The Impossibility of “Freedom as Independence”

Actual, Possible, and Minimally Probable Constraints

Let us begin by tracing the debate over the robustness requirement prior to List and Valentini’s contribution. Republicans initially criticized the liberal conception of freedom by arguing that it takes into account only interferences that occur in the actual world and ignores other possible interferences. On the republican view, by contrast, a person is free only if the non-interference she enjoys is robust. To illustrate this contrast, republicans pointed out that a person might be dominated, and hence under the power of others, while not actually suffering any interference on the part of those others. This situation might be a result of the fact that she is careful not to do anything her dominators do not want her to do, or that she manages to curry favor with them, or that her dominators happen to be kind, non-interfering persons, and so on. If one or another of these favorable conditions were to change, however, interferences would quickly follow. Consider the example of a slave with a non-interfering master. This relatively lucky individual appears to enjoy a high degree of non-interference in the actual world. Yet, as a slave, she remains unfree. Why? The reason must be that she does not enjoy non-interference in a range of possible worlds—that is, she does not enjoy it robustly. Republicans, unlike liberals, are therefore said to be concerned with the absence of both actual and possible interferences (Pettit, 1997: 24).

As several liberal theorists have pointed out, however, the liberal conception of freedom is itself very much concerned with the absence both of actual interferences and of a large class of possible-but-not-actual interferences. ² This concern is manifested in at least the following two ways. First, liberals characterize unfreedom in terms of interferences that exist both actually and counterfactually, in more or less nearby possible worlds, conditionally upon a series of actions or omissions on the part of the agent whose freedom is under consideration. Joel Feinberg (1978: 27–29) provided a helpful analogy long before the liberal-republican distinction took hold: the liberties available to a person are like a series of railway lines, inasmuch as the choices we make at the many points along the line may open up onto main lines presenting us with many other options, or onto sidings, closing off various options. Applying this image to the notion of social freedom, these hypothetical future openings and closings are actions or omissions performed by other agents in many possible worlds. Only a fraction of these actions or omissions ever become actual, but they are all obviously relevant to the freedom enjoyed by the agent at any one time in the actual world. ³ Liberals can take account of such non-actual preventive behavior by considering the extent to which an agent’s spatio-temporally specific liberties are conjunctively exercisable (Carter, 1999: 169–183; Kramer, 2003: 404–410).

Second, for each action or set of actions, we must ask how probably constrained it would be in the event of the agent attempting to perform it. The liberal conception of freedom therefore takes into account all possibilities of interference starting from the actual world at the time of the freedom being assessed (on some specified notion of possibility, such as logical, nomic, or technological possibility) but weighs each possible interference in terms of the probability of its occurrence (on the assumption that the agent attempts to perform the action that might be subject to that interference). ⁴ The greater the probability of interference (conditional upon the agent attempting to perform the relevant action), the more that interference contributes to the agent’s unfreedom (Carter, 1999: 233–245; Kramer, 2003: 174–178).

In light of these points, the liberal response to the republican insistence on a “robustness requirement” has been to say that such a requirement, taken on its own, fails to imply a significant difference between the judgments about freedom respectively implied by the liberal and republican conceptions (Carter, 2008). On a liberal conception of freedom, no less than on a republican one, a dominated person who does not suffer any actual interferences only because she is careful not to do anything her dominators do not want her to do, or because she curries favor with them, is less free than she would otherwise be, for the number of her sets of conjunctively exercisable liberties is smaller than it would otherwise be. After all, the performance of any action that her dominators do not want her to perform, or the omission of actions which curry favor with them, will lead to a reduction—perhaps indeed a very great reduction—in her subsequent opportunities. Similarly, a dominated person who enjoys an actual absence of non-interference only because her dominators happen to be in a good mood, or happen to be kind people, and happen not to get substituted with less accommodating types, is also less free than she would otherwise be: ceteris paribus, she is more likely to suffer interferences than someone who is not similarly dominated. Liberals have argued that these points suffice both to accommodate and to explain, in terms of freedom, the concern about the precariousness and constant fear of those who are subject to the power of others, even where such power is not actually exercised. Rather than defining freedom as the absence of domination, they say, we should explain our concerns about domination in terms of its effects on freedom.

