An unresolved problem: freedom across lifetimes

I’ve been lookin’ for freedom.

I’ve been lookin’ so long.

I’ve been lookin’ for freedom.

Still the search goes on.

David Hasselhoff.

Abstract

Freedom is one of the central values in political and moral philosophy. A number of theorists hold that freedom (or, relatedly, opportunity) should either be the only or at least one of the central distribuenda in our theories of distributive justice. Moreover, many follow Mill and hold that a concern for personal freedom should guide, and limit, how paternalist public policy can be. For the most part, theorists have focussed on a person’s freedom at one specific point in time but have failed to give proper attention to freedom across time. Given that we care about personal freedom now, we have reason to care about future freedom too. But what kind of distribution of freedom across a person’s lifetime should we promote as a matter of legislation and public policy? I argue that none of the candidate principles for the distribution of freedom across time is plausible. Neither a starting gate view, nor a maximisation nor a sufficientarian view is satisfactory, because none adequately reflects our various reasons to value freedom. I show that this result presents a tough challenge for theories of distributive justice and paternalism that set great store by personal freedom.

1 Introduction

Freedom is a fundamentally important ideal, yet, as David Hasselhoff observes, its nature is so elusive one can spend a long time looking for it. One well-known reason for this is that different concepts might be employed to spell out ‘freedom’.¹ In this article, I add a different and often overlooked challenge for invoking freedom in normative political theorising: even if we settle on a particular concept and conception of socio-political freedom at a particular point in time, this does not by itself tell us what kind of distribution of freedom we should seek across time. It seems that if we have reason to care about a person’s freedom now, we also have reason to care about her future freedom. However, how much and what kind of freedom a person has in the future depends, among other things, on her very own decisions, on how she exercises her freedom. This article is about the intrapersonal, crosstemporal distribution of freedom (henceforth just intrapersonal distribution). Specifically, I discuss what kind of distribution of freedom across persons’ lives we should promote as a matter of legislation and public policy. This article ultimately ends with a negative result: none of the prima facie plausible candidates for a standard for the intrapersonal distribution is plausible. This means we should either be less confident about the role of freedom as a value in political theory or we should intensify our search for a plausible standard for the intrapersonal distribution of freedom.

This result has important implications, as it presents a tough theoretical challenge for normative political theories that set great store by personal freedom. First, a number of theorists argue that freedom should be one of the central distribuenda in a theory of distributive justice. However, without an account of the intrapersonal distribution of freedom, such theories remain strongly underspecified. Second, how paternalist interferences affect a person’s future freedom is often thought be an important consideration in the justification of paternalism. Yet without an account of the appropriate intrapersonal distribution of freedom, the precise relationship between freedom and paternalism remains somewhat unclear. Given the importance of this question, it is surprising that—to my knowledge—only one systematic attempt has been made to theorise personal freedom across time (Carter 2013).²

I start by outlining how the intrapersonal distribution of freedom matters for freedom-based theories of distributive justice and paternalism (Sect. 2). In Sect. 3, I distinguish between a person’s freedom at a point in time and its intrapersonal distribution and explain how this distinction matters for discussions of paternalism. In Sect. 4, I discuss different standards for intrapersonal distribution and show that none is plausible. In Sect. 5, I return to freedom-based theories of distributive justice and paternalism.

2 Justice and paternalism

Before delving into the theoretical arguments of this article, let me add some context by outlining two debates for which the problem of the intrapersonal distribution of freedom matters in particular.

2.1 Freedom, opportunity and justice

A number of theorists hold that distributive justice is about the right distribution of freedom. Others hold that, if not the only, freedom is at least one central distribuendum amongst others (Carter 1999, chap. 3; Carter 2013; Norman 1987; Van Parijs 1997; Pettit 2014; Sen 1999; Spencer 1873, 35; Steiner 1974b, 1994). Other theorists defend the related view that egalitarian justice implies equality of opportunities (Arneson 1989, 1991; Roemer 2000). For simplicity, I exclusively focus on the distribution of freedom across time—though much of what I say applies, mutatis mutandis, to the related concept of opportunities too.³

For example, consider how the intrapersonal distribution of freedom presents a challenge for Philippe van Parijs’ theory of distributive justice. Van Parijs argues that to extend ‘real freedom’ to all we should provide everyone with a basic income (Van Parijs 1997). But, as van Parijs wonders himself, why not provide everyone with an equal basic capital at one point in time instead (a ‘lump sum’)? Of course, choosing the latter option risks ending up with undesirable outcomes. For example, some people might end up destitute having squandered their capital early on. However, opting for equal basic income, as van Parijs does, implies adopting a particular standard for intrapersonal distribution. What shape should such a distribution take exactly and why? In the absence of a standard for intrapersonal distribution, and a principled defence thereof, van Parijs’ preference of basic income over basic capital seems theoretically unmotivated.⁴ Note that the problem of intrapersonal aggregation is not specific to van Parijs’ theory. Though van Parijs raises the challenge most explicitly, other freedom-based theories will face it too.

2.2 Freedom and paternalism

How to distribute freedom across time is a challenge for discussions of paternalism too. Most—though not all⁵—paternalistic interventions affect a person’s set of available options and thus her freedom of choice. While not being the only ethical challenge for paternalism, its effect on people’s freedom requires a justification. Some argue that a concern for people’s freedom itself can provide one such justification. On at least some interpretations, Mill made such an argument in cases of voluntary slavery:

The reason for not interfering, unless for the sake of others, with a person’s voluntary acts is consideration of his liberty. […] But by selling himself for a slave, he abdicates his liberty; he forgoes any future use of it beyond that single act. […] The principle of freedom cannot require that he should be free not to be free. (Mill 1979, 173)

When engaging with Mill’s ideas on the topic, Dworkin (Dworkin 1972, 76) holds that ‘paternalism is justified … to preserve a wider range of freedom for the individual in question.’⁶ Similarly, Amartya Sen approvingly cites the above section from Mill on voluntary slavery when discussing tobacco control. He suggests a concern for personal freedom might sometimes favour more paternalist tobacco control rather than less. He asks:

…how should we see the demands of freedom when habit-forming behaviour today restricts the freedom of the same person in the future? Once acquired, the habit of smoking is hard to kick, and it can be asked, with some plausibility, whether youthful smokers have an unqualified right to place their future selves in such bondage. (Sen 2007)

Of course, Sen’s tobacco case also raises issues of volitional autonomy and addiction. But the intrapersonal distribution of freedom is clearly one important aspect the above authors care about.

