Freedom and its unavoidable trade-off

In the debate on how we ought to define political freedom, some definitions are criticized for implying that no one can ever be free to perform any action. In this paper, I show how the possibility of freedom depends on a definition that finds an appropriate balance between absence of interference and protection against interference. To assess the possibility of different conceptions of freedom, I consider the trade-offs they make between these two dimensions. I find that pure negative freedom is clearly possible. Repub-lican freedom might also be possible, though its protection requirement is too vague for a definitive verdict. Finally, the recently proposed ‘freedom as independence’ is impossible since it is an attempt to avoid the unavoidable trade-off.

In this paper, I clarify what makes a political conception of freedom impossible.By drawing on List (2004List ( , 2006) ) and Pettit (2001a), I identify two dimensions of freedom: scope and robustness.The former concerns absence of interference, while the latter concerns protection against interference.A definition of freedom that takes any kind of interference to make you unfree has maximal scope.The more kinds of interference not considered to make an agent unfree, the lesser is the scope of that conception of freedom.Robustness is indicated by the extent to which a definition requires that the interference specified by its scope does not occur in different socially possible worlds.A socially possible world is defined by a particular combination of preference orderings across all individuals within a society (List, 2006: 212).A definition of freedom with maximal robustness requires the absence of interference in all socially possible worlds.Maximal robustness therefore demands protection against interference to the extent that you will not be prevented from performing a certain action regardless of how others are disposed toward you.Minimal robustness, on the other hand, requires only that you do not experience interference in the actual world. 1  I understand freedom as a social value that two or more individuals can enjoy simultaneously.A central argument in the paper is that such a definition of freedom must be a trade-off between scope and robustness.This is evident in the observation that protecting agents against interference involves restricting their ability to interfere with each other, and such restriction is itself a kind of interference.The more types of interference we do not take to make the agent interfered with unfree, the more protection can our definition demand.Conversely, the more kinds of interference we understand to make the agent unfree, the less protection can our definition require.
Scope and robustness are thus inversely related, but only roughly so.An asymmetry between the two dimension appears when we observe that while maximal scope is compatible with (no more than) minimal robustness, maximal robustness is compatible with no scope at all.Freedom with maximal scope is possible but only if combined with minimal robustness.Individuals in a society can be free from each other's interference in a particular possible world.Taking freedom to require such non-interference also in other possible worlds, however, involves protecting individuals against each other's interference, and such protection necessarily involves interference.To combine maximal scope with more than minimal robustness is therefore to understand freedom to require both non-interference and interference.An increase in robustness must therefore come with a reduction in scope.Suppose we give freedom more than minimal robustness by saying it requires protection against violations of property rights.We then cannot count the interference of protective measures such as high electric fences around estates or patrolling guards as a source of anyone's unfreedom.By thus treating some kinds of interference not as a source of unfreedom, we reduce the scope of our definition of freedom for the sake of its robustness.
Maximal robustness, however, makes any definition of freedom impossible.People can never be entirely safe from one another's interference, as no kind of interference can be protected against to the extent that it occurs in no socially possible world.By supposing freedom to be possible for two or more individuals, there is no way of making sure that you are protected against interference no matter how others are disposed toward you.A definition of freedom requiring such protection is therefore impossible regardless of how much interference it is compatible with.
After developing this two-dimensional scope-robustness framework, I consider how different conceptions of freedom fit within it.Pure negative freedom is clearly possible, as it has minimal robustness and maximal scope. 2 The assessment of republican freedom is more complicated, as its robustness requirement is vaguely specified.But I suggest it can take a moderate position on both dimensions that allows for the possibility of freedom.Freedom as independence lies at the maximum end of both dimensions.This attempt to avoid the trade-off between scope and robustness results in impossibility.
1 Socially possible worlds form a subset of all biological possible worlds, which are a subset of all physically possible worlds, which are a subset of all logically possible worlds (List, 2006: 203-205).Each set is defined by positive-as opposed to normative-laws.Positive social laws are regularities in human behaviour, and List gives the example of the law of supply and demand. 2Pure negative freedom is often referred to as 'liberal freedom', especially when contrasted with republican freedom (Carter, 2008;Kramer, 2008;List & Valentini, 2016;Pettit, 2008a).I use the term 'pure negative freedom' because it refers to a specific conception of liberal freedom as non-interference.
In response, defenders might argue that while full freedom as independence is impossible, it can still be realized to a certain degree.I show, however, that freedom as independence cannot be achieved to any extent.We can try to measure and promote pure negative freedom and republican freedom, but freedom as independence falls out of the picture.