There remains, however, one important difference between the liberal and the republican conceptions of freedom regarding the effect that possible-but-not-actual interferences have on freedom. As we noted above, on the liberal conception, possible-but-not-actual interferences are taken into account but only in proportion to the conditional probability at the time of the freedom being assessed, of their becoming actual. This clarification has proved insufficient to placate the republicans. On the republican view, the latter have retorted, the sheer possibility of an interference, understood as the presence of that interference in a range of possible worlds, is sufficient to make an agent unfree regardless of how improbable that interference might be (Pettit, 1997: 74, 88; Skinner, 2008: 96–97). To illustrate this point, they have cited the case of a person who is dominated but is nevertheless extremely unlikely ever to be interfered with in the performance of any number of actions. Such a person is unfree, they say, whereas liberals would describe her as enjoying considerable freedom. Liberals have replied that such a scenario is highly unrealistic, and that if it were realistic, there would be no cause to characterize that person’s situation as one of precariousness and constant fear, nor therefore to call that person unfree.

List and Valentini recognize the internal consistency of the liberal appeal to conjunctive exercisability and probabilities, but they nevertheless take the republican side in the above-mentioned debate over possibilities versus probabilities. Like republicans, they argue that an extremely improbable constraint on the freedom to perform a specific action makes one unfree to perform that action, unless that constraint is absent from a range of possible worlds. Unlike republicans, they qualify this range of worlds by saying that it consists is a set of “nearby” possible worlds. Nevertheless, their agreement with the republicans on the issue of possibilities versus probabilities is crucial to their insistence on a contrast between “liberal freedom” and their own conception of “freedom as independence.” ⁵

Unfortunately, and as we shall now see, interpreting the robustness requirement in terms of possibility rather than probability raises serious problems, particularly when we drop the requirement that the interference should be arbitrary.

Probability Versus Sheer Possibility

As Matthew Kramer (2008: 45–46) points out with respect to Pettit’s republican conception of freedom, the eventuality of any particular person or group of persons acquiring the means that would enable them to possess arbitrary power to interfere in our lives is never a matter of binary distinction between possibility and impossibility. Instead, it is always a matter of greater and lesser probability. So the risk of any person or group of persons acquiring such means can never be categorically ruled out; it can only be rendered more or less probable. There is of course a distinction to be made between the likelihood that any person or group of persons will succeed in acquiring the means that would enable them to possess arbitrary power and the likelihood that persons who are already powerful will use their power. In effect, however, these two considerations will almost always converge. As Kramer (2008: 46) puts it, “[f]or anyone who desires to minimize the likelihood of the exercise of oppressive powers, the best route will typically be to minimize the accumulation of such powers in the first place.”

Now, as long as republican robustness is to be considered something distinct from the liberal minimization of probabilities, at least some of the minimally probable scenarios involving the accumulation of such powers must qualify as relevant possibilities from the republican point of view, that is, as occurring in possible worlds that are relevant in characterizing the robustness requirement. As a result, Pettit’s conception runs into the following problem:

The impossibility of republican freedom. Republican freedom entails that no one is ever free to do anything whatsoever, not even in a republican state, given that there is always a relevant possibility that a person or a group of persons will succeed in acquiring the means that would enable them to take over the state and thereby possess more or less absolute arbitrary power to interfere with every aspect of its citizens’ lives.

The important thing to note here is that, on Pettit’s conception of freedom, this problem arises only with respect to the particular case of a threat to the republican state as a whole. Other cases of people’s power to interfere with each other’s lives, including in the most severe and morally outrageous ways—for example, by stabbing each other in the back with a kitchen knife as they walk down the street—will of course be present, for they cannot be eliminated by any reasonable state, even a republican one. However, these other cases do not pose such a problem for Pettit’s conception of freedom. This is because, on his conception, in order to count as a source of unfreedom, the power to interfere has to be arbitrary. In order to be arbitrary, it has to be, inter alia, possessed with impunity (Pettit, 1996, 1997, 2011: 57–58). Given that no one would enjoy impunity within a republican state with respect to such cases, such cases would not involve arbitrary power. Therefore, they would not involve unfreedom. ⁶

This raises, of course, a different problem for Pettit’s conception of freedom. Stabbing someone in the back is a very clear and serious source of unfreedom, regardless of whether it is done with or without impunity. Yet, on Pettit’s conception, the possibility of doing so without impunity does not count as a source of unfreedom. This is, however, nothing but an aspect of the moralization problem which List and Valentini’s position avoids by dropping the arbitrariness requirement. Thus, in what follows, we will set aside this problem. We will continue to focus on the problem of the impossibility of freedom, which receives a whole new momentum under List and Valentini’s position.

