1 – Introduction

1Many of us care about future generations’ well-being. We are uneasy with the prospect of leaving a world with depleted resources, a degraded environment or an unbearable public debt to future generations. However, if future generations’ well-being is what matters, it is possible for them to learn to live well in such a world. Human preferences are malleable. We can shape them so as to prevent feelings of deprivation. This is an option welfarist accounts of intergenerational justice must envisage. Although I do not think subjective welfarism is the most plausible moral theory, I believe it is important to investigate its implications in the intergenerational realm for the following reason. Exploring the implications of subjective welfarism in the intergenerational realm directly highlights difficulties any theory of justice that gives some credibility to the individual’s evaluative perspective on how well off she is (that is, to “preferences”, “desires”, “ambitions”, “expectations” or “conceptions of the good” for instance) may face.

2Welfarism claims that the well-being of future generations matters. Well-being may be defined in an objective or in a subjective way. Subjective accounts of well-being involve specific difficulties in the intergenerational context, because each generation has the power to shape the content of the next generation’s preferences through education. As well-being does not hinge only on income and resources (Easterlin 1973), one might be equally well-off with fewer resources than another. Ceteris paribus, if subjective well-being is all that matters, then we might be allowed and even required to inculcate cheap or easily satisfied preferences in children (Barry 1997, 102; Gosseries 2001, 341; Page 2007, 455; Bykvist 2009; Lippert-Rasmussen 2012). Many find this conclusion counter-intuitive. Some fear that this would allow educators to teach children to adjust their desires to worthless circumstances (Bykvist 2009) such as a degraded environment (Page 2007). Let us call this objection the Objective Good Objection. Another concern is the compatibility of cheap preferences engineering with the development of autonomy. This is the Autonomy Objection. Finally, if cheap preferences engineering is permissible, this would allow the current generation to go ahead and deplete as much resources as it sees fit to satisfy its (expensive) preferences (Lippert-Rasmussen 2012). This counters the intuition that it is praiseworthy to save for future generations, which is captured in the public debate by the ideal of sustainability. I term this objection the Fairness Objection.

3The aim of this paper is to assess whether intergenerational welfarism can address these objections. However, I shall not discuss at length the Objective Good objection, because there is by definition no way for a subjective account of well-being to accommodate it, except by not being a subjective account of well-being [1]. I will rather focus on the Autonomy and on the Fairness objection in order to determine whether or not educational policies aiming at inculcating cheap preferences can accommodate them. As we will see, the answer to both objections depends on the kind of methods we choose to inculcate cheap preferences in children. In this paper, I will argue that cheap preferences engineering is compatible with the development of autonomy as an end-state as opposed to autonomy as a precondition. I shall even argue that, in most circumstances, it is better to equip the younger generations with autonomy-related skills to secure their well-being. Then I will draw the distributive implications of the discussion of autonomy. If, because they are autonomous, future generations could happen to revise their initially cheap preferences, then the current generation might not be allowed to deplete as much resources as it can.

4The paper is structured as follows. Section 2 outlines the main features of the account of intergenerational justice I discuss. Section 3 outlines the prudential argument in favour of cheap preferences cultivation. Section 4 shows that autonomy is compatible with cheap preferences cultivation, and should perhaps be preferred to other methods. Section 5 draws the distributive implications of an autonomy-compatible way to engineer cheap preferences. Section 6 concludes.

2 – Equal prospects for well-being for future generations

5To get started, let us outline the main features of the account of intergenerational justice I shall discuss in this paper. An account of intergenerational justice is concerned with the just distribution of advantages between generations rather than between individuals within a same generation. As any theory of distributive justice, this account specifies a pattern (a distributive rule) and a metric (a currency to measure the advantages to be distributed). A discussion of the merits of the various possible patterns of intergenerational distribution is beyond the scope of this paper [2]. In this paper, I will take the pattern of intergenerational distribution that claims that each generation should leave at least as much as it received from the preceding one. This is a cleronomic pattern of distribution, since the level of intergenerational savings depends on what each generation inherited from the preceding one (Gosseries 2011). Note, however, that any distributive pattern that amounts to require to secure a level x of subjective well-being for the next generations is concerned by the discussion of cheap preferences formation that will follow. The metric I use is the metric of “opportunities” or “prospects for well-being” [3]. “Well-being” refers here to what makes an individual life a good life, according to the individual’s standards. The paper will take well-being to be preference satisfaction. I define preferences in the following way: a preference is a judgement that a state of affairs is better than other ones accompanied by the desire that this state of affairs obtains. Well-being occurs when one’s preferences are being satisfied.

6Now, individuals do have preferences for things of all sorts, including for what happens to other. These other-regarding preferences may be motivated by ethical values, altruism, paternalism, but also by envy or even perversity. In this paper, I will only consider self-regarding preferences, that is, preferences for what happens to the agent herself. This is out of the liberal intuition that preferences over other people’s private sphere are nosy and illegitimate (Barry 1986; Dworkin 1981). Of course, the distinction between both kinds of preferences can be blurred in practice. For example, many consumption choices are motivated by envy rather than by prospective enjoyment. But we should not confuse preferences for what affects (in a broad sense) the agent herself but motivated by positional concerns (e.g. having more than others) with preferences for what affects others. The distinction between self- and other-regarding preferences is not based on the intentions or the motivations behind the preference. It is based on the practical consequences of satisfying such a preference. To a certain extent, if a social planner satisfies the Smiths’ preference for having a more expensive car than the Jones, she does not interfere with the Jones’ privacy (though she might interfere with their possible own desire to compete with the Smiths). But if a social planner satisfies the Smiths’ preference for the Jones to be converted to a specific religion, she interferes with the Jones’ basic liberties.