| THE TRADE-OFF
The debate over political freedom between proponents of pure negative freedom (Carter, 1999(Carter, , 2008;;Kramer, 2003Kramer, , 2008)), republican freedom (Pettit, 2008a(Pettit, , 2011(Pettit, , 2012: ch. 1;: ch. 1;Skinner, 2008), and, more recently, freedom as independence (List & Valentini, 2016) focuses on the extent to which we should understand freedom to conflict with interference and to what degree it requires protection against interference.It is, in other words, about the scope and robustness of freedom.While scope concerns what kinds of interference should count as sources of unfreedom, robustness is about the extent to which freedom should be understood to require protection against interference.
On the pure negative account, freedom is the absence of physical prevention, and you are made unfree only by other agents making it physically impossible for you to perform some action.As all kinds of prevention are taken to make you unfree, there is no need for evaluating the significance or permissibility of different types of prevention.This is what makes this conception of freedom purely negative.It has no concern with intention and autonomous motivation, which is associated with positive freedom (Steiner, 1994: 9-12, 17-21).
On the pure negative account, a prevention to perform some action, x, makes you unfree independently of any desire you might have to perform x.Pure negative freedom therefore avoids the counter-intuitive implication of 'liberation by adaption'.That is, an individual, A, cannot make herself free by adapting her preferences in response to what some other individual, B, allows her to do.Berlin (2002: 32, 169) points out this problematic implication of Hobbes's (1999: 81) non-frustration view of freedom, according to which you are not made unfree as long as no one prevents you from performing an action you want to perform.To avoid the liberation by adaption implication, Berlin proposes an account of freedom that takes B to make A unfree to do x by preventing her from doing x regardless of whether A wants to do x.Following Berlin, most freedom theorists now hold accounts of freedom that avoid this implication.
Pettit has repeatedly challenged freedom as non-interference generally (Pettit, 1997(Pettit, : ch. 2, 2001a(Pettit, , 2011(Pettit, , 2012: ch. 1): ch. 1), and pure negative freedom particularly (Pettit, 2008a).But he accepts that freedom cannot just require that you can do what you want to do.To avoid the liberation by adaption implication, we must also take freedom to concern what you do not want to do.Pettit (2001a: 5) therefore takes freedom to require that preferences be 'content-independently decisive'.That is, whatever A's preferences might be-whatever the content of A's preferences-A is made unfree if B prevents her from satisfying them.No matter how A's preferences over x, y, and z are ordered, freedom should require that A's preferences be decisive.
The problem with any account of freedom as non-interference, in Pettit's (2001a: 6-7) view, is its failure to capture that freedom must also require that preferences be 'context-independently decisive'.We cannot consider A to be free merely because B allows her to satisfy any preference, as A could then be free simply because B happens to be favourably disposed toward her.Freedom cannot depend on others' 'grace and favour', Pettit (2001a: 6) says.If it is, then we get the counterintuitive implication of 'liberation by ingratiation' (Pettit, 2001a(Pettit, : 12, 2008a(Pettit, : 216, 2011: 704-711): 704-711).That is, A can then make herself free by currying favour with B so that B will let her perform some action.We can only avoid this implication, Pettit says, by understanding freedom to require that preferences be decisive independently of social context.A particular social context, or a socially possible world, is the combination of preference orderings across all individuals in a society.A's preferences are context-independently decisive to the extent that they are satisfied regardless of A's social context.
Pettit thus identifies the two dimensions of freedom I focus on in this paper.The 'scope' dimension indicates the extent to which freedom requires that preferences be content-independently decisive, and the 'robustness' dimension requires that they be context-independently decisive.Pettit's (2001a: 8) own republican conception of freedom as non-domination demands that an agent's preferences be both content-independently and context-independently decisive.But he leaves it unclear to what extent republican freedom requires content-and context-independence. 'It will be a matter of judgement or stipulation', he says, 'as what degree of dependency on content or context means that the agent does not enjoy freedom' (Pettit, 2001a: 7).I return to the scope and robustness of Pettit's republican conception of freedom in Section 4. List (2004) demonstrates a limitation to how a definition of freedom can be placed within a two-dimensional scope-robustness picture.By defining freedom with maximal robustness, we make it incompatible with the Pareto principle, which says that if all individuals in a group, or society, prefer an alternative, x, to another alternative, y, then the group must prefer x to y.3 Whether the social outcome they produce satisfies the Pareto principle will depend on the content of their preferences, which is how we have defined context.Avoiding this result therefore requires weakening the robustness condition in the definition of freedom.This is actually an understatement, as the problem of full context-independence is not just that it makes freedom impossible to combine with some other value.It also makes freedom itself impossible.Freedom with maximal robustness cannot even be combined with minimal scope, as that would mean A's freedom depends on A being able to satisfy a preference for at least one social outcome, x, regardless of everyone else's preference concerning x.Since there will always be a possible world in which someone satisfies a preference for not x, A can never be free on such a definition.No one can ever be free to satisfy any preference, since there is no way of ensuring that an agent can satisfy any preference regardless of the agent's social context.Freedom therefore cannot be the ability to satisfy one's preference under all possible social contexts. 4e shall see, however, that maximal scope can be combined with minimal robustness.But we can nonetheless say that the two dimensions are roughly inversely related, so that we can only enhance robustness by reducing scope and vice versa.Freedom with more than minimal robustness requires protecting individuals against each other's interference.This protection itself means preventing them from satisfying preferences for preventing others from preventing their preferences.Greater robustness therefore means specifying a set of preferences the satisfaction of which freedom does not require.Robustness comes at the expense of scope.