However, not everyone has been impressed with this idea. Richard Arneson argues that Dworkin’s proposal puts us on a slippery slope: ‘Why not ban cigarettes and fried foods on the ground that these shorten the individual’s life span and thereby shrink the range of his freedom?’ (Arneson 1980, 474)⁷ As I discuss below, Arneson’s slippery slope argument is too quick. Whether there really is a slippery slope depends on what intrapersonal distribution we choose as a goal in public policy. In the absence of such an answer, it is not clear at all that banning fried food will be conducive to a desirable intrapersonal distribution of freedom.

Intuitively, ‘freedom-based paternalism’ seems a strong argument in favour of some paternalist interventions. However, without a precise standard for intrapersonal distribution, it remains unclear which interferences it does or does not condone.

3 Point freedom and its intrapersonal distribution

Before discussing the intrapersonal distribution of freedom, let me add two clarifications.

First, I here focus on freedom as an opportunity-concept (Taylor 1979). Freedom on this reading is about a person’s range of opportunities (though, again, how that relates to opportunity-based egalitarian theories of justice is another, complex question). Depending on how we spell out ‘opportunities’, we can broadly distinguish two classes of theories. Views we might label ‘Restraint Views’ hold that one is free, if and only if one is free from external interpersonal restraints (Berlin 1969b; Carter 1999; Steiner 1974a, 1994). Theories within this class vary, of course, depending on what obstacles they count as external interpersonal constraint (Carter 1999, chap. 8; Kristjánsson 1996; Miller 1983; Shnayderman 2013; Steiner 1994). Views we might label ‘Ability-based Views’ hold that freedom requires not only being free from external, interpersonal constraints but also having the capabilities to do things. This, in turn, requires having relevant external resources, being free from external constraints and having the necessary internal abilities (Kramer 2003; Van Parijs 1997; Schmidt 2016a; Sen 1988, 1991, 1999).⁸ For the most part, I remain neutral between these two classes of theories in this article, as questions regarding intrapersonal distribution apply to both.⁹

Second, my topic is the distribution of overall freedom across time. Overall freedom is different from specific freedom. A specific freedom is the opportunity to do a specific action or to be in a specific state. You have, for example, the specific freedom to read this paper at this moment. Overall freedom, in contrast, refers to how much freedom a person has overall. An account of overall freedom is typically developed by aggregating specific freedoms or sets thereof. For the purposes of this article, we can disregard the difficult question of how to measure freedom and work with the following (simplistic) model: a person’s level of overall freedom at t is simply the cardinality of the set of specific freedoms available to her at t. ¹⁰ To sidestep the debate about whether a person’s overall freedom is a function of the quantity of a person’s freedoms exclusively or whether it also needs to include their quality, I will assume that all options are equally valuable (in the sense that they all count equally no matter what the measure of freedom).

Now, remember that Mill argued we should prohibit voluntary slavery, as the ‘principle of freedom cannot require that he should be free not to be free’ (Mill 1979, 173). I think one—though not the only—important feature of the problem of voluntary slavery is the following. Removing the option to become a slave is, in some sense, an interference with a person’s freedom. At the same time, however, letting a person sign up to be a slave implies that that person would lose many of her future freedoms. I argue that we can make sense of these seemingly conflicting intuitions. To do so, we should distinguish between Point in Time Overall Freedom (henceforth just ‘Point Freedom’) and the intrapersonal distribution of Point Freedom across time (henceforth ‘intrapersonal distribution’).

Point Freedom asks how much overall freedom a person P has at a specific point in time t. P’s overall freedom at t is the sum of all the specific freedoms had by P had t. The temporal aspects of specific freedoms are a bit tricky. In general, a person’s specific freedom to do something carries two temporal indices. P at t has the freedom to φ at t + n. t is what I call the ‘existence-time’ of P’s freedom, and t + n is the ‘actualisation-time’ of P’s freedom. For example, I might have the specific freedom now (existence-time) to visit you in Toronto in 4 months’ time (actualisation-time). However, if I commit a felony next month and will be imprisoned for a year, then I will not have the freedom next month (existence-time) to visit you in Toronto in 4 months’ time (actualisation time).

To determine a person’s level of Point Freedom, we can use different ranges of actualisation-times. Accordingly, although Point Freedom focuses on one specific point in time—the existence-time of the specific freedoms—it can have different temporal extensions. It can cover all of the person’s future or a shorter span such as the next 20 minutes or the next 20 years. Thus we could calculate how much overall freedom I have now (existence-time) over the next 10 years (temporal extension), or how much freedom I had last month (existence-time) for the rest of my life (temporal extension) and so on. Because it does not matter whether a person is inclined to choose a certain option or not, a person’s Point Freedom will include all freedoms a person has no matter how unattractive they are and no matter how unlikely a person is to choose them.

Gauging a person’s Point Freedom is different from gauging its intrapersonal distribution, that is, how a person’s freedom is distributed across a period of time. Imagine you are at a funeral and find yourself discussing with a friend how much freedom Steve, the person who died, enjoyed over the course of his life. This is a question about Steve’s intrapersonal distribution of freedom. Accordingly, we are not gauging how much Point Freedom Steve had at a specific time t but how much he had over different ‘points in time’ of his entire life. Maybe Steve started with a life with lots of freedom when he was young. He then signed up for the army for 10 years during which many of this options were restricted. After these 10 years, however, he received a good deal of money which he invested wisely to then become a wealthy man. To gauge how Steve’s freedom was distributed over the course of his life, we would have to see how much Point Freedom he had at the different stages of his life. Of course, the intrapersonal distribution of Steve’s freedom strongly depends on the decisions Steve made at various points in his life. Had he moved to North Korea, for example, his freedom at subsequent times would very likely have been much lower than it was in the actual world.

