The view from Manywhere: normative economics with context-dependent preferences

Abstract

We propose a methodology for normative evaluation when preferences are context-dependent. We offer a precise definition of context-dependence and formulate a normative criterion of self-determination, according to which one situation is better than another if individuals are aware of more potential contexts of a choice problem. We provide two interpretations of our normative approach: an extension of Sugden’s opportunity criterion and an application of Sen’s positional views in his theory of justice. Our proposition is consistent with Muldoon’s and Gaus’ approaches of public reason in social contract theory, which account for the diversity of perspectives in non-ideal worlds.

1. Introduction

A central and recurrent empirical finding in behavioural economics is that individuals’ preferences are context-dependent. Preferences may depend on some aspects of the choice situation, such as the order in which the options are available or the way the choice problem is formulated.¹ A major challenge for normative economics is to determine whether such context-dependent preferences give evidence about individual welfare, or if they can be ignored and treated as normatively irrelevant. Throughout this article, we use the term ‘welfare’ in the economists’ sense of preference satisfaction and avoid references to the philosophically disputed notion of ‘well-being’.² We use the expression ‘standard normative economics’ to refer to the branch of normative economics that assumes that individual preferences are context-independent and satisfy some norms of rational choice (typically transitivity, reflexivity and completeness). By contrast, we use ‘behavioural normative economics’ to refer to the branch of normative economics that accounts for non-standard preferences (context-dependent preferences, intransitive preferences, etc.).³ The aim of this article is to shift the focus of behavioural normative economics from the individual’s welfare – either defined as the individual’s actual judgement or as his counterfactual enlightened judgement on what constitutes his welfare – to the individual’s ability to confront different judgements he may have on what is constitutive of his welfare.

As an illustration, imagine the ‘dessert problem’. You and your colleagues regularly go to the same restaurant for lunch. You know that the chef prepares very tasty desserts, although they are slightly oversized for you. You therefore face a recurrent choice problem. Should you order a dessert, knowing that it will give you a significant and immediate enjoyment but probably make you a bit sleepy in the afternoon? Your actual choice is very likely to depend on the context: whether some of your colleagues have already ordered a dessert, whether you are in a good or bad mood, whether the chef nicely displayed a dessert close to your table to arouse your appetite, etc. Since your preference for a dessert is context-dependent, it is not clear whether your actual choice constitutes a reliable evidence of what maximizes your welfare – and therefore, whether it is the ‘good’ choice in such a situation. Suppose, in line with the common practice in standard normative economics, that your welfare as measured by the theorist fundamentally depends on your own judgement about what constitutes your welfare.⁴ We can distinguish between three different approaches to define what your own judgement is.

1. (i) It is your current judgement in a given choice situation, e.g. it is a rainy day, you have just got a paper rejected, and the dessert looks very nice on the counter, so you order it.

2. (ii) It is the outcome of a kind of bargaining between your current self and what you imagine your future self will be this afternoon, weighing the pros and cons of both options, and whose result will very likely be context-dependent too.

3. (iii) It is your counterfactual ‘enlightened’ judgement about the choice problem, i.e. the rational tradeoff between an immediate enjoyment and a later cost, the result of which being determined by your deeply held values and preferences for different aspects of your life.

The argument we develop in this article is that option (ii) offers the most satisfying approach to determine individual welfare. Option (i) gives primary importance to the individual’s current personal judgement, and thus takes the first-person standpoint. Option (iii) considers an objective or context-independent perspective on what individuals ought to do, and thus takes the third-person standpoint. The approach we promote takes, instead, the second-person standpoint: the characterization of what individuals ought to do is the result of a fruitful confrontation of their different selves, who are all affected by the choice problem. Our aim is to develop a methodological approach based on the second-person standpoint to make normative evaluations, and to provide a brief sketch of its formal characterization.

