient power to cause B’s choice without caring to or needing to hide the fact of his interference. Such relationships characterize coercion, and shade towards manipulation as they become more subtle in their machinations. 12 A reason for making this conceptual cut when (many) others are available is that there is an important moral difference for B between being covertly influenced/controlled and having her hand forced overtly. B’s autonomy is impacted in both, but when those impacts are covert under my conception B’s autonomy is not only impacted, but she is left with distorted, incomplete or no knowledge of her predicament. She will thus be worse off with respect to her autonomy for not being aware that she is compromised and thus being unable to act coherently to rectify or resist ongoing interference. It may be that knowledge of her interferer’s intentions does not leave B able to do anything about it – A may have sufficient power that B cannot resist him in either case – but intuitively B faces a different and more vulnerable moral predicament under the covertness condition. This conception allows us to take that predicament seriously.
Now in the alliance case A has sought to manipulate B in order to secure his preferences against hers, but that description depends significantly on his intentions. If, instead of intending to choose for B, A acts out of concern for the risk C takes in treating with B, then he should be described as persuading, with B’s resulting actions taken as an unintended side-effect. Because his actions neither involve nor require covertness about his intentions, he, in Dowding’s terms, acts from reasons he believes he shares with C (Dowding, 2016: 12–13), and so his influence would not be attenuated should C receive a full accounting of A’s intentions. If, on the other hand, his intentions are to subvert what he takes to be B’s will, then we should say he attempts to manipulate B. And the same is true of the simpler case of direct manipulation/persuasion (Dowding, 2016: 12–13).
Dowding requires that reasons be in some sense shared between A and B in order to say that A persuades B, arguing that only when A gives B the reason that motivates A’s own assent to some sentence S does A attempt to persuade B of S (Dowding, 2016: 6). Interpreted strictly, this is too narrow a standard, as it seems to rule out attempts to persuade you via reasons I do not hold, but that I know to be the reasons you take seriously. There are certainly cases where a manipulator communicates reasons chosen solely by their likely influence over B, with no care as to whether they are good or worthwhile reasons for A or B. But also, surely there are cases where A carefully presents an argument in the terms in which B thinks arguments ought to be made, as according to his faith doctrine, preferred legal code or aesthetic sensibilities, for instance. These conjectural reasons may or may not be manipulative, and I see no reason to think they can never count as upright persuasion. Indeed, this is why I think that, of Dowding’s two reliability conditions, it is the second involving A’s intentions that does all the work. Reasons’ status as shared or not may often correlate with their status as manipulative or persuasive, but it should not be thought of as either a necessary or sufficient condition for indicating such.
On my account, the same case is then either manipulation or persuasion depending on A’s intentions about his actions. This is a feature of manipulation not incidental to a particular case. Much manipulation, to be effective, aims to appear as persuasion. By rendering manipulative actions observationally indistinguishable from persuasion, the manipulator makes it more difficult to reveal and challenge the manipulation. Deceit accomplishes this goal by offering some calculatedly misleading information to the subject, such that only a subject positioned to uncover the deception itself can discern manipulation from persuasion. But some manipulation involves no deception, and its target may then be worse off for having no objective grounds that could even theoretically be appealed to to reveal the manipulation. 13 Observational equivalency with various forms of persuasion is to some extent true of all cases of manipulation – its effectiveness often turns on it going undiscovered, and if discovered, being plausibly deniable. To see this, modify the alliance case: A tells B some lie about C, causing B to partner with A rather than C. C confronts A and B, claiming A has lied to B. A has chosen a lie that C is unable to refute. When confronted his denial is plausible, such that B does not revert. Here the lie is instrumental to interfering with B, but the deniability of the lie is essential for maintaining the new alliance against truthful revelation by C. B must believe she has been legitimately persuaded.
So manipulation can be indiscernible (at least obscure and unprovable) to the target and third parties because intent is a necessary condition. Relatedly, A cannot manipulate B accidentally. A can act in ways that serve to change B’s strategic position, information, choice architecture, etc., while not intending particular effects. In such cases, B may still have moral claims against A’s actions as reckless or otherwise insufficiently attentive to B’s interests (much as the republicanism’s gentle giant dominates the townsfolk, but isn’t aware of his effects (Kramer, 2008: 40–50)), but A has not manipulated. This is because manipulation, like coercion and persuasion, works by replacing B’s will with A’s. Persuasion works openly via reasons; coercion operates openly as well, via threats of force or force itself. Manipulation works covertly; A’s influence is intentionally obscure to others. So when A affects B’s will without intending to, we cannot say that A has replaced it with his own, since B’s new choice situation is merely incidental rather than intentional to A’s will.