7To avoid common pitfalls, let us assume that preferences are an accurate description of an individual’s well-being if they are (i) her true preferences and (ii) her rational preferences. True preferences are not always revealed by an individual’s statements, behaviours or choices (Sen 1977). Many factors prevent individuals from revealing their true preferences: material constraints, false beliefs, social pressure, to mention a few. The revealed preference account is thus a misleading account of well-being. Well-being is the satisfaction of an individual’s true preferences. But even our true preferences can be mistaken. I take thus preferences to be also rational. I follow here Arneson’s definition of “rational preferences” as “second-best informed preferences”: rational preferences are “those I would have if I were to engage in a thoroughgoing deliberation about my preferences with full pertinent information, in a calm mood, while thinking clearly and making no reasoning error.”. “Second-best” rational preferences should be favoured because “thoroughgoing deliberation” and “full pertinent information” are very demanding conditions for most of us. Hence second-best rational preferences fulfil the criteria listed by Arneson, but are reconsidered by taking into account the costs of information in terms of effort, resources and time (Arneson 1989, 83).

8This paper is concerned with opportunities or prospects [4] for well-being, as opposed to outcomes. Two generations enjoy equal prospects for well-being if the array of choices each of them faces are equivalent in terms of well-being. Suppose two successive generations, G1 and G2. If we constructed G1’s and G2’s decision tree, these trees providing all the complete life plans the members of each of these generations can achieve (on a per capita basis), and if we added the degree to which these individuals’ expected preferences will be satisfied in each of these lives, the sum should be equivalent for G1 and for G2. Why “prospects” rather than “outcomes”? First, since some of the conditions future generations will face are uncertain, it is better to provide them with more than one alternative. Second (I shall come back to this point in section 4), being able to choose whether, how and to what extent to achieve one’s well-being is valuable. Figure 1 displays the decision trees a representative member of G1 and of G2 could face if equality of prospects for well-being between generations was achieved. The sum (measured in terms of subjective welfare) is 40 for both. It is up to them to choose a life plan that brings a level of welfare of 3, 22 or 13, for example. Note that the life plan #1 has not been preserved for G2, but this is not a problem as long as the array of options she faces (which comprises a new life plan, #4) is equal to G1’s in terms of expected subjective welfare. Note also that the same life plans (#2 and #3) do not bring the same level of satisfaction for G1 and G2, which could be due to the evolution of preferences, values and conditions.

Figure 1

Equality of prospects for well-being between two generations (G1 and G2)

Equality of prospects for well-being between two generations (G1 and G2)

9Equality of prospects for well-being as a criterion of intergenerational justice inevitably raises the issue of how to make interpersonal comparisons of levels of well-being across generations. Harsanyi’s solution to the problem of interpersonal comparisons of well-being between contemporaries is quite appealing (Harsanyi 1955). Harsanyi starts from the fact that (most) human beings can empathize with others. Harry empathizes with Sally when Harry imagines rather accurately what it is like for Sally to have Sally’s (and not Harry’s) preferences satisfied. An individual can thus have empathetic preferences, that is, she can use a common standard to compare the levels of well-being of different individuals, including herself, in various states of affairs. According to Harsanyi, there is a broad consensus on the standards different individuals use to make interpersonal comparisons of well-being. We sufficiently understand each other when we talk about our preferences, others’ preferences and what it is like to satisfy them. Now, it would be more adventurous to suggest that we could reach such an agreement with distant future people. But a way out of the problem of interpersonal comparisons across generations is to construct a standard for overlapping generations. Suppose G1 overlaps with G2 for some time, and G2 with G3, whereas G1 does not overlap with G3. G1 could certainly empathize with G2 and agree with G2 on a common standard to compare their levels of well-being. And the same could hold for G2 and G3. And so on [5].

3 – The prudential argument for cheap preferences engineering

10Issues of justice between generations arise because the resources we need to satisfy our preferences are scarce. Prospects for well-being depend on resources, though to some extent. Well-being does not hinge solely on income, as shown by the well-known Easterlin paradox (Easterlin 1973). There are therefore two main directions policy-makers can take in order to promote future generations’ prospects for well-being: one is the “Resource Approach” and deals with scarcity by modifying either the allocation or the quantity of resources; the other is the “Taste Approach” and deals with scarcity by influencing preferences themselves (Otterholt 2010) [6]. The Resource Approach itself comprises two different kinds of policies, that is, distributive policies and productivity-oriented policies (they can be coupled). Distributive policies would require each generation to limit its prospects for well-being (to the extent that these prospects depend on resource consumption) in order to secure enough resources for the next generation(s) to enjoy the same amount of prospects (or more). Productivity-oriented policies would count on innovations and technological progress to improve productivity and thus future generations’ prospects for well-being without limiting the current generation’s ones [7]. But the natural resources needed to produce the desired goods are not endless. Moreover, in the light of population growth, technological progress might not be sufficient to secure future generations’ prospects for well-being. Indeed experts from international institutions now urge us to change our consumption patterns. The UN Agenda 21 prescribes “developing national policies and strategies to encourage changes in unsustainable consumption patterns” (United Nations Environment Programme 1992, chap. iv). The UN has reiterated this recently, for instance in the Third International Conference on Small Island Developing States (UN 2014, Article 68 and 69). In the light of this, the Taste Approach seems more reasonable, since it does not aim to increase the quantity of resources future generations receive. The Taste Approach aims to affect next generations’ preferences, so that they become easier to satisfy even with less resource.