Just as Point Freedom, we can gauge intrapersonal distribution with different temporal extensions. The normatively relevant temporal extension, which I mostly focus on here, is freedom across whole lives or lifetime freedom as I will sometimes call it.

We can determine a person’s intrapersonal distribution ex post by determining how much freedom she had over some period in the past. But we can also try to estimate how much freedom she will have distributed over a future period. Using probabilities over what choices a person might make in the future, we can estimate a person’s intrapersonal distribution of freedom ex ante.

Let us return to the case of voluntary slavery. For simplicity, I represent this as a choice between the ‘slavery option’ D at t which will only lead to one future freedom at t + 1 and other options A, B, C, which will lead to more future freedoms. If I choose A at t, I will have nine options at t + 1. If I choose B, I will have seven. If I choose C, I will have five. (For simplicity, I assume that the options at t + 1 made available through choices at t do not overlap) (Fig. 1).

Fig. 1

Voluntary slavery example

A person’s level of Point Freedom at t is—so I have assumed for simplicity—the number of options at t plus the options that ensue at t + 1, so we get: 4 + 22 = 26. If we now follow Mill’s advice and remove the slavery option, we would reduce a person’s Point Freedom at t to 3 + 21 = 24. Therefore, by removing an option we have reduced Point Freedom at t.

However, calculating a person’s intrapersonal distribution of freedom ex ante is less straightforward. To do so, we take account of how likely a person is to choose particular options. Using these probability judgements, we then calculate how likely a person will have different levels of Point Freedom in the future.¹¹ Note that, to keep things simple, I for now assume that when we gauge the intrapersonal distribution, we simply use the sum of a person’s Point Freedom at different points in time within that period. In Sect. 4, I will relax this assumption and discuss different types of intrapersonal distributions.

Let us use some toy numbers to show how to calculate ex ante a person’s intrapersonal distribution. Assume we have reason to believe that P is equally likely to choose either A, B, C or D at t, so the probability that she chooses any option is p _i  = 0.25. If we now add her Point Freedom at t (where the temporal extension is t) to the Point Freedom she is expected to have at t + 1, then we get the following results:

• Intrapersonal Distribution ex ante (with D) = number of specific freedoms at t + expected number of specific freedoms at t + 1 = 4 + 0.25(9 + 7+5 + 1) = 9.5

Let us now remove option D, the slavery option. Again, let us assume that P is equally probable to choose any of the remaining three options (p _i  = 1/3).¹²

• Intrapersonal Distribution ex ante (without D) = 3 + 1/3(9 + 7+5) = 10

Therefore, if we remove the slavery option D, we expect the person to have more freedom aggregated across time than if we did not remove D. So, if we care about the sum of Point Freedom across time, we should prohibit voluntary slavery.

Increasing a person’s ex ante intrapersonal distribution of freedom is different from increasing her Point Freedom. The ex ante intrapersonal distribution takes into account the probabilities with which a person will have certain options in future periods where these probabilities (partly) depend on how she is inclined to act.

This gives us the following result: whether increasing the ex ante intrapersonal distribution of freedom requires removing options like ‘voluntary slavery’ depends on the following three factors:

• a. What is the agent’s probability distribution over the different options?

• b. To how many future options do individual options lead?

• c. What standard do we use for the intrapersonal distribution of freedom?

In Sect. 2, I introduced ‘freedom-based paternalism’ as the idea that we can sometimes justify interfering with people’s freedom, if doing so is necessary to safeguard greater freedom in the future. We now see that whether interference increases how much future freedom a person can expect to have depends on (a) the overall probability distribution, (b) an option’s ‘fecundity’—that is, how many future options it will lead to—relative to the fecundity of other options, and (c) the standard we use for the intrapersonal distribution. Therefore, we cannot say that to increase the ex ante intrapersonal distribution, any ‘low-fecundity’ option, such as smoking or even becoming a slave, should always be removed. If one is very unlikely to choose to smoke and/or choosing to smoke results in only a little less freedom than the other options, then, depending on one’s standard for intrapersonal distribution, removing the option to smoke would not increase the intrapersonal distribution of freedom. This point also helps us answer, or at least qualify, the objection raised by Arneson. He argues that increasing a person’s future freedom leads us onto a slippery slope, as it would imply banning things like fried food. However, whether this is really so for individual options depends on the three factors mentioned above. It is not clear, for example, that the freedom-reducing effect of removing different options to eat fried food would be offset by an increase in the intrapersonal distribution of freedom. Arneson’s slippery slope objection to freedom-based paternalism is far too quick.

I think the distinction also dissolves the paradoxical nature around voluntary slavery: previous to making any decision, prohibiting slavery does in fact make people less free overall in terms of Point Freedom (at the time of the prohibition). However, prohibiting slavery will typically lead to more freedom in the sense of a higher ex ante intrapersonal distribution of freedom (depending on the three factors mentioned above).

Before moving on, I should address the following worry. In Sect. 3, I stipulated that this article is about freedom as an opportunity-concept, or what is sometimes labelled ‘negative freedom’. Theorists working on this concept typically argue that freedom cannot depend directly on a person’s preferences (or her pro-attitudes more generally). Because theories that make freedom directly preference-dependent would imply that a person can make herself (more) free simply by changing her preferences without anything else changing in the outside world (Berlin 1969a, xxxviii; Carter and Kramer 2008). Direct preference-dependence would betray a conceptual confusion between freedom as an opportunity-concept and psychological freedom.