The rest of the article is organized as follows. In section 2 we generalize the intuition of the three alternative standpoints in the dessert problem (the first-, second- and third-person standpoints) and propose a typology of normative evaluations as (respectively) views from ‘somewhere’, ‘manywhere’, and ‘nowhere’. Section 3 introduces the second-person standpoint and the view from manywhere. We advance a normative criterion of ‘self-determination’, according to which one state of affairs is better than another if individuals are aware of more factors that influence their behaviour. In section 4 we offer two justifications for our normative approach inspired by Sugden’s (2004, 2018) opportunity criterion and Sen’s (2009) notion of ‘positional views’. We then argue that our proposition is consistent with contemporary theories of justice that specifically account for the diversity of judgements in non-ideal worlds (Gaus 2016; Muldoon 2016). Section 5 concludes by discussing some practical implications of our normative approach for behavioural public policy.

2. To Whom Should Normative Economics be Addressed?

An important – although often overlooked – question that must be addressed when defining what makes an individual better off is ‘according to whom?’. While behavioural normative economists usually consider that context-dependent preferences are normatively problematic, it is also because what counts as the context is the result of their own value judgements about what is relevant in the given choice problem. For example, we could imagine that, in the dessert problem, an individual considers that the location of the dessert is a relevant property of the choice problem. It is unclear why the theorist should be entitled to impose her own normative judgement about what counts as a relevant property.

2.1. Which standpoint?

Most of the literature in behavioural normative economics considers that it is the role of the theorist, as an external observer, to characterize the individual welfare function. When the theorist does not imagine herself directly in such a position, she usually addresses her recommendation to an abstract ‘social planner’ according to the economist’s usual jargon (Kahneman et al. 1997), to a non-less abstract ‘choice architect’ (Thaler and Sunstein 2009) or to an actual consultant advising her clients (Harrison and Ross 2018). In this approach individual welfare is defined from a third-person standpoint: it is the impersonal perspective of an external observer who benevolently thinks about what is good for individuals. Sugden (2004, 2013, 2018) criticizes this approach, arguing that the theorist should refrain from making value judgements about individuals’ preferences, such as giving more weight to the individual’s counterfactual enlightened judgement. In his account, the adequate standpoint for making normative judgements is the individual himself: he is the only judge of what constitutes his welfare, not an external third party. According to the first-person standpoint, nothing requires that the individual ought to make enlightened choices. In that regard, context-dependent preferences are not a fundamental issue for normative economics. What matters here is not the content of one’s preferences, as judged by an external party, but that social institutions are designed in a way that each individual has the ability to satisfy any preference he might have.

Complementary to these two approaches, we argue that (i) theorists should avoid formulating value judgements about what is best for individuals, while (ii) recognizing that context-dependent preferences may still be a normative issue – in particular if individuals are not aware that their behaviour is likely to be influenced by the context. We are sympathetic to Sugden’s position that normative economics should be addressed to individuals rather than to the abstract social planner: what matters is not how individuals actually behave, but that they are able to satisfy their own preferences – or more abstractly, that they have the means to choose the kind of life they might desire. However, unlike Sugden, we recognize the possibility that individuals may consider some of their inconsistent choices as problematic, had they been aware of such an inconsistency.⁵

We advance that individuals should have the opportunity to avoid manipulation and the undue influence of factors that they are not aware of. This ability of critical self-reflection does not require individuals to form an enlightened judgement like the one of the idealistic inner rational agent (Infante et al. 2016). We refer to a much weaker condition that remains compatible with the actual cognitive capacities of psychological beings.⁶ In order to avoid paternalistic claims about which contexts and preferences are ‘problematic’, we argue that the role of the theorist should not be to give to individuals what they would choose if they were enlightened, but to ensure that social institutions are designed in a way that individuals have the opportunity to form their own enlightened judgements. The adequate standpoint to form a normative judgement is the second-person standpoint, which we define in detail in section 3.

As a matter of terminology, we use the expression ‘view from nowhere’ – as in Sugden (2013), who borrows the terms coined by Nagel (1986) – to designate the methodology of behavioural normative economics that consists in defining welfare from the third-person standpoint, i.e. the position of the theorist.⁷ We use the expression ‘view from somewhere’ to designate Sugden’s proposition that normative judgements ought to be made from the first-person standpoint, i.e. from the position of the individual himself embedded in a specific context (rather than his counterfactual enlightened judgement, as imagined by the theorist).⁸ In contrast with both the third-person ‘view from nowhere’ – for which the theorist defines what the enlightened counterfactual choice of individuals would be – and the first-person ‘view from somewhere’ – where normative judgements are expressed directly by individuals themselves – we label our approach as the ‘view from manywhere’.⁹