For example, if in the alliance case, A undermines B and C’s partnership without deception and clearly communicates his intentions to undermine B’s preference to partner with C, the account would not label this manipulation. Such cases are a type of influence on or domination of B that may itself call for theoretical analysis, but that lacks the covert or surreptitious element shared by cases of manipulation.
We have seen that an account of manipulation requiring instances of deceit and/or direct action on a target will be insufficient for several acts that should seem manipulative. We should thus expand our conception of manipulation from involving deceit to involving intentionally covert intentions to interfere with B’s will, pursued through direct or indirect means.
Indirect manipulation
Indirect manipulation is more complicated. The essential feature of this mode of manipulation is that A affects B’s will, not by interacting with B, but by acting on the situation in which B can be expected to operate such that B’s will is predictably affected. Examples include: altering some rule or standard B utilizes, as when A insists we use an opt-in system of voter registration rather than an automatic enrolment system; and altering C’s emotions, beliefs and/or information such that B’s situation changes in ways unconducive to achieving B’s putative will, as in the alliance case.
While direct manipulation comes in a great many varieties corresponding to the many ways in which one’s reason might be subverted, indirect manipulation is cleanly sub-categorizable. The clearest examples of indirect manipulation involve competitive games, where each of us seeks to alter the other’s actions through the manipulation of the other’s position, options and/or beliefs about those things. There are two general ways I can carry out this sort of manipulation. In the first, I work within the rules of the game to undermine your attempts to achieve the game’s objectives – call this strategic manipulation. In the second, I work at a higher level than the game itself by altering, challenging, reinterpreting, disregarding or adding to its rules in ways that undermine what I take to be your will – call this structural manipulation.
The former is the sort of action we associate with superior strategy in complex state-space games like chess or Go, where understanding an opponent’s strategy and strategically signalling about your own is part of good play. So limiting, suggesting or otherwise altering an opponent’s picture of one’s own strategy can be key to victory. Political examples include false troop retreat meant to provoke pursuit and premature attack, or indicating support for a policy so that your opponent invests resources opposing you on that issue over others on which she is stronger. This is not to be confused with ‘forcing an opponent’s hand’, which is how we might describe instances of strategic influence that involve no intentionality over deception or deniability, though will often share this structure.
The second type, structural manipulation, occurs when the manipulator affects a rule – or other structure – change in order to cause their opponent to act from a position that pushes B closer to A’s will. As before, this manner of manipulation may or may not involve deception. When I lie and convince you that left-handed players are always allowed to choose their side of the board, I am utilizing structural manipulation just the same as when I insist, at a point most conducive to my victory, that a heretofore forgotten rule be adhered to. Again, deception is unnecessary. All that is required is that the structures A acts upon influence, govern and/or constitute the framework within which B must act without sure knowledge of A’s intentions. Politically, we can think of examples like gerrymandering electoral districts, or a party leader advocating a new electoral system on the basis of its representative qualities while in truth wishing to secure partisan advantage. 14
Politics is made up of the kind of strategic behaviour covered by indirect strategic manipulation. Trivially, almost all actions I might take change the strategic position of some other agent. But most of those actions will not count as intentional or covert. Nevertheless, examples abound. For instance: candidate A signals support for a controversial position, knowing B will respond by shifting resources into research, policy platform drafting and ad campaigning on the issue, while really intending to disavow the position and launch a push for a different policy during the campaign. The ‘fake out’ or false signal expresses a false intention to B which then secures a preferred action for A.
Note that A’s power to manipulate B is always dependent on both A’s knowledge about B’s intentions/behaviour across varying choice sets, and A having the power to influence which choice(s) B will ultimately face. This is part of the conceptual core of manipulation across the categories I distinguish. In all cases, it requires knowledge about how a target is likely to react to different stimuli, and the power to select stimuli for the target. These properties will be essential to thinking about how to render potential targets resilient to attempts at manipulation.
Structural manipulation is common in political life, possibly because rule changes will tend to be highly useful for the manipulator, or because its impacts are less easily foreseen by subjects and are thus more easily hidden (Riker, 1986: Ch. 6). Further, it is both easy and tempting to offer a false justification for a rule change when the manipulator misleadingly appeals to plausible reasons. One case involves altering voting rules. Here a manipulator who favours the status quo appeals to democratic values in arguing for a supermajority condition on future votes. They may appeal to independent and strong reasons for instituting the new voting rule, but these are irrelevant to their true intent, which is to ensure the body’s decision aligns with their preference, rather than its ex ante putative will.