On my account, Point Freedom does not directly depend on a person’s preferences. However, we have seen that a person’s preferences do matter for our calculation of the intrapersonal distribution of Point Freedom (in influencing the probability distribution over options for ex ante calculations and in influencing past levels of Point Freedom for ex post calculations). Nonetheless, the idea of an intrapersonal distribution of freedom is not guilty of a conceptual confusion between freedom as an opportunity-concept and psychological freedom.¹³ For the intrapersonal distribution of freedom is not in itself a separate concept or conception of freedom. Freedom itself just is Point Freedom. Consider an analogy. Assume how much wealth a person has at any given point is determined independently of that person’s preferences (set aside complications regarding the behaviour-dependent nature of long-term assets and debt). Nonetheless, when calculating a person’s expected future wealth, her preferences should enter our calculations. For example, if a person prefers spending her life in an ashram over pursuing a career in investment banking, we have good reason to expect her to have less future wealth. But including preferences in our calculation of expected future wealth still leaves wealth itself a preference-independent quantity.

Similarly, a person’s preferences are clearly relevant when estimating future levels of Point Freedom without freedom itself being directly preference-dependent. For example, imagine your friend is planning a trip to North Korea to stage a political protest she knows will land her in prison. Or imagine someone who desires to do whatever is necessary to have low levels of Point Freedom (for example by getting herself incarcerated). In both cases, it would be irrational—even for negative freedom theorists—not to take into account the persons’ preferences when gauging her intrapersonal distribution of Point Freedom.

Of course, negative freedom theorists might have normative reservations about institutions or public policy makers trying to effect particular intrapersonal distributions of freedom. But such normative reservations are very different from the charge of conceptual confusion. While the aim of the present section has been merely analytical, I now turn to normative questions regarding the intrapersonal distribution.

4 Distributions of freedom across time

For simplicity, I have so far assumed that the intrapersonal distribution is simply the sum of Point Freedoms that fall within the relevant period. I now drop this assumption and discuss different procedures to distribute Point Freedoms across time. The question now is what kind of distribution of freedom across time we should promote as a suitable goal of public policy and legislation. I first briefly lay out (what are often taken to be) reasons to value freedom in the opportunity-sense. These reasons are relevant in the next steps of my main argument. The idea is that understanding what makes freedom valuable in general should also inform our discussion of the appropriate intrapersonal distribution of freedom. Note that, for the purposes of this article, I assume that freedom is valuable without defending this assumption here

4.1 The value of freedom

Freedom can be intrinsically or instrumentally valuable. I will not discuss whether freedom is intrinsically valuable or not. It is enough, if it turns out, as Ian Carter has argued, that freedom has non-specific value (Carter 1999, chap. 2). Freedom has non-specific value, if freedom is valuable over and above the goods that it facilitates in individual instances. Put a bit simplified, freedom is non-specifically valuable if having choice (or being from external constraints) matters over and above the value of outcomes of one’s choices. Let me briefly mention a few reasons to think that freedom is non-specifically valuable (not intended as an exhaustive list).¹⁴

The main consideration—which resonates with liberals in particular—is that freedom is necessary for individuals in their capacity as autonomous agents to pursue their own conception of the good. A number of more specific considerations support this more general thought. One championed by Mill for example, is that personal development and autonomy require experiencing different lifestyles. Freedom is required for personal development. Relatedly, Hurka (1987) argues that rejecting other options is necessary for autonomous agency.

We have further reasons to value freedom as a social value. In ‘real-life’ situations in which political decisions take place, we do not know everyone’s preferences nor do we always know what is good for them. Moreover—and this relates to the liberal thought above—we think it important to keep at last some neutrality about conceptions of the good. Allowing persons a range of choice is necessary, if we want to respect individuals, their decisions and underlying conceptions of the good. We need not understand respect here as a source of deontological duties (though we could). Instead, we can hold that personal freedom is necessary to facilitate valuable, respectful forms of social interaction and cooperation.

Another reason to value freedom—and one which is open to theories with very different ideas about what goals institutions should pursue—comes from personality change (Carter 1999, chap. 2). People like different things at different stages in their lives. As a child I hated olives and coffee but like both very much today. Freedom is also about safeguarding an adequate range of choice to allow for personality change (whereby I mean to include both smaller changes in tastes and preferences as well as more significant changes in character). This, in turn, is necessary for a person’s well-being across time given uncertainty about future preferences and likings. I think a failure to appreciate the normative importance of personality change comes from what has been called the ‘end of history illusion’. People typically acknowledge that their personality has usually changed significantly over the last 10 years. Yet people of all ages tend to systematically underestimate the extent to which their personality will change in the next decade. ‘History, it seems, is always ending today’ (Quoidbach et al. 2013, 98).

Now, given these various ways in which freedom is often considered valuable, how should it be distributed within lives?

4.2 The starting gate view

According to the Starting Gate View, we should concern ourselves exclusively with the level of Point Freedom a person has at a ‘starting gate’. Such a view implies that we should not try to facilitate any particular intrapersonal distribution after the starting gate. Once persons have passed the starting gate, we should not prevent persons from making choices, even if those choices leave them with very little Point Freedom (as in the case of voluntary slavery, for instance).

I think the main normative consideration motivating the Starting Gate View is respect for individuals and their choices. A commitment to preserving or promoting freedom, on such a view, is about respecting people and their decisions. Accordingly, if persons genuinely start with many alternatives and make their own choices, we should just leave them to it. The Starting Gate View also blocks attempts to justify paternalistic policies on the grounds that they redistribute freedom favourably across time. This anti-paternalist stance would also mean that a freedom-based theory of distributive justice would have to focus exclusively on the social distribution of Point Freedom at one particular point in time. The Starting Gate View would, for example, favour equal basic capital over equal basic income.

However, there is decisive reason against the Starting Gate View.

An initial problem is to find a temporal point at which to locate the ‘starting gate’. Starting at birth is too early:

Jack and James: in a world in which schooling is optional, Jack and James start with exactly the same level of freedom at birth (the same ‘starting gate’). When Jack is ten he starts skiving off school most days which is why he later misses out on a degree. James attends school regularly, graduates and goes on to university. When James turns 26, he is wealthy and has lots of options. Jack’s lack of a degree forces him to lead a very simple life with much lower income and shorter life expectancy.