11Welfarist theories have tended to take preferences as given. This stance relies either on the empirical assumption that preferences are stable or on the normative assumption that individuals should be considered as sovereign over their preferences. The Taste Approach challenges both assumptions (Otterholt 2010, 59). Setting aside the question of whether these assumptions can be made in the intragenerational context, I shall argue that they cannot be made in the intergenerational one. It is a fact that next generations’ preferences are to a significant extent within our control. We indirectly influence the preferences children will develop by leaving to them a world they will have to learn to like. But we also directly habituate them to a certain way of life (including basic standards of living, the kind of food they will like, and so on). And we inculcate many values in them. We can teach children values such as hard work, respect for the elderly, religious practice, political involvement, politeness or ambition. Most of these values become deeply entrenched and influence many preferences children will develop as well as their willingness to satisfy them or not. In the light of this, the assumption that we should consider our descendants’ preferences as already settled does not seem very reasonable. Now, one might wonder whether the normative assumption, that is, individual sovereignty over one’s preferences, is not unreasonable too. Should we avoid any kind of intentional shaping of children’s preferences? The problem is that, once we become aware of the various ways environmental factors as well as our actions influence children’s preferences, it becomes difficult to ignore them. We have to make decisions about how to educate children. Educational decisions cannot be perfectly neutral (Barry 1995, 143).

12Let us assume then that the next generation’s preferences are within the current generation’s control. “Control” refers here to two different processes. First, the current generation indirectly influences the next generation’s preferences just by what it decides to leave to satisfy the next generation’s preferences (Gosseries 2001, 341; Page 2007, 455). This is the phenomenon of adaptive preferences (Elster 1982): preferences are formed in response to what is available. Hence future generations might learn to like the options we leave to them. After all, quite a few of us regret dodos or Ancient Greek religion. We should then expect that our descendants will adjust their desires to the institutions, the culture and the natural environment we decide to leave to them. However, even if these “bequeath decisions” happen to shape our descendants’ preferences, they do not mean to. But, second, the current generation directly participates in forming the next generation’s preference through education (Gosseries 2001, 340-341; Page 2007, 455). In its broadest definition, “education” refers to any kind of action or process that contributes to shape the educated person’s beliefs, habits, preferences or values. In some way, it even includes the bequeath decisions mentioned above. But, in order to stick to a more common use of the word “education”, let us stick to a slightly narrower definition, which only includes “intentional” or “deliberate” education, that is, activities, programmes and policies whose primary purpose is to transmit knowledge, beliefs, habits and values. This does not mean that education, even in this narrower sense, should be thought to be only schooling. Besides schooling, education includes parental education as well as informal learning and socialization processes. Extra-curricular activities, TV-watching, advertising campaigns, relationships with peers, a neighbour’s reprimands, all of this contributes to education. Therefore, a policy aimed at promoting certain preferences should take all these processes into account.

13In order to know how to design education, we should answer the following question: which preferences should we inculcate in the next generation so that they can enjoy prospects for well-being that are equivalent or superior to ours? Our own prospects for well-being depend of course on various factors. One is resource consumption. But some people are more frugal than others. Some are able to reach a high level of well-being with few resources, while others need a lot to be equally satisfied. On the face of it, the current generation’s preferences are far from being frugal according to standards such as the ecological footprint. According to the World Wildlife Fund, “We would need the regenerative capacity of 1.5 Earths to provide the ecological services we currently use” (Living Planet Report 2014). And if all people on the planet had the living standards of the average resident of Denmark or Belgium, almost 4 planets would be needed. Neither our consumption patterns are sustainable, nor are our procreative choices. If our ecological footprint is increasing, it is also because the world population is increasing very rapidly. It is expected to reach 9 to 10 billion by 2050 (United Nations, Department of Economic and Social Affairs, Population Division 2013). Unless major technological breakthroughs allow us to multiply the efficiency of our use of Earth’s resources by four (which is a lot!), these consumption patterns seem unsustainable [8].

14In the face of the facts cited above, if future generations were to develop preferences that are as resource-intensive as ours (or even more), it is far from guaranteed that they will be able to satisfy them. These concerns motivate the claim that we should inculcate more frugal preferences in our descendants. It is time now to explain what “cheap preferences” mean in this context. A “frugal” or “cheap” preference is a preference that requires a low level of scarce resources consumption to be satisfied. A note on the vocabulary I use is in order. A person who has cheap preferences is not a person whose actual consumption choices are cheap, for cheap consumption choices might just reveal a state of poverty which leaves the person dissatisfied. A person who has cheap preferences can be genuinely satisfied with frugal consumption choices, and this whatever her economic situation is. In addition, I take the words “frugal” and “cheap” to mean the same thing, that is, to refer exclusively to the costs of a way of living in terms of material consumption, independently from any reference to a frugal or ascetic ethical ideal. Let us say that an agent has cheap preferences if she needs less resource to achieve a given level of well-being than another agent.

15Furthermore, we must emphasise the differences between the meaning of “cheap preferences” in the intragenerational context and of “cheap preferences” in the intergenerational one. In the intragenerational context, “cheap preferences” echo to Dworkin and Cohen’s expensive tastes. Dworkin and Cohen define the expensiveness of preferences by comparing a person’s preferences with her contemporaries’ ones (Dworkin 1981, 48; Cohen 2004, 5). However, in the intergenerational context, the picture is somewhat more complicated, for there is actually no intergenerational market. So in order to compare the costs of the preferences of different generations, Lippert-Rasmussen suggests comparing the costs of their preferences were these different generations have access to the same goods:

16

(…) by saying that members of one generation, G2, have more expensive preferences than members of another, G1, I mean that the former would reach a lower level of satisfaction than that actually enjoyed by members of G1 were they only to have the goods available to the latter, whereas members of G1 would have a higher level of welfare than actually enjoyed by members of G2 were they only to have the resources available to G2.