To recall, in order to avoid moralizing the concept of freedom, List and Valentini drop the arbitrariness requirement. Thereby, they also drop the impunity requirement. Consequently, their conception of freedom as independence magnifies the problem of the impossibility of freedom to the point of absurdity:

The impossibility of freedom as independence. Freedom as independence entails that no one is ever free to do anything whatsoever. But this entailment depends not only on the particular case of a threat to any regime that would otherwise secure its citizens’ freedom; it depends also on the countless threats which everyone constantly faces in everyday situations, given that these threats cannot be eliminated by any reasonable state.

So think again about the possibility of being stabbed in the back with a kitchen knife while walking down the street, or the possibility of being run over by a car while crossing the street, or the possibility of being pushed onto the train tracks while waiting at a station just as the train is arriving, or the possibility of being poisoned either by a waiter while eating in a restaurant or by one’s partner or kids while eating at home, and so on and so forth. On the conception of freedom as independence, given the mere existence of the possibility that anyone can be killed by someone so easily at any given moment, no one is ever free to do any of the things that their death would prevent them from doing, in other words, no one is ever free to do anything whatsoever.

On this conception, the very low probability of anyone actually killing us is irrelevant. Therefore, the liberal way out of this problem is unavailable. Similarly, the fact that no one can do so with impunity is also irrelevant. Therefore, the republican way out of the problem is also unavailable.

The cases presented above certainly satisfy List and Valentini’s “functional-role desideratum,” according to which an obstacle to an action counts as a source of unfreedom if and only if it stands in need of justification. As List and Valentini (2016: 1049–1050) point out, this desideratum implies not that every source of unfreedom is unjustified but merely that it is “a legitimate object of justificatory appraisal.” And it is certainly legitimate for pedestrians to ask for a justification as to why pedestrian crossings are not protected with barriers that would make it impossible for drivers to run them over when they use them to cross the street. Such a justification is of course readily available—the low likelihood of being run over, on the one hand, and the high cost of building such barriers on every pedestrian crossing, on the other. But this is not to say that it is inappropriate to ask for such a justification.

Notice, moreover, that the cases cited above are similar in all relevant respects to that of the slave with a non-interfering master, which serves as List and Valentini’s paradigmatic example of comprehensive unfreedom. ⁷ They are similar in terms of the strength of the constraints and the range of actions that are at stake: in all the above cases, just as in the case of a slave, there is someone who can physically prevent the agent from performing every future action. They are also similar in terms of the level of robustness of non-interference: in all the above cases, just as in the case of a slave, at any given moment there is someone who can very easily physically prevent the agent from performing every future action. Whether or not someone engages in such preventive behavior depends only on whether or not they decide to do so.

The point about robustness is particularly important. As List and Valentini point out, “freedom as independence” can admit different views regarding the required level of robustness. Moreover, “freedom as independence” does not have to be a binary attribute; it can also be reformulated as a matter of degree, corresponding to the relevant level of robustness (List and Valentini, 2016: 1071). It is therefore important to stress that our argument does not assume “freedom as independence” to require a high level of robustness. In fact, the level of robustness in the cases presented above in support of our argument is very low. As we have already seen, in these cases at any given moment there is someone who can very easily physically prevent the agent from performing every future action. Thus, the worlds in which they do so are all very close to the actual world (the improbability of a world should not, of course, be confused with its remoteness). And the number of nearby worlds containing similar constraints is presumably very high—setting aside, arguendo, any doubts about “counting” possible worlds. Above all, and regardless of how robustness is measured or where the threshold is set, it should be recalled that the level of robustness in the cases we have cited is similar to the level of robustness in the case of a slave with a non-interfering master. Thus, if on “freedom as independence” the slave is unfree (or highly unfree) to perform any action (or most actions), then so are the agents in the cases we have cited.

To illustrate this point more vividly, consider an excerpt from List and Valentini (2016: 1053), in which they use the example of the slave with a non-interfering master to argue against the liberal focus on probability and in favor of their own focus on sheer possibility. The following statement reproduces the excerpt verbatim, except that it substitutes the slave and his non-interfering master with a normal man called Mark and his wife who can easily stab him to death with a kitchen knife as he falls asleep:

Even when Mark’s ability to do some action X (e.g., “taking a nap”) is entirely contingent on his wife’s approval, the probability of his being constrained in doing X may vary a lot, conditional on his wife’s preference and the situation in which he tries to perform X. For example, if his wife is in a laissez-faire mood or happens to approve of X, the probability of interference is lower. Similarly, if Mark has made his wife happy or “appeased” her, the probability of interference is lower. Yet it would be out of line with ordinary-language use and misleading to suggest that Mark who successfully appeases his wife, or whose actions happen to be aligned with his wife’s mood of the day, is genuinely free to perform those actions, just as a person who lives in a society that made it effectively impossible to acquire knives.