Further, the greater the scope of a definition of freedom, the more dependent the definition becomes on context.That is, the more pairs of options the definition requires that an individual be decisive over, the more dependent on social context does freedom become.The set of possible worlds in which A can satisfy a preference for x is larger than the set of worlds in which she can satisfy preferences for both x and y, which in turn is larger than the set of worlds in which she can satisfy a preference for x, y, and z, and so on.Scope thus comes at the cost of robustness.

| FREEDOM IN ACTION
In what follows, I discuss freedom in action rather than in preference decisiveness, since this is the principal focus of the literature on political freedom.These two ways of understanding freedom come apart in the sense that whether or not your preference is decisive depends not just on whether you can perform some action but also on the consequences of that action (Sen, 2017: 446).But the two-dimensional view I have sketched can nonetheless be easily translated so as to apply to freedom in action.
The crucial similarity is that also freedom in action depends on social context.Just as the decisiveness of an individual's preferences depends on social context, an individual's ability to perform an action, x, depends on others letting her do x.We cannot say that an agent, A's, freedom to perform some action, x, depends on A being able to do x regardless of how other agents are disposed to behave toward A. We thus see that the trade-off between scope and robustness also applies to freedom in action.On either view of freedom, robustness cannot be maximal.
The scope of a conception of freedom expresses the various ways in which someone is said to be made unfree, according to that conception. 5We can measure the scope in various ways, and one way would be to include all kinds of prevention.Freedom with maximal scope would then imply that natural obstacles, as well as prevention imposed by an agent, make you unfree (Sen, 1999).But I shall take scope to refer only to the way in which preventions imposed by other agents are said to make you unfree.This is because the literature I primarily engage with here treats freedom as a social relation and this measure of scope is therefore better suited for comparing the different positions in this literature.As I understand it, then, the scope of a conception of freedom is greater the more kinds of prevention imposed by another agent are understood to make the agent interfered with unfree. 6cope therefore concerns the y variable in MacCallum's (1967) triadic formula of freedom, where x refers to an agent, y to a 'preventive condition', such as a constraint, restriction, or interference, and z to an action.An agent, x, is free, or unfree, from a constraint, y, to perform some action, z.The wider the range of preventions considered sources of unfreedom, the greater is the scope of that definition of freedom.I associate scope with the y variable because the literature I engage with, particularly on republican freedom, focuses on the type of constraint with which freedom is compatible.On an alternative view, scope concerns the actions an agent must be unprevented from performing.The scope of freedom would then be greater the more actions, or types of action, the prevention of which counts as an unfreedom.On this alternative view, scope is associated with the z variable.
To clarify the matter of prevention, or interference, I take B to prevent A from doing x regardless of whether A otherwise would have done x.If B prevents A from doing x, B interferes with A even if A in any case would not have done x.I count B's act of making x more costly and therefore less attractive and accessible to A as interference only if B thereby makes A unable to perform some action.For example, A might still be able to do x but unable to do other actions conjunctively with doing x because of the cost B imposes on her doing x.Furthermore, interference need not involve actual prevention, since prevention can also be subjunctive (Steiner, 1994: 33-41).B interferes with A by actually preventing her from doing x, but also if he, subjunctively, would stop A from doing x were A to try to do x.In either case, B makes it physically impossible for A to do x.
Robustness refers to the degree to which freedom requires that A can perform an action regardless of others' attitudes toward her.As in freedom as preference decisiveness, the robustness of a conception of freedom in action is measured in terms of socially possible worlds.The higher the number of such worlds in which freedom requires that the agent not be prevented from performing the action, the greater is the robustness requirement of that definition of freedom.On a definition of freedom that requires some degree of robustness, we cannot consider A free to do x simply because she is not prevented from performing x.She must also be unprevented from doing x across a certain range of possible worlds.Maximal robustness means all socially possible worlds.