On a version of the Starting Gate View that locates the starting gate at birth, a concern for Jack’s and James’s intrapersonal distribution of freedom does not speak for compulsory schooling. Even though Jack will have much less future freedom than James—and much less than he could have—the Starting Gate View would not favour compulsory schooling. We might think that, intuitively, this speaks for a later starting gate. Rather than at birth, we should set the starting gate at a time when persons typically become sufficiently autonomous to take important decisions. Of course, finding such a point is not trivial and will always involve some arbitrariness. Among other reasons, growing into an autonomous agent seems more like a gradual process than a binary one.

But even if we set the starting gate at a later time, the Starting Gate View still fails, because it does not account for some of our central reasons to value freedom.

One of the aforementioned reasons to value freedom was personality change. Freedom can facilitate a good life in the face of changing personalities across time. Safeguarding a suitable range of freedom in the future might sometimes be sufficient justification to remove low-fecundity options.¹⁵ The Starting Gate View fails to account for personality change. It pays no attention to uncertain interests of future versions of existing persons and does not prevent a person from abridging all those opportunities her future self might come to like or value.

Second, as mentioned above, freedom is required for personal development. Here, the Starting Gate View runs into problems too. The Starting Gate View ensures that agents have good opportunities for personal growth and development early on. However, individuals that start off choosing options with very low fecundity will lack opportunities for personal development later on. If we care about personal growth and development, it seems we should also care about a person’s future possibilities for personal growth and development.

Overall, the Starting Gate View is implausible, as it fails to account for some of our reasons to value freedom. If freedom is to figure as a suitable goal of public policy and legislation, focusing only on a person’s starting gate freedom is not enough. Caring about people’s freedom sometimes gives us reason to prevent very low levels of freedom, even if such low levels are the result of people’s own choices.

4.3 The maximising view

Given that the Starting Gate View seems implausible, should we, through our institutions and public policy, aim to maximise the sum of Point Freedom instead? If yes, then a concern for a person’s lifetime freedom would justify a number of paternalistic policies purely on the grounds that they would increase the sum of future freedom (provided certain conditions hold, as discussed in Sect. 3).

To assess the Maximising View, we need to specify what precisely we should maximise. Here, alas, we need another distinction. Aggregating freedom over time is more difficult than aggregating most other goods. Consider, for example, most standard views of wellbeing. When we consider how much wellbeing a person enjoys at a particular point in time, we focus on aspects of that person’s life at that point in time (for example, how happy that person is at that time, how much pleasure she is experiencing, whether she is satisfied and so on). Freedom, on the other hand, always ranges into the future (remember the existence-time and actualisation-time distinction). This leaves us with different temporal extensions over which we could aggregate. When we adopt an aggregative conception of the appropriate intrapersonal distribution of Point Freedom, we can either use a Discrete Segments Account or a Fresh Starts Account (Carter 2013, 139–40). When we use the Discrete Segments Account, we aggregate Point Freedom whose temporal range is only up to the next temporal point at which we measure Point Freedom again. Say we want to know how much freedom a person has in the next five years and we do so by aggregating her Point Freedom every year (I only assume for this example that segments are 1 year long).

On the Discrete Segments Account, we would add discrete segments in the following sense: we first calculate the Point Freedom that includes all freedoms whose actualisation-times lie in t, then we add this number to the Point Freedom that includes all freedoms whose actualisation-times lie in t + 1, then we add the Point Freedom with all freedoms whose actualisation-times lie in t + 2 and so on (Fig. 2).

Fig. 2

The Discrete Segments Account

Contrast this with a Fresh Starts Account. Here we aggregate over Point Freedom at different stages of a person’s life calculating each time her freedom from that point to her death (if we calculate lifetime freedom) (Fig. 3).

Fig. 3

The Fresh Starts Account

Accordingly, we would first calculate a person’s Point Freedom at t by aggregating all freedoms with an existence-time t and with actualisation-times between t and t + 5. Then we would add this to the Point Freedom at t + 1, which we get by aggregating all freedoms with an existence-time t + 1 and actualisation-times between t + 1 and t + 5. And so on.

Let us see whether the Maximising View works with either the Discrete Segment Account or the Fresh Starts Account. Let us start with the Fresh Starts Account.

When combined with the Fresh Starts Account, the Maximising View seems problematically biased towards ‘end-heavy’ distributions of actualisation-times. Let me explain what I mean. Imagine you could substitute a person’s freedom to do a particular act soon with a freedom to do such an act, or a similar one, at the end of her life (assume also that actualising either of these freedoms does not provide the person with additional freedoms). Consider:

Coupon Booklet: we are putting together a booklet with coupons for our friend Irina who is twenty-five years old. The booklet features coupons that give her the freedom to go on a free holiday, visit the opera or eat at a fancy restaurant. When setting the terms of the coupons, we have different options: we can make all coupons redeemable during the next five years, make each coupon redeemable during a different five year stretch, or make them all redeemable during the last five years of Irina’s life.

On the Maximising View, we furnish Irina with highest lifetime freedom, if we make all coupons redeemable during the last 5 years of her life (setting aside uncertainty), because this way they exist the longest. As a result, these actualisable options would be counted as separate freedoms in every measurement period. More generally, and other things being equal, the later an option’s actualisation-time, the longer it exists and the greater its contribution to the sum of Point Freedom across time.

Coupon Booklet shows that the Fresh Starts Account is biased towards freedoms with late actualisation-times (other things being equal). Practically more relevant, such a bias will also countenance relatively frequent interventions to curtail freedoms that are freedom-reducing (again, depending on relative fecundity and an agent’s probability distribution over the different options). Combined with the Fresh Starts Account, the Maximising View would advise curtailing low-fecundity options to safeguard future freedoms comparatively often. For on the Fresh Starts Account it becomes imperative to keep freedoms as long as possible, seeing that they are counted in every measurement-period.