To illustrate with a trivial example: few people play the board game Monopoly using the full set of its official rules. One often forgotten rule stipulates that when landing on any unowned property a player must purchase it or see it auctioned to all players. If several moves into a game A invokes the auctions rule on B, insisting they employ it for the remainder, where A does so not due to an epiphany about the forgotten rule, but out of a strategic sense that it would impede B, then A has manipulated B, and his protestations that ‘the rules state …’ are really an attempt to create plausibility in denying his manipulative intentions. That A insists on the correct rules means that his offered justification is a reason for enforcement, but it is not his reason; his reason is purely strategic. The fact that the written rules have not changed in this case, while they do in the former, is immaterial. In each case, the de facto rules are altered while A maintains deniability about his true reasons.
Indirect manipulation has two functions within our account. First, it shows the insufficiency of Sunstein’s conception that says manipulation is any effort to influence agents’ choices that ‘does not sufficiently engage or appeal to their capacity for reflection and deliberation’ (Sunstein, 2016: 82). In the simple example, I can manipulate the path you take by placing rougher or easier terrain in your way, precisely because I know the kind of reflection and deliberation you engage in over hiking paths. But I do not thereby remove or reduce your reflective and deliberative engagement with your choices. Indeed, the more I know about your tendencies, principles, thoughts, decision rules and the like, the more I am able to confidently gear the choices you face to ensure that you both deliberate on them and make decisions I prefer. Given the power to set the choices you face, some sufficient mix of knowledge about your psychology and of formal work on agenda setting or structural control gives me more or less complete choice over the choices you make, within the domain of choices you would make.
In the limit this is the full and terrible power of manipulation: to ex ante select your decisions for you, while limiting your sense that you have not been the author of your own will. Of course, knowledge about decision-making and the powers to set agendas and structures are very rarely so complete as to allow such an extraordinary level of interference, but much the same is true about coercion. Our questions are about the degree of knowledge and power others often have, and what it means to wield those things to influence others’ choices.
In Sunstein’s account then, either ‘sufficiency’ stands as a placeholder for a begged normative account of reflection and deliberation capable of making sense of indirect influence, or he simply does not consider pre-deliberative manipulation. In either case, there is much left wanting in that work’s ability to deal with cases.
Second, indirect manipulation captures important insights that liberals have been slow to adopt from a broader literature on structural power (i.e. Hayward, 2000; Lovett, 2010; Lukes, 1986; Young, 1990). The normative evaluation of the use of indirect manipulation is a subject for further work, but it is clear from existing normative treatments of neighbouring concepts like structural power that liberal-democratic thought needs language for taking these issues seriously. However, note that this conceptual language is sufficiently broad to capture the machinery of structural power, but in a more precise manner than those scholars have used the term manipulation, as when Iris Young talks about the ‘manipulat[ion] of language and symbols’ in order to structure meanings to the advantage of the powerful (Young, 1990: 87).
Distinctions in hand, we are now in a position to come back to the precise definition of manipulation that we began with:
An act of manipulation is any intentional attempt by an agent (A) to cause another agent (B) to will/prefer/intend/act other than what A takes B’s will, preference or intention to be, where A does so utilizing methods that obscure and render deniable A’s intentions vis à vis B.
A can accomplish this by acting directly on B in a variety of ways, some utilizing deception but not necessarily, or through acting indirectly on B either by influencing B’s strategic position or by changing the rules that structure B’s choices.
Note that both B’s objective ex ante will and the direction of A’s influence on B are irrelevant in discerning cases of manipulation. The latter is simple: A manipulates B when he covertly influences B into accepting his will, but also when he covertly influences B into accepting any other position/belief/intention; whether the result is agreement with A’s own ideal point is immaterial. On the former, the irrelevant feature is the objective nature of B’s will, since our concern is with A’s beliefs about B’s ex ante will rather than its actual nature. A need not have any very precise or confident judgment about B’s ex ante will. He need only seek to influence her to pursue an option closer to his ideal for her decision, or to pursue such an option with a greater likelihood. This captures cases where A has no clear picture about B’s intentions, but removes tempting (to B) alternatives to increase the likelihood that A’s preferred option is pursued. The account then also covers cases of manipulation of targets that have not yet formed relevant wills, or who even do not yet exist. By publishing a deceitful textbook, for instance, A can hope to thereby manipulate readers not yet born. Notice how a textbook full of falsehoods that A did not know were falsehoods on publication, with which he did not intend to mislead readers, is on this conception not manipulation, while the former is.