The crucial observation is that the trade-off between scope and robustness is unavoidable.Making sure that individuals can perform a range of actions across different socially possible worlds means preventing them from preventing each other from performing those actions.Robustness thus comes at the cost of scope.And increasing the range of actions interference with which is taken to make individuals unfree, means reducing the range of possible worlds in which individuals are not prevented from performing all those actions.There are more worlds in which others will let A do x than there are worlds in which they let her do x and y, and there are even fewer worlds in which they let her do x, y, and z, and so on.Scope thus comes at the expense of robustness.

| POSSIBLE TRADE-OFFS
For pure negative freedom theorists, any interference makes you unfree.To use a classic example, when the ship crew tie Ulysses to the mast, they make Ulysses unfree to steer the ship off its course, since they prevent him from doing so.It makes no conceptual difference that Ulysses has instructed the interference.Pure negative freedom therefore has maximal scope: any act of prevention makes the agent interfered with unfree.
Freedom concepts that are not purely negative take certain acts of prevention to not make the agent interfere with unfree.This identification of preventions that do not make us unfree implies a reduction of scope.We can, for example, reduce the scope by counting only constraints for which some agent is morally responsible as a source of unfreedom (Kristjánsson, 1996;Miller, 1983), or only constraints unjustified under a libertarian theory of justice (Nozick, 1974).By thus narrowing the set of preventive acts that can make us unfree, we reduce the scope of freedom.
Republican freedom reduces the scope of freedom by saying that interference does not make an agent unfree if the agent has instructed it.Interference that is under the control of the agent interfered with is non-arbitrary and not a source of unfreedom. 7For republicans, Ulysses tied to the mast while listening to the song of the sirens is not unfree to steer his ship off course, or to perform any other action he could have performed had he not been tied to the mast, since he instructed his crew to tie him to the mast (Pettit, 2012: 152-153). 8Freedom is the status Ulysses enjoys with respect to his crew, and the crew cannot make Ulysses unfree as long as they are somehow constrained to respect this status.
By analogy, Pettit thinks citizens are not made unfree by government interference that tracks their 'common avowable interests' (Pettit, 2001b: 156-160).These are interests citizens can express in 7 Pettit has recently abandoned the arbitrary/non-arbitrary terminology and now prefers 'uncontrolled' and 'controlled' interference.He gives two reasons for this change (Pettit, 2012: 58): First, he wants to avoid any association with 'arbitrary' as it is commonly used to describe actions not conforming to established rules.Interference conforming to established rules may still be uncontrolled, Pettit argues, since rules need not serve the interests of those subject to them.Second, he wants to avoid any connotation of arbitrary with morally wrong or objectionable.I continue to use the arbitrary/non-arbitrary distinction because it remains in use in the literature on republican freedom, including in List and Valentini (2016) and Carter and Shnayderman (2019), and it will likely be more familiar to many readers. 8Pettit (1997: 26, fn. 1) adds that while B's non-arbitrary interference does not make A 'unfree', it does make A 'non-free'.If B is appropriately made to interfere only in accordance with A's instructions, B's interference makes A 'non-free but not unfree'.public without embarrassment because they are shared by all citizens who are willing to treat each other respectfully as free and equal members of their society.More specifically, these are interests in exercising the basic liberties. 9Police preventing crime, he says, make no citizen unfree insofar as they act in accordance with common avowable interests (Pettit, 2001a: 15).And imprisoning people do not make them unfree as long as they are sentenced under a law tracking citizens' common avowable interests (Pettit, 1997: 56, fn. 3). 10 The citizens then remain free persons.Republican freedom thus picks out a set of constraints that do not make us unfree, and thus reduces the scope of freedom.But we shall see in the next section that measuring the scope of republican freedom requires an understanding of its robustness requirement.
On the robustness dimension, we find pure negative freedom at the minimum end.A is free to do x in the actual world as long as no one prevents her from doing x if she tries.There is no requirement that no one prevents A from doing x also in other possible worlds.We can therefore position pure negative freedom at the maximum end of the scope dimension and the minimum end of the robustness dimension.But let us note that protecting individuals against interference is a plausible way of promoting their overall pure negative freedom-that is, increasing the number of combinations of actions they can perform conjunctively (Carter, 1999: 169-183;Kramer, 2003: 404-413).I return to the issue of measurement of freedom in Section 6.Here I am concerned only with the freedom to perform some particular action.
Republican freedom, on the other hand, requires that agents be robustly protected against interference they have not instructed.If A is not robustly protected so that she can do x in a range of possible worlds, then she is not free to do x, even if she in fact does x in the actual world.And B's non-arbitrary interference makes A unfree if B is not sufficiently constrained to make sure he only interferes non-arbitrarily-that is, robustly in accordance with her instructions.When the crew members tie Ulysses and refuse to untie him, they must do so because of Ulysses' instruction.They must be constrained so as to act on this instruction also in possible worlds in which their preferences are different.