Is this implausible? On the one hand, one might think that keeping one’s freedom involves keeping one’s options open. In this sense, the later your freedoms’ actualisation-times, or the less likely you are to choose early low-fecundity options, the more freedom you will have across your life. But such thinking might become implausible, if carried ad absurdum. Taken to extremes, we could imagine lives in which people can actualise a great deal of freedoms at the end of their lives but can actualise very few freedoms (with relatively low fecundity) beforehand.¹⁶ Such an implication, I think, conflicts with some of our reasons to value freedom. The bias towards late actualisation-times might significantly reduce a person’s opportunities to pursue her own conception of the good. Most conceptions of the good life seem to imply being in a position to pursue projects and make actual choices at earlier and not only at later stages in life. Such choices require freedoms with earlier actualisation-times, oftentimes including those that do not expand or even preserve all of a person’s future freedoms.¹⁷ Moreover, a very ‘end-heavy’ distribution of actualisation-times does not always sit well with a concern with personality change. People change throughout life and this might require being able to actualise many freedoms throughout one’s life (again, even low-fecundity ones). Moreover, personality change typically slows down somewhat in old age, which further speaks against a strong bias towards late actualisation-times (Caspi and Roberts 2001, 50–51). Finally, a person’s prospects to experience different lifestyles are diminished, if too many of her freedoms can only be actualised towards the end of her life or if she is prevented too often from choosing options relatively low in fecundity. The Starting Gate Account’s bias towards late actualisation-times and towards curtailing low-fecundity options does not sit well with our reasons to value freedom.

A related, more fundamental, worry is that ‘slicing up’ a person across time to then maximise the sum of freedom contradicts the liberal notion that freedom is a value for respect-based social institutions. I do not attempt an analysis of ‘respect’ here. But a pre-theoretical understanding of respect suggests that we accept some or most of an individual’s decisions and understand her to be a person pursuing plans and taking responsibility for her decisions across time. Maximising the sum of a person’s Point Freedom across time will often lead to tension with such respect for individuals and their choices.

This example brings together the two aforementioned problems.

Juliette: Juliette is in her thirties. She is a ‘savoir-vivre kind of person’. She loves opera, travel and good food. She would like to spend a good portion of her income on these enjoyable pastimes.

On the Maximising View, a concern for Juliette’s freedom implies we ought, other things being equal, prevent Juliette from spending her money as she sees fit. More money means more freedom (on most accounts of freedom).¹⁸ But Juliette wants to use her money for a type of consumption that does not preserve her wealth. Instead, she should spend her money on things—such as real estate in upcoming neighbourhoods—that preserve or even increase her spending power. Juliette could maximise the sum of Point Freedom by spending her money in ways that expand her options. So, to maximise Juliette’s sum of freedom across time, we might have to prevent her from spending money on her favourite pastimes (depending on the probability with which Juliette is going to choose a ‘savoir vivre kind of lifestyle’ and depending on its relative fecundity). As a result, Juliette will be able to actualise high-fecundity options, such as buying the right real estate, earlier on. Much later in life, she will then be able to spend the saved money on low-fecundity options. However, forcing Juliette to save her money this way seems, first, to exhibit an excessive bias towards an end-heavy distribution of actualisation-times and, second, sit uneasily with respecting Juliette and the choices she makes as part of her conception of the good.¹⁹

I think the Maximising View combined with the Fresh Starts Account is not plausible. Is the Maximising View more plausible when combined with the Discrete Segment Account? At least two considerations speak for such a combination.

First, freedoms enter our calculations only once within their respective segments, even if their actualisation-times are at the end of a person’s life. Therefore, unlike the Fresh Starts Account, the Discrete Segments Account is not biased towards late actualisation-times.

Second, the Discrete Segment Account is somewhat less speculative in its calculation than the Fresh Starts Account. Because predicting what options a person can actualise within limited time-segments is easier than gauging all of the options she can actualise between now and her death.

However, despite these advantages, the Maximising View combined with the Discrete Segment Account is still implausible. Consider first these three general reasons against the Discrete Segments Account.

First, Carter argues²⁰:

The discrete-segment version is the simpler of the two, but it is also the less plausible one. After all, the real degree of a person’s freedom at any given time surely depends on her freedom to bring about events that are temporally located at any subsequent time within her expected lifetime. Focusing at t1 only on those freedoms to perform actions located between t1 and t2 seems plausible only if we think of the life of the person under consideration as itself coming to an end at t2. (Carter 2013, 139–40)

When determining a person’s Point Freedom, it seems counterintuitive to focus exclusively on options a person can actualise within a limited temporal segment. Rather, all of her options, even those with later actualisation-times, seem relevant. But if this is so for Point Freedom, the Discrete Segments Account involves a theoretical disconnect between Point Freedom and its intrapersonal distribution.

Second, being able to make decisions that influence one’s life a few or even many years down the line matters, if we care about how people can exercise agency across time. Indeed, many of our central life projects involve long-term options, for example long-term financial planning, the kinds of professions we choose, the family plans we make, and so on. The Discrete Segments Account seems to lose relevant information about the extent to which agents are able to shape their lives across a longer time horizon.

Finally, we might worry about the arbitrariness of fixing the length of these temporal segments. Should the segments be a day, a year, 5 years, 10 years etc.? That this is an open question is not a decisive reason against the Discrete Segments Account. But it is a legitimate worry that no straightforward way to answer this question suggests itself.

Besides these general worries, consider a problem particular to the combination of the Maximising View and the Discrete Segments Account.

As said above, the Discrete Segments Account avoids a bias towards late actualisation-times. But a different problem emerges. Compare the following distributions of freedoms across segments (t to t + 5) within a person’s life (Table 1).

Table 1 Counterexample to the Maximising view with the Discrete Segments Account

The Maximising View combined with the Discrete Segments Account simply adds up the numbers in each segment. Accordingly, there is as much lifetime freedom in D ₁ as in D ₂ , as the sum is 605 for both. For D ₃ , the sum is only 604. So, on the Maximising View, the person has more lifetime freedom in D ₁ and D ₂ than in D ₃ . On the Discrete Segments Account, it is irrelevant how freedoms are distributed across time. This is problematic. If we care about freedom, it seems we have reason to disfavour intrapersonal distributions, such as D ₁ or D ₂ , in which the person can make more or less no choices for long stretches of her life. The more specific reasons to value freedom mentioned above reinforce that point. For example, to account for personality change and personal development, we have reason to secure a certain range of actualisable options within segments across various stages of a person’s life.²¹

In conclusion, the Maximising View is implausible, both when combined with the Fresh Starts and the Discrete Segments Account.