2.2. Defining the context

Since context-dependent preferences can be seen as the core of behavioural normative economics, we first propose a precise definition of the ‘context’ in order to clarify the terms of the discussion. It is indeed noteworthy that the literature lacks a precise definition of the notion, mostly because it may lead to circular definitions – as e.g. the ‘irrelevant’ factors, without defining a clear criterion to identify what the ‘relevant’ factors are. We propose a definition in line with the casual use of the expression in behavioural economics as the ‘background’ of choice, that seems to be irrelevant from a normative perspective, while, however, still influencing to some extent the choice of boundedly rational individuals.

We consider an individual $I$ who must choose an option $x$ among the non-empty set of available alternatives $X$ . Each option is described by a list of properties $P$ , with ${\cal P}$ the set of properties. Formally, each property $P \in {\cal P}$ is a function assigning to each option $x \in X$ a value $P\left( x \right)$ from some range. In the case of a binary property, the range is $\left\{ {0;1} \right\}$ , where $P\left( x \right) = 1$ means that $x$ has the property and $P\left( x \right) = 0$ means that $x$ does not have the property. More generally, the range could be some interval of values, where $P\left( x \right)$ represents the degree to which $x$ has the property (e.g. the distance between the option $x$ and a reference point). Properties can either refer to intrinsic properties of the alternatives (e.g. the location of the dessert, its colour, its shape, its distance) or properties of the situation (e.g. norms about how to behave in a restaurant). We consider different types of properties in our analysis: (i) motivational properties $P \in {{\cal M}_I} \subseteq {\cal P}$ , (ii) known properties $P \in {{\cal K}_I} \subseteq {\cal P}$ and (iii) relevant properties $P \in {{\cal R}_I} \subseteq {\cal P}$ . The distinction is the following. The motivational properties are the properties which influence the actual choice of the individual, the known properties are the properties of which the individual is aware – i.e. when considering the options, the individual can determine the value $P\left( x \right)$ – and the relevant properties are the properties which are normatively relevant for the individual – i.e. the properties that determine whether an option is ‘better’ than another for the individual.

The set of motivational, known and relevant properties may overlap, and there is a priori no relation of inclusiveness between ${{\cal M}_I}$ , ${{\cal K}_I}$ and ${{\cal R}_I}$ . Imagine, for instance, the dessert problem discussed above. Several desserts are available and displayed on a buffet, all of them within sight of $I$ , including a cake. The cake is presented in a blue plate, is of medium size, turns out to be sugar-free, and was voluntarily positioned slightly in front of the other desserts – because the restaurant owner realized that clients rather tend to choose the desserts which were in front. We have several properties characterizing the cake, which could be formalized as follows:

• ${P_b}\left( {Cake} \right) = 1$ , meaning that the property ‘blue plate’ is satisfied.

• ${P_{sz}}\left( {Cake} \right) = 0.5$ , meaning the relative size of the cake, on a range of real numbers from 0 to 1, is in the middle.

• ${P_{sg}}\left( {Cake} \right) = 0$ , meaning that the property ‘sugar’ is not satisfied.

• ${P_\ell }\left( {Cake} \right) = 1$ , representing the relative location of the cake on the buffet, on a range of real numbers from 0 (in the back) to 1 (ahead of others).

Suppose that ${{\cal K}_I} = \left\{ {{P_b},{P_{sz}}} \right\}$ , ${{\cal R}_I} = \left\{ {{P_{sz}},{P_{sg}}} \right\}$ and ${{\cal M}_I} = \left\{ {{P_{sz}},{P_\ell }} \right\}$ . The client is aware that the cake is displayed in a blue plate and is of medium size, although he did not realize that it was ahead of the other desserts and sugar-free. He considers in the current situation that only the size of the cake matters, but if he was aware that the cake was sugar-free, he would also consider this property as relevant. Lastly, the only factors that will influence his actual choice are the size of the cake (which is a known and relevant property) as well as the relative position on the buffet (which is an unknown and irrelevant property). We have here a situation in which a property is relevant, motivational and known (the size of the cake), another which is also relevant, but neither motivational nor known (the sugar content), a property which is motivational, but neither known nor relevant (the relative position on the buffet) and another which is known, but neither relevant nor motivational (the colour of the plate).¹⁰