Thus, on List and Valentini’s conception of freedom as independence, because of his wife’s ability to stab him as he falls asleep, Mark has no freedom whatsoever, just as the slave of a non-interfering master has no freedom whatsoever.

Indeed, on their conception, freedom-wise Mark’s predicament is not only as bad as that of a slave of a non-interfering master. It is also as bad as that of a slave of an interfering master, and indeed of someone who is just about to be stabbed by his or her partner. The reason for this is simple. If the mere possibility of interference is enough to render us unfree, then the interference itself cannot do anything more to our freedom. It cannot further limit our freedom as there is no freedom left to be limited. ⁸

Note further that, contrary to what is suggested in the above excerpt from List and Valentini, on the liberal conception Mark’s freedom is affected adversely by the possibility of being stabbed by his wife, but only in proportion to the conditional likelihood of her doing so in the various hypothetical scenarios available to him. ⁹ Therefore, all other things being equal, he is not as free as a person who lives in a society that makes it effectively impossible to acquire knives; rather, he is slightly less free. (All other things are of course not equal: given that knives are necessary tools for many non-freedom-restricting actions, their unavailability would very likely lower the freedom of most citizens.) However, and again assuming all other things to be equal, Mark is substantially freer that someone who is just about to be stabbed by his or her partner. Similarly, on the liberal conception of freedom, a slave of an interfering master is less free than a slave of a non-interfering master. This is not to say, however, that a slave of a non-interfering master is as free as someone who is not a slave. Even in the presence of a non-interfering master, a slave is normally much more likely than a non-slave to suffer future interferences. Only with respect to the mind-bogglingly unrealistic scenario of a slave whose non-interfering master is (a) extremely unlikely to ever interfere with the slave’s actions in the future in any of the hypothetical scenarios accessible to the slave, (b) extremely unlikely ever to sell or give the slave to another person, and (c) extremely unlikely to die without having first provided for the manumission of the slave in the event that the master dies before the slave does—and despite all this, for reasons known only to herself, the master chooses not to free the slave within her own lifetime—does the liberal conception entail that, all other things being equal, a slave can be virtually as free as a non-slave.

Now it might be suggested, in response to our argument, that there is nevertheless a relevant difference between the counterexamples we presented above and the case of a slave with a non-interfering master: the types of possibilities of constraint present in our counterexamples are non-normative, while those present in the case of a slave with a non-interfering master are normative. More specifically, the latter are legal normative possibilities. In a legal system recognizing slavery, the non-interfering master has the legal permission to constrain the slave, whereas an ordinary member of a contemporary democratic society does not have the legal permission to just stab another citizen. List and Valentini (2016: 1070–1071) themselves claim that “freedom as independence” can be concerned with different types of possibility, depending on the theoretical task at hand, and that this versatility is one of the virtues of their analysis. They might claim, then, that their argument about the unfreedom of the slave focuses only on normative possibilities (i.e. permissions) of interference.

However, List and Valentini cannot help themselves to the distinction between permitted and unpermitted interference in order to drive a wedge between the case of the slave and the other cases cited above, for to do so would be to contradict their “functional-role desideratum.” They indeed invoke that desideratum when arguing that legal constraints, no less than humanly caused physical ones, should count as sources of social unfreedom. Their reason for extending the set of relevant constraints in this way is that legal constraints, no less than humanly caused physical ones, stand in need of justification (List and Valentini, 2016: 1057). But the argument goes both ways. Indeed, the implausibility of focusing exclusively on permissions of interference can be clarified in List and Valentini’s (2016: 1057) own words, again replacing the slave and his master with Mark and his wife:

We would have to acknowledge that Mark is subject to some modal constraints on action that call for justification—that is, his wife does have the power to interfere, she is simply not inclined to use it—and yet insist that these should not be classified as restrictions of freedom.