17Lippert-Rasmussen thus suggests taking the other generations’ preferences as a benchmark for comparison. The problem of the absence of the intergenerational market is solved by comparing a generation’s actual state of satisfaction with the other generation’s hypothetical state of satisfaction were it to have access to the same market. But we can propose another benchmark to assess the costs of preferences in the intergenerational context. The benchmark proposed by Lippert-Rasmussen is the cheapness of other generations’ preferences. The benchmark proposed in this paper is the cheapness of the preferences a generation could have formed under counterfactual conditions, that is, had the previous generation decided to educate the next generation in another way. This second benchmark for comparison is the one we are interested in when we want to determine which preferences we should inculcate in children. In order to grasp the difference between these two benchmarks of cheapness, suppose once again two generations, G1 and G2. Each of these generations could have developed two sets of preferences, P1 and P2, had they been subjected to different conditions or had they acted differently. Then the benchmark for cheapness can be defined in two ways, (i) and (ii):

1. G1’s preferences are cheap if, when having access to the same resources, its preferences need less resource than G2’s preferences to be satisfied.
2. G1’s preferences P1 are cheap if (a) P1 requires less resource than P2 to be satisfied and (b) G1 could have developed either P1 or P2.

18In (i), the benchmark for comparison is another, referent, generation’s set of preferences, while in (ii) it is the same generation with a different process of preference formation. If G1 and G2 had the same initial capacity to develop equivalently cheap preferences, then (i) and (ii) would coincide. But if G2’s capacity to “produce” a given level of well-being from a given stock of resource is lower than G1’s capacity, then (i) and (ii) will differ. To grasp what it means, suppose G2 is very tall on the average, while G1 is very short. Size is one of the factors that influences the capacity to “produce” well-being from resources. Taller people need more calories to meet their daily energy needs. Hence G2 is more likely to need more food to satisfy its (informed) preferences. Let us illustrate the comparison now. Suppose G2 has the initial capacity to develop preferences allowing its members to reach a level H of satisfaction with a daily intake of bread going from 5 pieces of bread a day to 10 pieces of bread a day. G1 is shorter and thus has the initial capacity to develop preferences allowing her members to reach H with a daily intake of bread going from one piece of bread a day to 7 pieces of bread a day. According to (i), G1’s preferences are cheap as long as their satisfaction requirements do not exceed 5 pieces of bread a day. According to (ii), G1’s preferences are cheap if they require the smaller possible amount of bread to be satisfied, that is, one bread a day.

19The choice of the benchmark for comparison depends on the question we want to answer. If the question is: “should we compensate a generation for having less cheap preferences than another, and this at the expense of other generations?”, then the relevant benchmark for comparison is these other generations’ preferences. This questions pertains to the distributive obligations a generation has towards the others [9]. If the question is: “which preferences should we inculcate in the next generations in order to secure their prospects for well-being?”, then the relevant benchmark for comparison is the counterfactual set of preferences these generations could develop alternatively [10]. This question pertains to the educational obligations a generation has towards the others. In this paper, I am interested in the latter.

20With a given quantity of resources, having cheap preferences rather than expensive ones increases a person’s prospects for well-being. Now, one might point out that the kind of preferences we have may have an impact on the production of resources (Arneson 2006, 13). The person who has expensive preferences has more incentives to satisfy them. A generation with expensive tastes will probably be more hard-working and more innovative. This objection pertains to the question of whether and to what extent we should assume that resources are scarce within the intergenerational context. The answer to this objection is that it is reasonably risk-averse not to take for granted that resources will grow indefinitely in the future. And, if resources are to remain scarce, who has cheap preferences has more chances to achieve well-being than who has expensive preferences. On these grounds, our generation is morally permitted to inculcate cheap preferences in the next one, since it is one of the ways to secure equal prospects for well-being for future generations (Lippert-Rasmussen 2012). And, if it is true or likely to be true that cheap preferences are a necessary condition to achieve equality of prospects for well-being between generations, then intergenerational justice requires us to inculcate these cheap preferences in children.

21Figure 2 displays a very simple model showing how, in a world of scarce resources, where population is given, cheap preferences engineering would increase the likeliness preferences will be satisfied. Cheap preferences engineering increases the probability that people will choose frugal lifestyle instead of lifestyles that require a high level of consumption (I have labelled them “consumerist lifestyles”). Note that cheap preferences engineering by itself does not remove any lifestyle from the option set (although it makes it more difficult for this lifestyle to be adopted, first, by not developing a taste for it, and, second, to the extent that, the less frequent a way of life is, the more it is difficult to live this way).

Figure 2

Cheap preferences engineering, scarce resources and preference satisfaction

Cheap preferences engineering, scarce resources and preference satisfaction

4 – Three objections to cheap preferences engineering: the Objective Good Objection, the Autonomy Objection and the Fairness Objection

22I will now consider the main potential and existing objections to cheap preferences engineering. I will especially focus on two of them, namely, the Autonomy Objection and the Fairness Objection.