The extent to which republican freedom demands robust protection has been a focus point in the literature on republican freedom.List (2006: 218) observes that 'republicans may have to reduce the scope of freedom to such an extent that very little individual freedom can be preserved if maximal robustness is to be achieved'.But this is an understatement, as we have already seen that maximal robustness is not compatible with any scope at all.The maximal-robustness interpretation has therefore led critics to the conclusion that republican freedom is impossible; there is no way an agent can be completely protected against arbitrary interference in all socially possible worlds (Carter & Shnayderman, 2019: 139-140;Dowding, 2011;Gaus, 2003: 69-74;Goodin & Jackson, 2007;Kramer, 2008: 45;Simpson, 2017).There is always a possible world in which A is interfered with in a way A has not instructed.In terms of the two-dimensional picture, this line of criticism says that maximizing robustness entails reducing the scope of freedom to zero.In other words, no one is ever free to do anything whatsoever. 9For Pettit, the basic liberties are the liberties that meet three conditions.First, they can be enjoyed, or exercised, without thereby preventing any other number of people from exercising them at the same time.Second, they are widely considered within a society to have an important role in the lives of normal people.And third, the set of basic liberties are limited only by these first two conditions (Pettit, 2008b).This definition leads Pettit (2008bPettit ( : 220, 2012: 103) : 103) to a list of basic liberties that includes at least freedoms of thought, expression, religious practice, association, assembly, private property, employment, movement, as well as to take part in public life as a voter, candidate, or critic. 10Carter (2008: 65) rejects Pettit's analogy between Ulysses' control over his crew and popular control over the government.After all, it is implausible that an individual wants to go to prison even under popularly controlled law enforcement.Others have also questioned this analogy between individual and popular control, including List and Valentini (2016: 1061-1063).But see especially Sharon (2016).The criticism seems appropriate when we note that Pettit (1997Pettit ( : 88, 2012: 60) : 60) himself requires that arbitrary interference be made 'inaccessible'.Other agents can always access a possible world in which they interfere with A. However, inaccessible cannot mean strictly impossible, since Pettit (1997: 22) takes republican freedom to only require the absence of arbitrary interference with impunity (see also Skinner, 1998: 72).What makes an agent unfree, Pettit (2011: 708) says, is another agent's 'power of relatively costless interference'.This impunity condition weakens the scope of republican freedom by making it compatible with more interference than simply the set of all uninstructed interference.If B prevents A from doing x despite not being instructed by A to do so, A is not made unfree, in the republican sense, as long as B is appropriately punished for his interference.The range of interference understood not to make the agent unfree is thus expanded, which implies a reduction in scope.
This scope reduction is insufficient to justify maximal robustness, however.As Dowding (2011: 311) points out, criminals often get away with their crimes.And the extent to which the law is enacted will always depend on the motivations of government officials and citizens' disposition to report every crime they observe (Kramer, 2010: 842-844).Pettit (2012: 173) says citizens must remain vigilant of their government officials to make sure they act in the people's interest.But such vigilance depends on the wills of the citizens, and therefore appears to add little to the robustness of the protection against arbitrary interference.And even with such virtuously disposed citizens and government officials, a monitoring problem seems likely to persist.Ingham and Lovett (2019: 778-779) agree that a 'strong' protection requirement of perfect law enforcement would make republican freedom impossible.They add, however, that Pettit takes a 'moderate' and possible view.Pettit (2012) weakens the law enforcement requirement when he says people are free as long as they can exercise the basic liberties without fear of the consequences of doing so.The probability of B being punished must be high enough for A to feel assured that B will not interfere.A will then 'have no good reason to be anxious' (Pettit, 2012: 71).Pettit (2012: 68, fn. 38) admits that this requirement is vague, but he sees no way of specifying more exactly how probable successful law enforcement must be.The range of possible worlds without interference with impunity, he says, 'is discernible only on an intuitive, context-sensitive basis' (Pettit, 2012: 32, fn. 8).Under the required conditions, Lovett and Pettit (2019) explain, no collection of individuals can coordinate so as to take control of legal institutions and gain the power of uncontrolled interference.It remains the case, however, that the required social norms underpinning such effective institutions can change, as norms often do, and people therefore remain dependent on each other's wills (Simpson, 2019).