4.4 The sufficiency view

We could take inspiration from Dworkin (1982, 81) and hold that ‘[i]n the realm of choice, as in all others, we must conclude—enough is enough’. Instead of maximising freedom across time, maybe we ought to ensure that people have enough freedom at all times. Legislation and public policy should make sure a person’s freedom remains above a certain threshold at all times.

However, the Sufficiency View alone is implausible. Compare two lives. Life 1 starts well above the threshold and remains high throughout. Life 2, on the other hand, is always just slightly above the threshold (Fig. 4).

Fig. 4

Counterexample to the Sufficiency View

If we adopt a person’s freedom across time as a suitable goal for institutions and public policy, it seems we have good reason to facilitate lives that are more like Life 1 than Life 2. A normative political theory—for example, a theory of justice—that would treat them equivalently would miss something important. For this reason, Ian Carter suggests combining the Sufficiency View with the Starting Gate View (Carter 2013). Let us call this the Combined View:

• C1: allow only those lives in which Point Freedom is above the threshold at all times.

• C2: out of these, choose the life with the highest starting gate freedom.²²

What speaks for the combined view?

First, the Combined View seems to take on board the idea that one aspect of valuing a person’s freedom is about respecting her decisions. Unlike the Maximising View, the Combined View gives people a good starting position and then largely refrains from trying to effect a specific distribution of people’s freedom across time. Thus, the Combined View does not require, for example, that we put a prohibitive cap on Juliette’s expenses on good food, travel and opera.

Second, unlike the Starting Gate View, the Combined View takes into account personality change, personal development and autonomy—at least to some extent. Future versions of current people have interests and one of them is freedom. Through securing enough freedom in the future, we try to prevent that future versions of a person lead lives with very little Point Freedom.

The Combined View seems more ‘balanced’ than the Starting Gate and the Maximising View. However, Carter’s Combined View—the only systematic account offered to date—faces a number of theoretical challenges. The sufficiency criterion C1 invites the question as to where such a level of sufficiency is to be had—a notoriously difficult problem for any threshold view. Moreover, at which temporal point in a person’s life should we locate a person’s starting gate for our criterion C2? I doubt there are ‘correct’ theoretical answers to both questions. But even if Carter can provide satisfactory answers to these questions, other more troubling problems already make the Combined View implausible, as I argue now.

A major problem for the Combined View is to deal with situations in which people have very few freedoms they can actualise close in time. Consider again the distinction between the Fresh Starts and the Discrete Segments Account. If we understand the sufficiency criterion C1 according to the Fresh Starts Account, the Combined View runs into problems similar to those seen in connection with the Maximising View. In some scenarios, a person might have many freedoms with actualisation-times much later in life but very few freedoms with close actualisation-times. Imagine, for example, a person is thrown in jail for a long time but then compensated with numerous freedoms with actualisation-times much later in her life. She has extremely few freedoms with actualisation-times that fall within her time in prison and would thus be able to make very few choices during that time (though she has numerous freedoms with actualisation-times after her time in prison). On the Fresh Starts Account, she would not have insufficient freedom during her time in prison. For she has a lot freedoms with existence-times that fall within her time in prison. Unfortunate for her, most of those freedoms can only be acted on—or actualised—much later in life. Therefore, if C1 is understood along the lines of the Fresh Starts Account, C1 would condone imprisonment that is compensated with freedoms with late actualisation-times. I think this result should strike us odd, particularly if we are concerned with making sure people have sufficient choices throughout their lives. As before, the distribution of actualisation-times seems to matter too. This is a serious issue for the Combined View as presented by Carter, because Carter favours the Fresh Starts Account.

Can we circumvent this problem, if we understand the sufficiency criterion C1 according to the Discrete Segments Account? Remember that I discussed some general problems with the Discrete Segments Account in Sect. 4.3. These problems would apply here too. Of these, setting the length of the segments will pose a particularly important challenge. Consider:

Flying: I am on a 24 h flight from Europe to Australia. During the flight I have extremely few freedoms I can actualise.

Flying shows that our segments should not be too short. If we used very short segments, such as two days, then C1 would rule out the option to fly to Australia, because I would have insufficient freedom during the flight (remember the Discrete Segments Account only includes freedoms with actualisation-times that fall within the specific segments). If segments are longer, say one year each, C1 would typically not discountenance the option to fly long distance. For the options I can actualise after my flight should typically move me above the threshold for the segment as a whole. The challenge then is to fix the length of segments in a theoretically non-arbitrary fashion.

More problematically, a different example shows that temporary lack of freedom causes trouble for the Combined View both when coupled with the Fresh Starts and when coupled with the Discrete Segments Account. Both views would rule out lives in which insufficient freedom comes at the end of a person’s life.

Football: Carla is very particular about sports. The only sport she likes to play is football. If she is allowed to play football she will live to become 80 years old. However, because Carla has a hereditary illness that will break out by the time she reaches 78, she will be bedridden for the last 2 years. If she is not allowed to play football, Carla will choose a very unhealthy lifestyle which will result in a fatal heart attack when she is 55.

Assume Carla’s condition between 78 and 80 grants her insufficient freedom while her level of freedom during her unhealthy life is always above the threshold.²³ According to the Combined View, a concern for Carla’s intrapersonal distribution of freedom means we should prevent Carla from playing football, because this would result in a life in which Carla never falls below the sufficiency level. This is so on both the Discrete Segments and the Fresh Starts Account. This prescriptive conclusion, however, seems implausible. For Carla would still lead a life with high levels of freedom between the age of 55 and 78 and preventing Carla from playing football means she would be missing out on those years.

I conclude that the Combined View is implausible.