23But before, let us have a look at the Objective Good Objection. It starts with the worry that cheap preferences could be preferences for worthless states of affairs (Arneson 2006; Bykvist 2009; Barry 1997). Some argue that the permissibility of “cheap” or “easy to satisfy” preferences engineering is a good reason to abandon preferentialism or subjective welfarism (Barry 1997; Page 2007). Others suggest that preferentialist theories should deal with the problem of cheap preferences by putting constraints on the kinds of preferences that can be developed. These preferences ought to be compatible with some objective theory of well-being (Arneson 2006; Bykvist 2009; Otterholt 2010). But this discussion tends to take that “cheap” or “easy to satisfy” preferences are always “worthless”. But “cheapness” only refers to the costs of satisfaction of the preference in terms of material resources [11]. There is no necessary strict correlation between the costs of satisfying a preference and its ethical or hedonic value. A preference for watching sunsets is cheap, and yet its satisfaction is deemed valuable by many accounts of well-being [12]. Furthermore, the account of well-being developed in this paper avoids some pitfalls, since it takes well-being to be the satisfaction of rational preferences. It is less than likely that a rational preference as understood here (that is, a preference resulting from a thoroughgoing deliberation) could be satisfied by worthless states of affairs. For example, even if it is cheaper to eat food that is insufficient in quantity and in quality, a rational person will not develop such a preference, because she is able to foresee the consequences of this diet on her health and life expectancy (unless, of course, she prioritizes other values, like hunger strikers do). In other words, this paper argues for the cultivation, not of the cheaper preferences tout court, but of the cheaper preferences among rational preferences [13].

24Let us now turn to the Fairness Objection and the Autonomy Objection. I shall deal with them simultaneously since my solution aims to accommodate both. Let’s start with the Fairness Objection. Nowadays our preferences are expensive, probably more expensive than the cheap preferences prudence recommends us to inculcate in our children. Since well-being depends on our rational preferences being satisfied, our prospects for well-being could be enhanced without harming future generations by allowing ourselves to go ahead and deplete natural resources (Lippert-Rasmussen 2012). Cheap preferences engineering would legitimize Thénardier-like parents bringing up Cosette- and Gavroche-like children the hard way while leading an easy-going life themselves. This sounds unfair. One could think we need to give up the “prospects for well-being metric” and replace it by a “resource metric” if we want to accommodate this intuition. My aim is to argue that this is not necessary. Subjective welfarism can meet the objection. How?

25First of all, an education aiming to inculcate cheap preferences is less successful if educators do not preach by example. Examples are a key element of value-shaping education and parents are one of the most common role models (Bucher 1998). If G1 wants to succeed in inculcating cheap preferences in G2, to the extent that it is true that a value- and habit-shaping education is more efficient when educators preach by example, then G1 must also change its preferences to adopt cheaper ones. As Van Parijs puts it:

26

(…) not the slightest chance of success if one does not preach by example. I therefore regard at least a modicum of conspicuous frugality as an intrinsic component of my parental responsibilities. Hence my resolute policy of wearing a watch which my children would not dream of wearing, and clothes which my sons have grown out of (…).

27If G1 has to shift from expensive to cheap preferences, then welfarism does not allow it any more to deplete natural resources, since G1 does not need all these resources to be satisfied any more. Now, we do not know exactly to what extent children and adolescents let role models influence their own behaviours (Kristjánsson 2006). And not all members of G1 are involved in the education of G2. Childless people who do not happen to be teacher or educator could still indulge themselves in satisfying their costly preferences.

28There is another, more compelling reason why G1 should not deplete most of the resources even if it inculcates cheap preferences in G2. This reason pertains to the kind of educational methods G1 is allowed to use and its effects on the kinds of preferences G2 will develop. It is now time to tackle the Autonomy Objection. Some might object to cheap preferences engineering, not because they believe cheap preferences are worthless, but because they distrust educational projects aiming to inoculate specific preferences or behaviours in children. They fear such projects would undermine the development of autonomy. But there are in fact many methods educators could use to cultivate or inculcate cheap preferences. Suppose two ideal-types of educational method. “Liberal Education” will refer to the most autonomy-compatible ways to inculcate cheap preferences. “Conditioning” will refer to the most autonomy-circumventing ways to inculcate cheap preferences. A Liberal Education is an education that, among other objectives, aims to render the prospective adult autonomous. Liberal educators facilitate the development of autonomy by teaching children all the capacities involved in critical reflection (Brighouse 1998).

29Can Liberal Education be compatible with cheap preferences engineering? First of all, let us note that it is perfectly possible for an adult to deliberately cultivate cheap preferences without losing her autonomy (Elster 1982, 235; Arneson 2006, 19; Bykvist 2009, 317). Stoic philosophers, who subscribed to quite high standards of autonomy, were also notoriously frugal. Now, if I can autonomously change my preferences, can I also change my children’s preferences without interfering with their autonomy? The difference between children education and adult character planning is that children acquire preferences and habits before they are able to exercise the capacities associated with autonomy. Hence they cannot deliberately choose to acquire cheap preferences. However, they can endorse them retrospectively. The compatibility between autonomy and cheap preferences cultivation is possible if autonomy is aimed to be an end-state, as opposed to autonomy as a precondition, a distinction defended by Clayton (Clayton 2006, 89-92). Autonomy is an end-state when children are enrolled in specific preferences and values but are also taught the autonomy-related skills, so that they can in principle challenge these preferences and values in the future. When the autonomous adult reflects on her previously acquired cheap preferences, she has the genuine possibility to either reject them or reflectively endorse them. Autonomy is a precondition when educators avoid imparting children any kind of values or preferences. Clayton argues that, insofar as comprehensive conceptions of the good (especially religions) are concerned, we should take autonomy to be a precondition and avoid comprehensive enrolment. Would this line of reasoning hold for consumption habits too? Those who think that deliberate preference-shaping undermines autonomy as a precondition should be equally worried about the impact of commercial advertising campaigns and of the consumerist culture on children’s capacity to form autonomous preferences later on. Empirical research reveals that children under 8 are not able to understand the commercial intent of advertising; they view it as funny and informative (e.g. Blatt, Spencer, and Ward 1972). If education should aim to facilitate autonomy as a precondition even with respect to consumption habits, then the least a liberal society can do is to ban commercial advertising. But even though, parents and schools cannot help shaping children’s consumption habits, by the very fact that they habituate them to specific food, clothing, housing, toys, hobbies, and so on. We should not frame the problem as if education and socialization processes had so far be neutral with respect to consumption habits, as if we had the choice between “no preference engineering” and “cheap preferences engineering”. This would be very naïve. In our world, the default option in terms of preference formation is consumerism [14]. The choice is thus rather between “cheap preferences engineering” and “expensive (or consumerist) preferences engineering”. Therefore, with respect to consumption habits, autonomy as an end-state is the only attainable goal.