The sets ${{\cal M}_I}$ , ${{\cal K}_I}$ and ${{\cal R}_I}$ are the representation by the theorist of the choice problem faced by $I$ . For simplicity, we assume that the theorist correctly identifies the set ${{\cal M}_I}$ , i.e. she precisely knows what the properties which influence the choice of the individual are. Our definition of the context is based on the premise that it refers to what we, theorists, consider as the ‘irrelevant’ properties of the choice situation (Bacharach 2006: 13). Formally, a context property is a property that is motivational but not relevant: $P \in {{\cal C}_I} = {{\cal M}_I}\backslash {{\cal R}_I}$ . A context is any combination $C = {({C_P})_{P \in {{\cal C}_I}}}$ of values of the context properties. In the illustration above, there is only one property – the relative location of the cake on the buffet – that is motivational and not relevant, i.e. ${{\cal C}_I} = \left\{ {{P_\ell }} \right\}$ , and the context is defined as the set of values of relative location of the different desserts on the buffet. With these notations, we have context-dependent preferences when the set of contexts ${{\cal C}_I}$ is non-empty, i.e. there exist properties which are motivational although not relevant (even though the individual might be aware or not of their existence).

Given that the choice of the individual (and therefore, his preferences) may depend on the context, we define the individual $I$ as a collection of selves $\left\{ {{I_{C{\rm{'}}}},{I_{C{\rm{''}}}},{I_{C{\rm{'''}}}}, \ldots } \right\}$ . $I$ is an enduring living entity, whose selves ${I_{C{\rm{'}}}},{I_{C{\rm{''}}}},{I_{C{\rm{'''}}}}$ , etc. are connected over time by some relations (which we do not need to specify here).¹¹ When $I$ faces a choice problem in a context $C$ , he exists through a particular self ${I_C}$ embedded in this context $C$ . Furthermore, we define the counterfactual enlightened self ${I_0}$ describing how $I$ would choose in a ‘context-free’ situation – guaranteeing that his choices only depend on properties that are considered as relevant from the theorist’s perspective. Note that questions regarding the relationship between multiple selves for normative concerns are not new. Schelling (1980) already pointed out in early works the issue of ‘self-management’ (in his own terms), defined as the way of dealing with self-control problems over time:

People behave sometimes as if they have two selves: one wants clean lungs and long life and another who adores tobacco, or one who wants a lean body and another who wants dessert, or one who yearns to improve himself by reading The Public Interest and another who would rather watch an old movie on television. The two are in continual contest for control … which is the authentic “I”? (Schelling 1980: 95–104)

Our approach differs from the one of Schelling in at least two respects. First, and as previously argued, we do not endorse the view according to which welfare is determined by one’s deeply held values and preferences for some aspects of one’s life. This is what Schelling suggests when he introduces ‘the notion of a consumer’s having reasonably stable values, knowing his values, and having the intellectual capacity to make choices that satisfy those values’ (1980: 96–97). Second, we do not suggest that there is an ${I_C}$ to be ‘more authentic’ than the others, as Schelling sees it by considering such a problem as a ‘problem of authenticity’ (1980: 104). Indeed, this would imply that ${I_0}$ always has the final word on $I$ ’s welfare, while there is no obvious reason to give normative authority to this self. Our approach is thus aligned with the critical assessment of the inner rational agent of Infante et al. (2016).