By extension, the account includes cases where A has a mistaken belief about B’s ex ante will. This is because the core concept the conception seeks to capture is that A makes some decision for B, which B might have otherwise made for themself. It is the covert shifting of the locus of choice from B to A that marks out instances of manipulation. Of course, A could manipulate B to pursue an option which, unknown to A, B was already set to pursue. Here we cannot say that A has necessarily caused B’s choice; rather, B’s choice is causally overdetermined by both her ex ante will and by A’s manipulation. So, in fact A has sufficiently but not necessarily caused B’s decision. This may seem a small point, but it means that alternative decisions B might have made are removed or made much more costly to the point where should B wish to select a different option, she would be sufficiently influenced by A not to.
Covertness
The covertness condition is controversial, and requires some elaboration and defence. First, to clarify the condition we need to decompose covertness into both intentional deception and sought plausible deniability – each in reference to A’s intentions over B’s will/beliefs/decisions. In each case, A looks to leave others either unaware of his real actions and intentions, or unable to counter his denials should they confront him. Deception is straightforward: A directly or indirectly misleads B about the nature of his true intentions, often by presenting information that is simply false.
Plausible deniability admits of more techniques, varying in how deniability is achieved and degree of plausibility. In a general sense, deniability means the manipulator seeks to be able to deny their intentions were manipulative, or that they were involved in the action in question, and the manipulated will have no sufficiently immediate objective recourse for showing that claim to be false. 15 Insincere justification in deliberation presents a relatively clear example of such manipulation. 16 Here a deliberator presents a justification as compelling for themselves, which seems to others a plausible justification, but by stipulation they select it not because they believe it ought to compel the other, but because they eschew reasoning with the target, selecting whatever reasons they believe will cause the target to assent (Dowding, 2016: 7). Here he establishes intentional deniability by presenting his reasons as good reasons, as though they were selected for their content-dependent properties and not their effects on B. Intentionality thus features twice in manipulation, since A must intend both that B’s will be undermined, and that his intentional causal role be covert. Cases where an agent is causally effective over another’s will but does not intend to hide their interference are then properly called something other than manipulative.
We have seen and dismissed arguments over the first (deceptive) aspect, and while philosophers have been indirect or vague in addressing the second aspect, it seems nonetheless controversial. Moti Gorin’s chapter is an illustrative example (Gorin, 2014b). He presents the case of a professional rivalry between A and B, whom A happens to know is a recovering alcoholic. A seeks to expose B to alcohol in various ways: ensuring it is nearby, offering her drinks and speaking to her evocatively about the joys and benefits of alcohol. A does so making ‘no attempt to mask his intentions’ from B. When B eventually breaks her sobriety at A’s tempting, her work performance suffers and A is offered promotion ahead of B. According to Gorin, ‘[A] has manipulated [B] by engaging her compulsion to drink alcohol. And [B’s] awareness of [A’s] intentions does not undermine the intuition that this is a genuine case of manipulation… [and therefore] the Deception-Based View is false’ (Gorin, 2014b: 81).
But the alcoholic case does not establish Gorin’s strong claim. He writes that A ‘makes no attempt to mask his intentions’ and moves directly to the fact of ‘B’s awareness of [A’s] intentions’ (Gorin, 2014b: 81). However, there are many cases where A makes no direct effort at concealment, while nonetheless intending his intentions not become known. We should be careful here to specify just what is required with respect to intentions, specifically B’s knowledge about A’s intentions, A’s intentions about B’s knowledge about A’s intentions, and A’s intentions about others’ knowledge about A’s intentions. Covertness as either deception or plausible deniability calls us to take each seriously.
For the sake of argument, though, we should proceed with an alcoholic example where A announces openly and repeatedly to B that he seeks to tempt her into failure. Is this sort of case manipulation? I think not, and offer two arguments as to why.
First, an argument about intuitions. Gorin’s example of the alcoholic leverages a special kind of vulnerability stemming from a tenuously broken dependency on alcohol. This is a poor choice of case to prime an unalloyed intuition about covertness because it builds in an orthogonal set of intuitions involving an already compromised condition for B’s autonomy.