30Conditioning, broadly understood, refers here to a learning process through which the agent develops the preferences and behaviours the educator wants her to have. We can safely predict that an agent that is conditioned to have cheap preferences will indeed have desires for cheap things and act upon these desires. Nowadays conditioning seems to be rather used for the purpose of inoculating expensive preferences: researchers warn us that commercial advertising campaigns activate the brain reward-system and thus give us little freedom of choice (Dumas et al. 2012). If neuromarketing works, and if it could work for the specific purpose of inoculating cheap preferences, it could be used for the sake of intergenerational justice. Let us note that the very justification for choosing Conditioning is similar to the justification for human engineering as a response to environmental issues. Because biomedical human engineering would be an effective way to change human behaviours, some view it as a promising solution to environmental issues and climate change and claim our aversion for it is mainly due to a status quo bias (Persson and Savulescu 2012; Liao, Sandberg, and Roache 2012). Perhaps our current aversion for conditioning, be it through education or biomedical means, is just due to a status quo bias. And, in a world of scarce resources, conditioning children to develop cheap preferences is like delivering them a happiness pill, like Huxley’s Soma pill. In his novel Brave New World, Huxley described the effects of the Soma pill, a happiness pill, in the following way:

31

The world’s stable now. People are happy; they get what they want, and they never want what they can’t get.(…) they’re so conditioned that they practically can’t help behaving as they ought to behave. And if anything should go wrong, there’s soma.

32And indeed Conditioning would allow the current generation (G1) which was educated to have expensive preferences to deplete as much resources as it needs to reach maximal satisfaction, at least to the extent that it does not undermine the efficacy of cheap preferences engineering. If the Conditioning method is chosen, then it is very likely that G2 will not change its preferences later on and keep them cheap. G2 will not suffer from frustration in a world of scarce resources. If we were to be concerned only with G2’s prospects for well-being, we should choose Conditioning. Now, advocates of cheap preferences engineering still tend to think that educational methods should not undermine the development of autonomy (Otterholt, 64; Lippert-Rasmussen, 512). Hence they would favour Liberal Education. How could we justify this choice from a welfarist perspective [15]?

33Liberal education might be less efficient than Conditioning in inculcating cheap preferences. However, welfarism may still require educators to foster autonomy-related skills and attitudes if being autonomous could make a decisive contribution to one’s well-being. Note that the suggestion that being autonomous can contribute to one’s well-being does not imply that one should lead an autonomous life in order to live well. That would be a too strong claim. It just implies that possessing autonomy-related skills and attitudes increases people’s prospects for subjective well-being. But why is that so? A first line of answer could be that an autonomy requirement is built into the requirement that preferences should be informed, or rational. But this line of answer does not provide a specific case for the cultivation of autonomy-related skills. It is true that, in order to develop rational preferences, an agent needs some of the skills that also tend to favour the development of autonomy. The skills needed to form a rational preference include the ability to check the truth or at least the probability of a claim and the skills involved in logical reasoning. A person who has rational preferences is able to select the option that is the most likely to help her to achieve her goal, whatever this goal is. Rationality obviously facilitates autonomy. However, on most accounts, autonomy requires more. An autonomous agent is also a “person”, in the sense Frankfurt gives to this word (Frankfurt 1971). What makes an agent a person is not only that she employs rational faculties to achieve her desires. It is that she is capable of forming second-order volitions, as opposed to first-order volitions. An autonomous agent is thus aware of her own will and is not neutral with respect to her possible first-order desires. Then, imposing a requirement of rationality on preferences is not sufficient to justify the choice of autonomy-compatible methods for cheap preferences engineering.

34Nevertheless, two specific arguments can be advanced to support the claim that being autonomous increases the prospects for subjective well-being. Let us label these arguments the context argument and the endorsement argument.

35The context argument points out that autonomous persons are likely to live better in the modern world. Being able to develop and revise autonomous preferences improves one’s chances to live well in our circumstances, because modern societies are technologically, politically, morally complex and are continuously evolving. Being autonomous enables a prospective adult to develop and to revise her preferences in a way that is more suitable to the world that awaits her than what she would have done had she been conditioned. This is Joseph Raz’s argument for autonomy:

36

[The ideal of autonomy] is an ideal particularly suited to the conditions of the industrial age and its aftermath with their fast changing technologies and free movement of labour. They call for an ability to cope with changing technological, economic and social conditions, for an ability to adjust, to acquire new skills, to move from one subculture to another, to come to terms with new scientific and moral views. Its suitability to our conditions and the deep roots it has in our culture contribute to a powerful case for this ideal.

37Now, this argument is contingent upon the characteristics of the conditions we live in. Such conditions might not maintain in the future, although we can have good reasons to believe that they will. In the near future, we can expect technologies to change even faster, workers to become even more mobile, bureaucratic complexity to increase, ethical pluralism to become more and more a characteristic feature of the moral landscape, possibilities to choose one’s form of family, when and how to end one’s life expectancy or the genetic makeup of one’s children to become real. These evolutions involve difficult personal choices and being equipped with autonomy-related skills will be helpful, if not essential to make these choices.