The vagueness of its robustness condition makes it unclear whether republican freedom's position on the robustness dimension makes it possible or not.But we can stipulate that it can go some way out on the robustness dimension to a point at which it requires that actions one is free to perform is successfully protected in a certain range of possible worlds.That is, very vaguely, institutions can protect individuals' unprevented ability to perform a certain set of actions across a certain range of possible worlds.Of course, such a gain in robustness comes at the cost of scope.Given the sensitivity to social context, the exact position will differ from society to society and from one time to another.While this vagueness in the definition of republican freedom appears to allow for a possible trade-off, it leaves it unclear precisely what republican freedom is.

| IMPOSSIBILITY
Freedom as independence occupies a distinct position in the two-dimensional scope-robustness picture.List and Valentini (2016)  concepts provides all we want from a conception of political freedom, they argue, because they both fail to satisfy the following two desiderata: The functional-role desideratum: The conception picks out as sources of unfreedom those modal constraints on action (by which we mean actual or possible constraints) that stand in need of justification (List & Valentini, 2016: 1049).
Ordinary-language plausibility: The conception displays an adequate level of fidelity to ordinary-language use (List & Valentini, 2016: 1051).
To meet both of these desiderata, a definition of freedom must treat any modal constraint that is imposed by one agent on another and stands in need of justification as a source of unfreedom.The definition must also be true to ordinary-language use of the term 'freedom'; it must be compatible with commonly held intuitions about freedom.
List and Valentini take any act of interference to stand in need of justification, however justifiable it might be.The functional-role desideratum therefore requires that all kinds of interference be understood to make the agent interfered with unfree.Freedom must, in other words, have maximal scope.On the basis of the discussion in the previous section, we see that the scope of pure negative freedom satisfies this desideratum, whereas that of republican freedom does not.
List and Valentini show how the scope of republican freedom fails to meet either desideratum by referring to the case of a prisoner sentenced under a law made and enforced under popular control.This prisoner is not made unfree in the republican sense.11But she can still demand a justification for her imprisonment (List & Valentini, 2016: 1059).Republican freedom's scope restriction therefore means it fails to satisfy the functional-role desideratum.The prisoner example also implies that the scope of republican freedom does not meet the ordinary-language plausibility desideratum, since we ordinarily think of prisoners as paradigmatically unfree regardless of why they have been imprisoned (List & Valentini, 2016: 1051).
As for robustness, List and Valentini take the two desiderata to require a definition of freedom according to which A is unfree to do anything as long as B can access a possible world in which he interferes with A. They understand republican freedom to satisfy this requirement, while pure negative freedom fails to do so.They reach this view with reference to the example of a slave with a benevolent, non-interfering master (List & Valentini, 2016: 1054-1056).The slave, qua slave, can always demand a justification from her master for her inferior position, no matter how non-interfering the master happens to be.The functional-role desideratum therefore requires that the slave can never be considered free to do anything.Also the ordinary-language desideratum conflicts with the idea of the slave being free in any action, they argue.After all, they say, we cannot plausibly think the slave is free to do anything, since anything he does is due to the master allowing him to do it, perhaps just because he happens to be in a good mood (List & Valentini, 2016: 1054-1055).(They say nothing about the conflicting intuition that we are free to perform the actions we actually perform.) 12or pure negative freedom theorists, the slave is free to do x as long the master (or anyone else) does not prevent her from doing x.It does not matter whether the master prevents the slave from doing x in other possible worlds, where his preferences are different.For republicans, however, the slave is not free to do x, since the master can easily interfere arbitrarily with his slave with impunity.List and Valentini (2016: 1051-1058) therefore reject the robustness of pure negative freedom and adopt that of republican freedom. 13hey conclude that meeting the two desiderata requires a definition of freedom with the scope of pure negative freedom and the robustness of republican freedom.They therefore present their conception of freedom as independence as a combination of the scope of pure negative freedom and the robustness of republican freedom.On this account, A is free if and only if no one prevents her from performing any action (scope) and no one possesses the power to do so (robustness).Imprisonment will therefore always make prisoners unfree, since it involves preventing them from performing certain actions.And slaves will always be unfree because they are subject to the power of someone with the power to interfere with them.Freedom as independence is therefore the robust absence of all interference, not just arbitrary interference.As List andValentini (2016: 1044) say, freedom as independence is 'the robust absence of constraints simpliciter, not only of arbitrary constraints' (emphasis in the original).