Imagine instead a narrower case where A influences B’s will. A does not deceive B; A influences B in a way or direction which A knows B would object to and resist, and that by his own lights she would have good reason to object to. A is clear with B that his influence over her is intentional and objectionable; he does not hide his interference or his purpose. Like a cartoon villain, A twirls his moustache and laughs, declaring: ‘And there’s nothing you can do to stop me!’. Intuitively, to my mind, his straightforward approach to influencing B places this case outside of the category of manipulation and into some other type of interference: A controls B, or A forces B. A has the sort of uncontrolled power over B that leaves him with no reason to hide or mask his interference – neither B nor anyone else who B might appeal to could stop A’s interference. This is quite unlike the canonical examples of manipulation either in philosophy or literature. In Othello, for instance, Iago’s plot to tempt Othello into suspicion and eventual rage towards Desdemona would have been undone had Othello known Iago’s intentions.
Instead we might colloquially distinguish the brash villain case by saying that ‘A has forced B’s hand’ in her decision, since he has openly given her no other choice. But other cases of more straightforward manipulation we have discussed have involved A winnowing down B’s options so that his preference is selected, so we need to do more to sharpen this conceptual line. The difference is that A has been influenced with or without some significant intentional subtlety, and thus the brash villain is engaging in something more like coercion than manipulation. The cost of thinking otherwise is blowing up the category of manipulation beyond useful application.
A tempting implication would be that manipulation requires the ignorance of B about A’s intentions, meaning any attempted manipulation would be resisted or foiled by B’s discovery of A’s true intent. This will often be the case, but part of the point of plausible deniability is to broaden the scope to include third parties. Here, A interferes with B in a situation where some other agent or set of agents C could effectively stop him doing so iff C were convinced that A were intentionally interfering, but where A has sought to make that realization impossible or unlikely by tailoring his actions to deprive B of effective communication with C, or of sufficient evidence to present to C to elicit help.
Return to the alliance case of A, who undermines B’s alliance with C by lying to C about B’s trustworthiness in order to ensure that B must instead ally with A. Here B can know A’s true intentions, but is nonetheless still influenced to act according to A’s will because she cannot correct A’s lie to C – her trustworthiness has been undermined such that she cannot credibly argue she is trustworthy – but needs a partner nonetheless. This case is emblematic of a wide class of political manipulation, where the target understands or plausibly suspects A’s true intentions, but A has established sufficient plausible deniability vis á vis relevant parties that B is unable to re-establish her credibility to an extent that would enable her to resist A’s influence.
Consider a more real-world example to see the importance of this category. A legislator (A), in parliamentary debate over a bill to legalize same-sex marriage, argues against the bill because its effects would fall upon innocent children. This reason is, let us stipulate, both false and known by A to be false, but like all reasons it affects distinct receivers differently. Let B represent a receiving constituency who believe correctly that the outcome of the bill would not involve harm to children and are thus not persuaded by the argument. C, however, is a decisive coalition who are not aware of the truth of the matter, but are swayed by the plausible suggestions that children will be harmed. Absent this argument, C would have voted with B, but is deceived into voting instead with A, thus ensuring the bill’s defeat. C doesn’t believe B when their representative correctly tells them A is being deceptive, instead accepting A’s denials because the reason is plausible and B has no ready ability to contradict A to C’s satisfaction. This category of manipulation is especially common because it requires less of A, who does not need to craft a careful manipulation of A, but only needs to find a vulnerable constituency and a difficult to refute but plausible reason.
The second and more broad response to critics of covertness is to acknowledge that our basic linguistic or conceptual intuitions may diverge. There are hard to settle questions about just where to draw the discriminating line through edge cases and fuzzy category distinctions about which we may have unsettled, unclear or simply departing intuitions. I have argued for my intuitions about the concepts involved, but these may not be universally felt. Even when suitably refined via careful analysis of cases and thought experiments, we may simply not share relevant intuitions. This leaves us with two options. We can craft separate conceptions of manipulation distinguished first by whether or not they eschew a covertness condition, or we could attempt to give arguments on other grounds that a covertness criterion is a good or negative inclusion. The first is a credible approach if intuitions simply diverge, but the account will be on stronger footing if I offer something of the latter as well. To that end, consider a normative case for a covertness condition. On this basis, covertness should be part of manipulation not (merely) because it fits with conceptual intuitions we may hold, but because it allows us to constructively grapple with a set of political practices that are normatively distinctive, in the sense that they potentially wrong subjects in a distinctive set of ways such that accounting for those wrongs is aided by the conceptual fence of the covertness condition. The wrong of these practices is unlike the wrong of practices not characterized by covertness to a degree sufficient for the conceptual distinction to be normative in application.