Suppose, then, that List and Valentini were to jettison their “functional-role desideratum.” This would allow them to employ a purely normative (though non-moralized) conception of social freedom, according to which freedom is the robust absence merely of (weak or strong) permissions of interference, specified on the basis of some positive normative code. ¹⁰ Even in this case, however, the level of freedom of every ordinary citizen would turn out to be implausibly low. Consider the normative liberties normally enjoyed by the citizen whose freedom is under consideration. Any such normative liberty will presumably count as an unfreedom on the part of that citizen, on this revised version of List and Valentini’s conception, as long as it remains in some respect a “naked” Hohfeldian liberty (i.e. a permission to act that is not accompanied by claims to non-interference). But in every known normative system (whether legal or non-legal), most normative liberties are naked with respect to some class of permitted preventive actions (Steiner, 1994: 74–76). Moreover, a great many of these preventive actions occur in nearby possible worlds. Mark’s wife has the normative liberty to prevent him from entering the bathroom at any one time by getting in there first. His neighbor has the normative liberty to mow the lawn next door at any time of the day (though not of the night), thus preventing him from taking a nap. Protecting naked liberties with constitutional immunities (List and Valentini, 2016: 1073) does nothing to remove their nakedness.

Furthermore, by employing an exclusively normative conception of social freedom, List and Valentini would no longer be challenging the liberal conception of social freedom, which is non-normative: the whole point of the liberal conception is to provide a useful criterion on the basis of which to judge real and hypothetical enforced structures of normative liberties, claims, powers, and immunities. In fact, by giving up on a non-normative conception of freedom in order to avoid our critique, List and Valentini would not merely be talking past the liberal conception; they would also be acknowledging its superiority, as a non-normative conception over “freedom as independence.”

Perhaps there is yet some other set of constraint-types on which we ought to have focused in interpreting List and Valentini’s framework. Note, however, that the “virtue” of versatility can easily degenerate into the vice of vacuity: as each successive attempt to apply the framework falls foul of one or another of the problems identified above, there will be a growing temptation to refuse to venture beyond the framework itself. By abandoning the functional-role desideratum and remaining completely open about which kinds of possibilities to focus on, the advocate of “freedom as independence” may, in principle, try to answer any counterexample with the suggestion that it focuses on the wrong kinds of possibilities. Such a position will still fail to establish the possibility of “freedom as independence” because it will fail to establish in a sufficiently clear way what “freedom as independence” is.

Consider finally the real-world case which List and Valentini present in support of their conception of freedom as independence and against the liberal (and the republican) conception of freedom: the US Patriot Act and its effect on the freedom of US citizens and residents, and particularly of Joe—a White, middle class, Christian Male who lives in rural South Dakota with his wife, two kids, and a dog and works for a local insurance company. On the liberal conception, the effect of the Patriot Act on the freedom of each US citizen and resident is proportional, ceteris paribus, to the conditional likelihood that any of the powers it grants the executive will be applied to that citizen. Thus, freedom-wise Joe is affected to a much lesser degree than Musa—a young, single, devout Muslim student from Pakistan who studies nuclear physics in the United States. On List and Valentini’s conception, by contrast, Joe’s freedom and Musa’s freedom are equally affected by the Patriot Act. Thus, even if “freedom as independence” were possible for people like Joe and Musa, it would fail to capture any of the differences in the effects of such laws on the lives of different people. ¹¹ To many this implication will signal a defect, rather than a strength, of “freedom as independence.”

We conclude that the sheer possibility of constraint, as opposed to its probability, is not a plausible criterion to adopt when attempting to make sense of the notion of social unfreedom. List and Valentini’s contribution has advanced the debate by clarifying the nature of the robustness requirement. However, that clarification tells against the sheer-possibility view of constraints rather than in favor of it. As long as we take account of physical constraints intentionally imposed by other human agents—as any theorist intent on engaging with the liberal conception of social freedom surely must—then the sheer-possibility view renders unfreedom a ubiquitous phenomenon and freedom a virtually non-existent one. We have seen that this implication holds even where the robustness requirement is interpreted in a very undemanding way. Moreover, even if the advocate of the sheer-possibility view were to interpret “possible constraint” as meaning only “permitted constraint” (on some positive normative code), thus giving up on the idea of challenging the liberal, non-normative conception of freedom, she would see many unfreedoms in the place of the numerous normative freedoms people enjoy in virtue of their more-or-less naked Hohfeldian liberties, given that interference with such liberties is permitted. The sheer-possibility view of constraints can be saved from some of these pitfalls by re-introducing the arbitrariness requirement endorsed by republicans. But the latter requirement is itself seriously problematic for other reasons. Only the liberal conception, with its helpfully fine-tuned assessments of people’s non-normative freedom, provides us with a plausible basis on which to assess the freedom-promoting effects of the more-or-less effectively enforced norms entailed by various institutional set-ups, republican or otherwise.