38The endorsement argument states that being autonomous enables people to form and revise their preferences in a way that is more suitable, not to the external conditions they live in, but to their own personality [16]. By “personality”, I mean the assemblage of qualities and traits that are specific to a person and that she could identify as being hers. A person that is equipped with autonomy-related skills is more able to change her preferences whenever there is a lack of fit between her own personality and the current preferences she acquired through education and socialization. Being autonomous is valuable because this lack of fit can dramatically decrease the level of subjective well-being one can enjoy. And this lack of fit jeopardizes one’s prospects for subjective well-being because, if a person has acquired, through education and socialization, preferences and values that conflict with her personality, and if she is unable to revise these preferences and values, then it will be impossible for her to live her life “from the inside” (Brighouse and Swift 2014, 167), that is, to endorse it. To what extent does the lack of endorsement affect a person’s prospects for well-being? According to Dworkin, endorsement is of primary importance if we adopt a constitutive view of how to evaluate one’s life instead of an additive view (Dworkin 2000, 218). The additive view states that the value of a life is equal to the sum of its components (the more or less intense preferences that have been satisfied). If a component was endorsed, this adds to the value of this component. In the additive view, endorsement is, writes Dworkin, like the “frosting on the cake”. The constitutive view states that a component adds to the value of a life only if it is endorsed. Hence, while the additive view merely grants that endorsement adds value to an already valuable life, the constitutive view goes as far as saying that a life without endorsement is worthless. The constitutive view is preferable, according to Dworkin, because it better makes sense of why a person’s life, as a whole, is valuable for the person whose life it is.

39Now, there is a difficulty with the endorsement argument. It cannot, by itself, address to the following issue: if endorsement is so crucial for well-being, couldn’t it be inculcated simultaneously with cheap preferences, through more elaborate conditioning methods perhaps (assuming such methods exist and work)? For Dworkin, endorsement ought to be genuine, that is, the mechanisms that induce a person to endorse a particular set of preferences should not lessen her ability to evaluate this set of preferences in a reflective and critical way. Another advocate of autonomous preference formation, Sumner, also appeals to the agent’s counterfactual judgment to check whether the agent’s preferences are authentic, that is, whether they are truly the agent’s own preferences (Sumner 1999, 156-170). But if a pseudo-endorsement could be instilled in the same way as cheap preferences, why would genuine endorsement matter from a welfarist perspective? Sumner suggests that an autonomy-as-authenticity requirement could help the agent to correct regrettable preferences, not in the sense that they are misinformed, but in the sense that they reflect problematic socialization processes. These are the processes that involve coercion, manipulation, oppression, and so on. Women’s adaptive preferences in sexist societies are a typical instantiation of such processes. But what if coercive and manipulative processes were not aimed to subject women or other groups, but to serve the best interests of those who will live in a world of scarce resources? In other words, educators primarily concerned with the well-being of the future adults they educate would face a trade-off between prudential considerations that require engineering cheap preferences in the most efficient way, and securing genuine endorsement.

40There are two possible ways out of this difficulty. Firstly, we can appeal to the fact that it is hard, even impossible, for many people to change their constitution [17]. And our constitutions restrict the range of the ways of lives that could make us happy. Suppose, for instance, that the cultivation of cheap preferences involves developing a preference for a vegan diet. But the metabolism of some of those who have learned to like this vegan diet suddenly becomes, for some reason, incapable of using efficiently plant proteins. Hence those people are not well constituted to live well by satisfying their cheap preferences. At a first glance, it would be rational for them to change their preferences. But if they had been brainwashed to endorse values that sustain their vegan diet (suppose they would feel ashamed or impure if they ate meat), they would resist any preference change and stick to a way of life that does not do them a lot of good. But if they had the tools needed to revise their vegan preferences, they could have access to a higher level of well-being, given their particular condition. Given that some features of our constitution cannot realistically be changed, prudence recommends giving people some room to decide whether they want to endorse the preference they have learned to have or not. This is for the first way of addressing the difficulty.

41Secondly, we can appeal to the fact that education and socialization involve a variety of influences and develop many other preferences and values besides cheap preferences. Therefore, even if cheap preferences were in everyone’s best interest, which means that whether these preferences are pseudo- or genuinely endorsed would not matter for well-being, other preferences and values that are inculcated in children might not be in their best interests. Parents, schools, churches and corporations will try to impose on children preferences and values for other aspects of their lives than consumption (sexual or religious preferences, for instance). Some of these preferences and values will help children to live well. Others will not. Hence children will need to acquire autonomy-related skills to challenge these preferences and values if necessary. But if, as a result of their education, they will be able to revise their sexual or religious preferences, they should also be able to revise their cheap preferences. Therefore, more elaborate methods of conditioning cheap preferences are not advisable, because these methods would preclude the option for children to challenge other aspects of their education that are crucial for their well-being. To conclude, given the diversity of people’s constitution and given the various influences children are exposed to, prudence recommends teaching them autonomy-related skills so that they can end up being able to revise all their preferences and values if needed. As a result, even from a welfarist perspective, conditioning methods for inculcating cheap preferences are not desirable.