The defence of the conceptual distinction between those things that a covertness condition rules out of manipulation and those things that Gorin’s conception would not, then, is that his conception groups together a set of practices that are unlike one another in the kind and degree of wrongs they present for their victims. For instance, when A ‘forces B’s hand’ in some interaction such that B cannot but choose A’s preference, but A has been forthright in how his actions caused that situation for B, then B’s normative situation is quite unlike the situation where A has manipulated B such that B chooses A’s preference but is unaware of A’s causal role in that choice.
Group manipulation
Thus far I have been primarily concerned with individual political manipulation, which analytically describes a species of interference with others’ political wills and associated conduct. But the political focus of this conception of manipulation pushes us to also account for the manipulation of politically oriented group agents. Call this level of manipulation group manipulation. As we will see, like individuals, groups can be manipulated directly or indirectly, but will pose a distinct set of conceptual and normative puzzles.
By group agents, I have in mind a wide set of bodies, committees, legislatures, individuals, corporations and others that make political decisions as, on behalf of or significantly for collections of people. I mean this concept to be capacious, such that the Catholic Pope, the US President, the board of the Coca-Cola Co., the UK House of Lords and the country of South Africa should all properly count as group agents in some contexts. Of course some of these examples are individual agents, such that we can speak of their being manipulated in various individual/personal ways, but since they also make decisions on behalf of the group they lead or represent, there are some contexts in which they count as standing in for that group, and thus as a group agent in this broad sense. So President Obama being individually manipulated concerning his daughter’s curfew, say, would not count as group manipulation despite him having acted as the group in various contexts. On the other hand, when he is individually manipulated in negotiations at the G20 summit he is both individually manipulated as Barack Obama, and a subject of group manipulation as the US President.
There are thus two central possibilities for group manipulation: manipulation of a group itself, and manipulation of key or sufficient members of the group so as to achieve manipulation of the group as a whole. Both have as their aim the replacement of some putative group-level will with the manipulator’s own, and each can be accomplished through any of the methods of manipulation covered thus far. Each presents distinct normative questions from individual manipulation simpliciter, and group manipulation involving multiple individuals cooperating through some sort of social choice mechanism complicates the conceptual picture by multiplying potential targets and strategies for interference.
In the first instance, we can imagine groups being subject to the kind of strategic manipulation that Gibbard (1973), Satterthwaite (1975) and Riker (1982) are concerned with in certain preferential voting systems. In such cases, a group member seeks to manipulate the social choice the group aims at by submitting a false or strategic preference profile. Supposing that at least some groups are constituted with voting procedures expecting or requiring honest preference profiles, strategic voting is then manipulatory insofar as it pretends to honesty.
In the second instance, group decisions are affected through the manipulation of certain or sufficiently many individuals engaged in some manner of group decision-making. There are in this case two further possibilities. First, the manipulator may act (manipulatively or otherwise) on individuals in order to undermine the putative will of the group to which the individuals belong and thereby manipulate the group agent. Second, the manipulator may not in fact intend any group-level manipulation, but may nonetheless cause similar effects through the manipulation of individuals. We must consider whether and to what degree the manipulator is culpable when her individually manipulatory actions result in or risk group level consequences. 17 An example of this kind of phenomenon, where individual political manipulation rises to the level of group manipulation, is when a significant number of registered voters in an upcoming election are given misleading information about their polling places, such that on election day they find themselves unable to cast a ballot. Each of them has been individually manipulated, but this rises to the level of group manipulation if, for instance, the effect of their suppression is unevenly distributed across candidates, such that their individual manipulation causes or risks a candidate to win who would have lost in a counterfactual world where they were not interfered with.
It might be objected here that the group is not like the individual; that while individuals have wills, intentions, identities and thought processes that ought to be considered theirs in one way or another, these concepts do not map cleanly onto groups. The critic holds that individuals are agentive such that we ought to be concerned about their self-authorship of an identity continued across time, while groups, even if we think of them as making single decisions about some matters, are not agents in this or other relevant ways. I am not concerned here to say very much about the ontological status of groups. I take it as uncontroversial that social choices do exist – that collections of people make choices that are, descriptively at least, both theirs and choices. The nature of those choices and the conditions for being a group such that those choices should be considered agentive are important questions, addressed by List and Pettit (2011), List (2018) and Korsgaard (2009), among others. Those questions will be important in further work on the moral status of manipulation – the nature and extent of something’s agency will factor into the normative assessment of actions that curtail or disrupt that agency – but do not bear on the conceptual work at hand. It suffices that insofar as social choices exist, we can think of those choices being influenced in the manipulative ways discussed thus far, and let ‘group agent’ stand as a thin term requiring only that premise.