In light of the earlier discussion, combining the scope of pure negative freedom with the robustness of republican freedom suggests maximal scope dimension and a position some way out on the robustness dimension.However, freedom as independence is positioned at the maximum of both dimensions.Explaining why is the beginning of showing why freedom as independence is impossible.List andValentini (2016: 1048) suggest a moderate position along the robustness dimension by understanding their freedom ideal to require protection against interference only in nearby possible worlds, not in all possible worlds.The distance from the actual world to other possible worlds is measured in terms of accessibility-that is, how likely B is to succeed in accessing a possible world where he interferes with A if he tries.The master may be benevolent and therefore unlikely to interfere with his slave, but there is nonetheless a nearby world in which the master interferes, since he can easily do so if he tries.In List andValentini's (List &Valentini, 2016: 1053) example, the master makes the slave unfree to take a nap, since there is a nearby possible world in which he prevents his slave from taking a nap.
However, the maximal scope of freedom as independence-that is, its view that any interference makes an agent unfree-disables List and Valentini from finding a moderate position on the robustness dimension.The reason is that making some possible worlds less accessible means imposing constraints on people.Protecting A's robust ability to perform certain actions involves restricting how B can behave.Maximal scope is therefore compatible with no more than minimal robustness, as we have seen.To say that A is free to do x only if B does not interfere with her doing x in any nearby possible world, List and Valentini must reduce the scope of their definition of freedom.
To gain robustness, they must follow republicans in treating the interference of protective institutions as no source of unfreedom.But this reduction in scope will make freedom as independence appear indistinguishable from republican freedom.It also violates the functional-role and ordinary-language plausibility desiderata on List and Valentini's own terms.And by holding onto maximal scope, freedom as independence cannot accommodate the impunity condition that allows republicans to give their freedom concept some robustness (Carter & Shnayderman, 2019: 140).
As already noted, List (2006: 218) realized that enhanced robustness necessarily implies a reduction in scope when he wrote that 'republicans may have to reduce the scope of freedom to such an extent that very little individual freedom can be preserved if maximal robustness is to be achieved'.A decade later, however, List has abandoned this trade-off view between scope and robustness.Like any other type of interference, the interference involved in gaining robustness stands in need of justification and therefore makes you unfree, according to List and Valentini's account of freedom.
No definition of freedom can therefore satisfy List and Valentini's functional-role desideratum without making freedom impossible.A definition of freedom cannot require the absence of all 'actual and modal constraints'-that is, the absence of any kind of interference both in the actual world and in a range of possible worlds.For the same reason, no possible definition of freedom can meet the ordinary-language plausibility desideratum if 'adequate level of fidelity to ordinary-language use' is interpreted as demandingly as List and Valentini appear to understand it.Freedom cannot demand both actual and modal non-interference, as such protection against interference itself necessarily involves interference.Common intuitions about freedom, as List and Valentini understand them, are therefore inconsistent.I suspect the observation that people commonly hold conflicting intuitions about freedom will come as no surprise to many readers (Pettit, 1998: 277-279;Steiner, 1994: 6-9).
We see, then, that freedom as independence is an attempt to avoid an unavoidable trade-off between scope and robustness.There is no way of making sure a person will not be prevented from performing an action in all socially possible worlds, and this level of robustness is certainly not compatible with the view that the interference necessary for enhancing robustness is itself ruled out as a source of unfreedom.This move, Carter and Shnayderman (2019: 140) argue, 'magnifies the problem of the impossibility of freedom to the point of absurdity'.No one can ever be considered free to perform any action.List andValentini's (2016: 1072) view of freedom as independence as 'arguably demanding' is certainly an understatement. 14 I should qualify the impossibility objection to freedom as independence by emphasizing that my discussion in this paper is based on the assumption that freedom is a value enjoyable for two or more individuals.We can imagine a society in which everyone except the dictator is constrained to an extent that makes it impossible to interfere with the dictator.The dictator then enjoys freedom with both maximal scope and robustness, but only because everyone else enjoys no freedom at all.So, we could perhaps say that at least the dictator has freedom as independence because no one else in her society does.The same holds for an isolated individual, such as a Robinson Crusoe on a desert island no one else can access.But List and Valentini do not have such cases in mind when they define freedom.After all, only a conception of freedom that allows for the possibility of two or more individuals being free can be the challenge to pure negative freedom and republican freedom that freedom as independence is intended to be.It is therefore safe to assume that freedom as independence is meant to be enjoyable for more than one individual per society.But that is, as we have seen, impossible.

| MEASURING FREEDOM
But perhaps freedom as independence can still be a value to promote, though we can never fully realize it.If we can measure freedom, we can talk about promoting it without necessarily having any hope of achieving it in full.
On the pure negative account, a specific freedom to perform an action does not exist by degrees, as you either have a freedom do to x or you do not (Kramer, 2002).But overall freedom is a matter of degree, as it is measured in terms of the combinations of actions no one prevents you from performing 14 Ingham and Lovett (2019: 778, fn.13) also suggest that List and Valentini are committed to an impossibly strong protection requirement.
conjunctively.We can therefore talk about maximizing people's freedom or promoting some other distribution of freedom (Carter, 1999: ch. 3).