5 – Distributive implications

42If the arguments advanced above are sound, then, in order to secure G2’s prospects for well-being, G1 should inculcate cheap preferences in G2 but in an autonomy-compatible way. But liberal methods of education will result in G2 being able to revise its previously cheap preferences. In the future, G2’s preference changes may result in G2 having more expensive preferences. This implies that, in order to enable G2 to satisfy its revised preferences, G1 should leave more resources to G2 than the amount of resources needed to satisfy its previous cheap preferences. We now have the means to address the Fairness Objection. There is indeed something unfair in the situation where G1 depletes all resources and inculcate cheap preferences in G2. But this is not because equality of resources between G1 and G2 matters as such. This is because, even if it is more prudent to cultivate cheap preferences among G2, a richer account of G2’s prospects for well-being requires providing G2 with the capacity to revise its cheap preferences. But this revising process would be meaningless if G2 did not also have access to more resources than what it needs to satisfy cheap preferences.

43Then, one could first ask how much resources, and which kind of resources, would be needed by G2 to be able to meaningfully revise its preferences. In other words, what level of intergenerational savings (in terms of resources), and what types of savings, can guarantee that G2 will enjoy equal or higher prospects for critical well-being than G1? One could think that such a requirement would lead to excessively high savings. After all, if I wish my children to be able to satisfy whatever preferences they might decide to develop (instead of their previous cheap preferences), I should leave as much resources as possible to them. And these resources should be as diverse as possible (they should include environmental goods with the greatest possible biodiversity as well as a lot of machinery and material capital, investments in education as well as in health care, preserved artworks as well as investments in artistic creation, natural resources as well as new technologies…). If that is true, then a welfarist theory of intergenerational justice like the one discussed here would in fact not allow G1 to deplete as much as it can. It would require G1 to live as austerely as possible (though enough to enjoy the prospects for well-being that are equivalent to the ones the preceding generation enjoyed). But two further considerations need to be taken into account here. Firstly, let us not forget that G2 has been enrolled in cheap before being able to change its preferences. Even if G2 has the opportunity to revise its cheap preferences, it is likely that many members of G2 will continue to live frugally. In other words, cultivating cheap preferences plus the capacity to revise them would allow G1 to dis-save, but to dis-save less than cultivating cheap preferences without the capacity to revise them. Second, one does not need to have access to a maximally big and maximally diverse basket of resources to be able to revise one’s preferences in a meaningful way. In fact, we can and do actually revise autonomously our preferences without a millionaire’s fortune and power.

44The quantity and the type of resources G1 ought to leave to G2 should thus be constrained by an additional normative requirement, which corresponds to the external conditions for autonomy. For a person to be autonomous, that is, to be able to undertake a meaningful revision of her preferences (with a reasonable expectation to have them satisfied), there must be adequate options available to her (Raz 1986, 373). As Raz puts it, for options to be adequate, “clearly not number but variety matters” (Raz 1986, 375). Raz provides the example of a choice between houses to illustrate his point. A choice between a great number of very similar houses is less meaningful (as a “choice”) than, say, a choice between only two genuinely different options, an apartment in the city and a cottage in the countryside. Formally, according to Raz, options are sufficiently varied if they meet a “variety test”. A sufficiently varied set of options “enables the person to sustain throughout [her] life activities which, taken together, exercise all the capacities human beings have an innate drive to exercise, as well as to decline to develop any of them.” (Raz 1986, 375). The activity of choosing between these options must be sufficiently challenging for the person to have to exercise certain abilities, including stimulating our sense, engaging our imagination and affection, occupying our mind. In order to secure G2’s prospects for critical well-being, G1 does not have to leave a great number of options to G2. G1 just has to leave enough and sufficiently varied options to secure G2’s access to meaningful preference revision. Now, the costs of securing an adequate set of options for future generations to be able to meaningfully revise their preferences might still be high. More research needs to be done to translate the normative criterion of adequacy into concrete intergenerational policies.

45To summarize, I argue that prudence requires G1 to inculcate cheap preferences in G2, but in an autonomous-compatible way. The distributive implications of autonomy-compatible methods of cheap preferences engineering are different from the distributive implications of autonomy-circumventing methods of cheap preferences engineering. The latter allows high intergenerational dis-savings because resources ought to be allocated according to the costs of each generation’s preferences. But the former method does not allow such dis-savings. Even if members of G2 are more likely to keep their previous cheap preferences once grown up, they should be provided with the internal as well as the external conditions to be able to meaningfully revise their preferences. The external condition that has distributive implications is that the set of options younger generations should have access to must be adequate, in the Razian sense. To put it simply, G1 must not dis-save to the point that G2 cannot even consider the possibility to shift from cheap preferences to less cheap ones [18].

6 – Conclusion

46Cheap preferences engineering is a promising way to secure prospects for well-being for future generations that are equal or higher than ours. But it raises many worries. This paper has proposed an answer to two major objections to cheap preferences engineering, that is, the Autonomy Objection and the Fairness objection. The main argument can be summarized as follows. Educational methods to cultivate cheap preferences are compatible with the development of autonomy understood as an end-state. It should even be advisable to prefer autonomy-facilitating methods to conditioning because being autonomous increases one’s prospects for well-being for at least two reasons. First, being autonomous enables one to have a better life in the modern world, where technologies, mobility or ethical pluralism require individuals to make more difficult choices by themselves than before (this is the Razian case for autonomy). Second, being autonomous enables one to revise the preferences educators have inculcated in her if these preferences did not fit well with her personality. Hence she will be able to endorse her preferences, endorsement being a condition for preference satisfaction to be of high subjective value. This answers the Autonomy Objection. Now, if autonomy-compatible methods of education are chosen, it is possible – though less likely to happen – for G2’s member to change their preferences so that they end up having more expensive preferences. For G2 to be able to meaningfully revise its preferences, the amount of resources G1 leaves to G2 ought to be higher than the amount of resources needed to satisfy G2’s initial cheap preferences. Therefore, G1 is not allowed to deplete as much resources as it can and the Fairness Objection can be accommodated.