The threat of manipulation is multiplied for group agents, who may be manipulated by other agents (group or individual), but also by their constitutive member agents. The first happens through mechanisms already discussed, while the second involves any of the several voting strategies that social choice theorists describe as manipulative (Saari, 2001: 22–23; Taylor, 2005: Ch. 2), but also deceptive argument in public deliberation/justification, and votes for rule changes/invocations that surreptitiously push for particular favoured social choices. In each case, one or more non-decisive influential member of a group seeks an additional share of influence over the collective decision of the group. For in order to be a group rather than a latent team (Simpson, 2016: 40), a mob or a mere collection of random people (French, 1984: 5), there must be a norm- or rule-governed procedure for arriving at a social choice that effectively and consistently apportions influence. In ideal democratic groups, that procedure specifies an egalitarian distribution of individuated influence: each citizen has one vote’s worth of influence in any decision. Anyone seeking to vote twice thus seeks either a general power-grab (if they are open about it) or to manipulate the group (if they are deceptive or surreptitious).
Two typifying examples are worth discussing further. First, deceptive justification involves the public presentation – whether deliberatively, in a judicial setting or written into a piece of policy/legislation itself – of some reason in favour of a policy when that reason is deceitful, misleading or surreptitious vis à vis the presenter’s intentions or beliefs about the policy itself. 18 An example might be the invocation of a misleading report during debate on a bill to regulate a toxic substance. If A knows the report to be false or otherwise misleading, yet invokes it to muddy the waters over the bill, he attempts to manipulate the legislative body while obscuring his intentions with the plausible deniability of a seemingly legitimate report.
Second, in polities with regional constituency representation, redistricting as populations grow and shift is a recurring political task. Manipulation through that process is referred to as gerrymandering and is a key example of structural manipulation on a group. Here new district lines are drawn so as to isolate opponents’ supporters into fewer districts and/or spread them thinly over as many as possible, so that the gerrymandering party secures more final representatives than opponents. It is only very rarely announced publicly that this is the intention of officials engaged in redistricting, and the geometry of districts does not make obvious how much change from one map to another is likely to indicate partisan intent (Altman, 1998; King and Browning, 1987).
Gerrymandering is an especially difficult to parse instance of manipulation, because it is not immediately clear what will is being affected. The problem is in the role of representation against proportionality in democratic institutional design. If proportionality were the only principle then districts would be unnecessary, or mere conveniences for constituency services. Most democracies employ some principle – often un- or under-specified – of local representation pulling against proportionality. 19 In effect, the principle says that certain collections of citizens, whether due to an identity of interests, geographical proximity or some other reason, should jointly decide whom to elect, and that the losers and winner alike in that decision should be represented by that choice. Such principles are irreducibly normative in character and are produced and re-affirmed through the continued operation of the wide set of decision-making mechanisms they regulate.
Gerrymandering, then, is a drawing of district lines that runs afoul of the substantive requirements of the operative representation principle. 20 When that redrawing is carried out surreptitiously or deceptively (i.e. when it is done with the intent, but without recognition, of undermining the representation principle), it is manipulative of the group’s will over representation and by extension their will over the resulting decision(s).
The roots of our concern over group manipulation will not be identical to those related to individual manipulation. While the latter reduce to moral concerns about autonomy, self-authorship, liberty or the like, our moral assessment of group manipulation will turn on the sort of group in question and the type of decision-making authority attaching to it. For now we need not settle questions on the precise extent of collective rights to self-determination of various groups, or the scope of matters that a group ought to be sovereign over, or the exact shape of members’ rights to influence and collective control, and so on. Rather, so long as we think that some political communities, nations, states, churches, corporations, etc. have a sufficient level of legitimacy such that they should be protected from interference in their decision-making over matters they have rights both to decide for themselves, and in the manner of their own choosing, then we face the problem of external and internal manipulation of those decisions. 21 Of course, the reasons we have for thinking a given group is the legitimate decision-maker on some issue of their concern are not incidental to the moral assessment of interference in those decisions, but the general concern need not be attentive to those differences.