So, where does that leave us? Various normative reasons to value freedom pull in different directions. For example, freedom is valuable, because it can secure good options in the face of nearly everyone undergoing personality changes at various points in their life. But freedom is also important for respect-based interpersonal relationships. Moreover, we have seen that it is not clear whether we should apply the Discrete Segments or the Fresh Starts Account—different cases pull in different directions. We have therefore not found a straightforward intrapersonal distribution of Point Freedom that does justice to the various normative considerations typically connected to freedom. Neither a Starting Gate, Maximising, Sufficiency nor Combined View does the job.²⁴

5 Distributive justice and paternalism

The prima facie plausible candidates for the intrapersonal distribution of freedom all failed to do justice to our various reasons to value freedom. Of course, this is not an impossibility theorem—someone else might find a standard for distributing freedom intrapersonally that solves these problems. Nonetheless, let me mention some of the implications this (potentially provisional) result has for freedom-based paternalism and distributive justice.

5.1 Paternalism

Freedom-based paternalism holds that some paternalist interventions can be justified, if they are necessary to improve someone’s intrapersonal distribution of freedom. On the one hand, my discussion brought out that a person’s freedom across time is often a salient consideration for paternalism. Distinguishing between Point Freedom and the intrapersonal distribution of Point Freedom showed how paternalistic interferences are often not just simple infringements of freedom but redistributions thereof across time. Paternalist interventions, such as the prevention of voluntary slavery, might reduce Point Freedom at some point but increase its intrapersonal distribution overall. Similarly, a law to make seatbelts compulsory might prevent freedom-reducing, incapacitating accidents. Preventing drug addictions might put people in a position to enjoy more future freedom. Compulsory schooling might be a suitable means to safeguard a person’s future freedom as shown in Jack and James. As seen in Sect. 3, whether a concern for freedom across time implies removing a freedom-reducing option depends on that option’s fecundity relative to other options, the probability distribution of an agent’s decisions over these options and on one’s standard for intrapersonal distribution.

On the other hand, however, freedom-based paternalism appeared to have significant justificatory limitations. Because we have not found a plausible standard for distributing freedom intrapersonally, freedom-based paternalism does not by itself provide a clear, principled justification for paternalistic policies. How far is this a problem?

First, in some situations, a paternalistic interference might be necessary to safeguard a certain level of freedom across a number of standards for intrapersonal distribution. For example, some interferences might be called for on both the Maximising and the Sufficiency View and on different combinations with the Discrete Segments and the Fresh Starts Account.

Second, freedom-based paternalism seems to provide one but clearly not the only relevant consideration when discussing paternalism. We might need other principled criteria, deontological constraints for example, to construct a plausible ethical theory of paternalism. Such a comprehensive theory would put us in a better position to determine which instances of paternalism are justified and which ones are not. So, by itself, freedom-based paternalism often fails to provide clear principled guidance. But this does not mean, of course, that a principled basis from which to justify certain paternalist interventions—and rule out others—is not to be had in general.

5.2 Distributive justice

The implications for freedom-based theories of distributive justice are, I think, quite serious. Without a plausible standard for the intrapersonal distribution of Point Freedom, it remains unclear what it means to distribute freedom in a just way between persons. For example, many authors think that justice requires that we aim for greatest equal freedom (or, relatedly, for greatest equal opportunities). What does this mean exactly? A Starting Gate View would answer this question very differently than a Maximising View, for example. This is particularly problematic for theorists—for example, left-libertarians such as Steiner and van Parijs—who hold that freedom is the only relevant distribuendum.

What about theories of distributive justice that take freedom (or opportunities) to be one amongst other relevant distribuenda? Such theories might simply pick one of the intrapersonal distribution principles discussed here and add further principles. These other principles might feature different distribuenda, could be deontological side-constraints or something else. This way, we might arrive at a set of principles that together gives us plausible judgements in cases in which any of the principles of the intrapersonal distribution of freedom led to implausible judgements on their own. Together, such principles might give us a plausible extension. However, if the arguments of this article are successful, such a strategy will at least be theoretically unsatisfactory. While we might arrive at a plausible extension, we might not do so for all the right reasons. For the chosen principle for the intrapersonal distribution of freedom will still fail to adequately reflect our different reasons for valuing freedom. For example, imagine our pluralist view would assume the Starting Gate View of freedom and add a proviso that tells us to secure the social and material conditions minimally necessary to safeguard a person’s dignity. Now, the latter principle would presumably imply outlawing slavery contracts to prevent undignified social relationships. But this theory will still involve an inadequate picture of what it means to care about a person’s lifetime freedom. For it is not only because of a person’s dignity that we should prevent voluntary slavery, a concern for personal freedom seems to require that too. And the Starting Gate View fails to capture this point.

This article thus ends with an unresolved problem. It is intuitive to believe—and often believed—that freedom is an important distribuendum in a theory of distributive justice. Moreover, it seems rational to not only care about one’s current freedom but also one’s future freedom. However, it is not clear what precisely follows from such a commitment, if we do not answer the question as to how freedom should be distributed across lifetimes. And none of the suggestions discussed here provides a plausible answer.

The negative result in this article leaves us with the following options (and possibly others).

First, we might continue searching until we find a plausible standard for the intrapersonal distribution of freedom. (As David Hasselhoff sings: ‘still the search goes on’.)

Second, we might call off the search and be more sceptical about the power of freedom-based views of distributive justice to generate clear prescriptions. Those who are critical of the role of freedom in normative political theorising in general will see this as an additional objection to freedom-based views of anything. Others might want to continue endorsing freedom as a relevant distribuendum and assume one of the principles of intrapersonal distribution of freedom discussed here but also add further principles to arrive at intuitive all-things-considered judgements. However, as noted before, while this might give us intuitive judgements in individual cases, this option is theoretically unsatisfactory inasmuch as our principle of freedom therein would not be fully plausible in itself.

Finally, we might interpret the absence of a clear intrapersonal principle as just another example of the limitations of normative political theorising in general: normative political theorising will always leave us with significant indeterminacy; it is the task of political theory to reduce such indeterminacy, not to eradicate it.

I leave it to the reader to choose from these options.