Republican freedom is a status and therefore cannot be measured by counting specific freedoms.But Pettit thinks we can meaningfully talk about having this status to some degree: you are free to the extent that you are protected against arbitrary interference.15List andValentini (2016: 1071) take a similar view of freedom as independence when they say it can be 'reinterpreted as a matter of degree'.We can say an agent has more or less freedom as independence, they explain, depending on the number of possible worlds in which she is not interfered with.While freedom as independence still has maximal scope and robustness, they suggest taking degrees of robust protection to correspond with degrees of freedom as independence.
Carter and Shnayderman (2019: 141) also consider this possibility.They conclude, however, that even a low degree of freedom as independence is impossible, since there will always be nearby, easily accessible possible worlds in which interference occurs.This is right, since we have seen how the protection that can make interference less accessible itself involves interference.By killing A, B makes A unable to perform any action.Any protection of A that means there is no easily accessible world in which B kills A must involve interfering with B so as to make it harder or more costly for him to access this world.Any move up along the robustness dimension must therefore correspond with a reduction in scope.So, we cannot measure degrees of robustness while holding the scope fixed at the maximum.
But what if we instead hold the robustness dimension fixed at the maximum and measure freedom as independence by counting the number of actions an agent can perform in all socially possible worlds?An implication of the previous section is that also this measure of freedom as independence will fail, since maximal robustness implies that no one is free to do anything whatsoever.There is not a single action an agent can perform in all socially possible worlds.A measurement along the scope dimension will, in other words, be stuck at zero.
A third suggestion would be to measure freedom as independence in terms of both robust protection and the number of unprevented actions.We can aim for an optimal trade-off between the degree of actual non-interference and the degree of protection against interference.It is not clear what this optimal trade-off point would be, but we can plausibly take it to be a point at which everyone can pursue a wide variety of courses of action.That is, citizens can perform various combinations of conjunctively exercisable actions without being interfered with.But this is exactly how overall pure negative freedom is measured (Carter, 1999;Kramer, 2003).Protection against interference itself involves interference, as it makes people unable to perform certain actions, thus taking away some of their specific freedoms.But the protection can nonetheless enhance people's overall freedom by enabling people to pursue courses of action that would otherwise be unavailable to them.
This third way of measuring freedom as independence also looks like that of measuring republican freedom.Pettit (2017: 338) understands the 'common avowable interests' to be in the ability to exercise of the basic liberties, which gives people 'a generous set of compossible choices'.He sees the interference necessary for protecting this ability as non-arbitrary and therefore compatible with republican freedom.Interference that enhances individuals' range of choices, or actions they can perform, therefore appears to be the kind of interference Pettit does not see as a source of unfreedom.
It is therefore understandable that Carter (1999Carter ( : 240, 2008) ) and Kramer (2003Kramer ( : 128-137, 2008) argue that the measurement of republican freedom is indistinguishable from that of overall pure negative freedom. 16Pursuing the ideal of republican freedom appears reducible to the promotion of pure negative freedom (Moen, Forthcoming).This is no place to get involved in that debate, however.My point here has been to show how pure negative freedom and republican freedom are both possible, though the latter is vaguely defined, and that freedom as independence is impossible.And its impossibility disables freedom as independence from providing a measure of freedom and from functioning as a value we can try to realize to a certain degree.

| CONCLUSION
To avoid impossibility-that is, the implication that no one can ever be free to perform any actionwe cannot define freedom so that it requires the absence of interference in all socially possible worlds.We need a weaker robustness requirement, and to the extent that we want more than minimal robustness, we must reduce the scope of freedom.This is evident in the observation that protecting agents against interference necessarily means preventing them from interfering with one another.
List and Valentini ignore the inevitable trade-off between scope and robustness.Their freedom as independence is therefore impossible.Both pure negative freedom and republican freedom, however, appear to be based on trade-offs that allow for the possibility of being free to perform an action.Pure negative freedom is obviously possible, since it requires minimal robustness-that is, A is free to do x in the actual world as long as no one prevents her from doing x in the actual world.Republican freedom has been criticized for being impossible, but the robustness requirement is not specified precisely enough to allow for this conclusion.This vagueness makes it unclear what exactly republican freedom is, but it also appears to enable an interpretation of the republican scope-robustness trade-off that avoids impossibility.
This paper is indeterminate as to which conception of political freedom we ought to prefer all things considered.It has merely clarified one consideration, namely that of possibility.It has clarified the bounds of possibility, and while they can accommodate more than one conception of freedom, they can exclude some, and thereby disqualify them as contenders for the best way of conceptualizing political freedom.