Conclusion
The conceptual apparatus describing manipulation is meant to help analyse a distinctive type of interference. A maximally elegant ordinary language definition of such a term would capture the shades of linguistic meaning consistently ascribed to it. It would not be susceptible to reductios when intuitively non-manipulative interactions are labelled manipulative, because our linguistic usage is not expected to map perfectly to our conceptual intuitions. My account is not meant to capture usage in that way. We need not, for instance, encompass the sense used when instructing a student to manipulate a focusing knob on a telescope. The account is thus pragmatic in nature: a conception judged by its usefulness in distinguishing between political phenomena that we think ought be treated distinctly from other sorts of influence. The flaws in rejected accounts are then not down to lack of fit with linguistic intuition, but because cases we have independent reasons for treating together are instead separated, or vice versa.
A relevant test of this sort of conception requires its application to cases. Will it, as Goodin hopes for his conception, give us ‘foundations for empirical political studies with a sharper evaluative focus’? (Goodin, 1980: ix). A critical answer to this question is worth cutting off by way of conclusion. Manipulation, on my account, requires that A hold some belief about B’s will and that A intentionally work to undermine B’s pursuit thereof. One worry is that since assessing manipulation requires evidence or otherwise grounded beliefs about A’s intentions – his internal mental states – the account offers too little empirical traction for those who wish to stop manipulators, protect the manipulated or sort out which is which. One might worry it is not meaningfully operationalizable.
But this is no serious problem for an account of a practice which itself aims to be observationally equivalent with upright or mundane actions. Manipulators seek to camouflage their puppeteering as unrelated or merely persuasive interactions. The well-manipulated should go about their lives thinking themselves the authors of their various decisions, while in fact it is the manipulator who has chosen their actions. It is thus unsurprising and indeed unavoidable that targets and third parties will not have confident assessments absent confessions of intentions. Operationalization remains a difficult question, but not an unfamiliar one from legal work on crimes requiring intentionality. In any case it would be a mistake to craft an account meant only for easily empirically verifiable manipulation.
I will not speculate about empirical hypothesis tests and data collection, but there is good reason – evidenced by the cases discussed throughout – to think that I have described ‘a concept … that would help us track, measure, and evaluate an important set of phenomena in the social and political world’ (Rehfeld, 2018: 216). The question of evidentiary standards for separating cases absent full knowledge of intent is not a conceptual but a normative question. Any standard we set will specify levels of sensitivity and errors in both false-positive and false-negative directions. How those errors ought to be distributed is an irreducibly normative question. Answering it would beg commitments to fuller political theories of justice, democracy and ethics than I have sought to offer or appeal to here. More complete normative analysis in this and other open questions is the subject for further work.
And further, those phenomena are of special normative political interest. Indeed, we should care about the distinctions this conception lays out, since realism about democratic decision-making requires clear-eyed analysis of the ways which political decision-making might be undermined in its normative promises. Part of the difficulty that plagues thinking about those promises is a fuzziness about not only the complexity and diversity in our political psychologies (Mendelberg, 2002) but also what is possible from social decisions in the first place (Arrow, 1950; Riker, 1982). It is for future work to flesh out the normative analysis of these distinctions and apply them to cases. So long as the present account plausibly describes intuitive differences, then it should stand as the basis for that project.
One feature of that normative analysis worth advertising here is that a key characteristic of politically manipulatory actions – whether on groups or individuals – is the attempt to subvert others’ decisiveness by gaining influence while obscuring that fact. Individually, when A manipulates B, A then has replaced B as the decisive vote over B’s decision. In group contexts, A either gains for himself an outsized impact on the group’s decision-making process – thereby shifting the decisive coalition towards himself – or he is simply decisive, thereby claiming for himself the group’s decision (depending on the magnitude of his interference). In either case, his decision-making influence is increased without the full knowledge of his impact by others. But to describe this normatively we would appeal to a normative theory specifying principles about the significance and distribution of decisiveness to say that he has violated relevant principles in subverting that distribution. This is precisely the worry that has inspired decades of work in social choice and democratic theory (Ingham, 2014, 2018; Mackie, 2004; Patty and Penn, 2014; Pettit, 2012); that the normative promise of equal individualized influence on social choices is a chimera, or too easily usurped by manipulative participants in voting or institutional design. While the normative importance of autonomous individual decision-making is perhaps the most thoroughly studied ethical issue, the normative weight of the principles that groups evoke to underline the worth and legitimacy of their decisions is comparatively under-pursued. Sorting out the normative issues suggested by social choice literature is thus a problem that receives too little attention, and this conception helps us to collect common instances with an eye towards their normative analysis in future work.
Footnotes
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