Knowing that the individual $I$ is embodied within different selves over time ${I_{C{\rm{'}}}},{I_{C{\rm{''}}}},{I_{C{\rm{'''}}}}$ , etc., this will likely lead to inconsistent choices as judged by ${I_0}$ (or equivalently, as judged by the theorist). According to the view from nowhere, the normative authority of $I$ lies in ${I_0}$ . This means that $I$ is willing to let the theorist choose for himself because nothing guarantees that the choice of ${I_C}$ will correspond to the choice of ${I_0}$ . In this sense, and with respect to our framework, Schelling (1980) takes the view from nowhere. By contrast, Sugden (2004, 2018) intends to address his recommendations to the ‘responsible person’ $I$ , who turns out to be embodied in every ${I_C}$ . The view from somewhere thus locates the normative authority of $I$ in each self ${I_C}$ . We share the position of Sugden that the correct addressee of normative economics should not be ${I_0}$ but $I$ – whom we call the enduring individual, to be distinguished from his multiple selves ${I_{C{\rm{'}}}},{I_{C{\rm{''}}}},{I_{C{\rm{'''}}}}$ . We suggest, however, that it is necessary for every ${I_C}$ to have a minimal ability of self-reflection, giving him the possibility to imagine himself in another context, and then to form counterfactual judgements about how ${I_{C{\rm{'}}}},{I_{C{\rm{''}}}},{I_{C{\rm{'''}}}}$ , etc. would choose in their own context. According to the view from manywhere, not only each self ${I_C}$ matters, but also his shared ability to imagine the choice problem from different perspectives, which eventually gives a normative authority to the enduring individual $I$ .

3. The view from Manywhere

Rather than leaving the task of defining the correct context to the theorist, or merely accepting the current context (which can be set by a third party, for better or worse), we propose that it should belong to the enduring individual to define what makes him better off. It is only through the fruitful confrontation of his different selves ${I_C}$ that $I$ will be able to choose what makes him better off, while limiting the risk of being manipulated or unduly influenced by a third party. We introduce the second-person standpoint in order to highlight the distinctive feature of our approach, and propose a normative criterion of self-determination that we define in this section.

3.1. The second-person standpoint

The second-person standpoint locates normative authority in others. It is because $I$ recognizes the authority of others to make claims and expect something from him that $I$ can be held morally responsible for his actions when they impact others. In his moral theory of the second-person standpoint, Darwall (2006) uses as an illustration the right a person $A$ has that others (a person $B$ , for instance) do not step on his feet. $B$ ’s stepping on $A$ ’s feet is morally questionable not only because it causes harm to $A$ (which is a third-person reason for not stepping on $A$ ’s feet), but also because $B$ violates the right $A$ has that others do not step on his feet. Unlike a general moral principle such as ‘do not cause harm’, this is a second-person reason for not stepping on another one’s feet because it has a direct addressee: the morally responsible individual $B$ . This implies that individuals should have the following characteristics:

1. (i) They have an authority, which is a condition for the validity of the reason one purports to address.

2. (ii) They must all presuppose that they share a common ‘second-personal’ authority, as ‘free and rational agents’ in Darwall’s (2006: 5) terms.

3. (iii) They share accountability relations, meaning that they have the ability to communicate and understand their demands to others.

In this sense, Darwall’s approach to morality is based on the search for the conditions that must be satisfied for normative claims to be considered as legitimate. His approach is reason-based. That is, not only $B$ should share a Humean sympathy for $A$ to perceive that $A$ is in pain, but should also consider ‘being in pain’ as an undesirable state of the world that $B$ has the possibility and reason to change. Just as $A$ has a second-person reason that $B$ does not step on his feet (because it is his right and a legitimate demand that $B$ has the ability to understand), $A$ has a second-person reason to not step on $B$ ’s feet since he may later feel guilty and hold himself responsible for violating $B$ ’s rights in return. What is fundamental to the second-person standpoint (compared with the first-person and third-person standpoints) is that the nature of normative statements is in individuals’ relations and ability to deal with these relations (e.g. demands, complaints, promises, consents, contracts), while first-person and third-person standpoints make normative statements based on an entity on its own: either the individual himself or the theorist herself who is confronted to a given problem.

Our approach can be seen as an extension of Darwall’s (2006) second-person standpoint between individuals to oneself. In the dessert problem, $I$ ’s afternoon self (who intends to be productive) has the right to claim that $I$ ’s current self does not ruin his later productivity by ordering a dessert. If $I$ ’s current self recognizes that this is a legitimate demand, he is accountable for actions that will affect his afternoon self’s productivity. However, the same is true of $I$ ’s current self. He has the right to claim that $I$ ’s afternoon self does not impose the ruining of the end of the meal by skipping dessert. If $I$ ’s afternoon self recognizes that this is a legitimate demand, then he cannot blame his current self for ruining his later productivity by ordering a dessert. Given the potential difference between the claims of the current and afternoon selves, it is necessary to settle this dispute by establishing a contract between $I$ ’s different selves.¹²

A way to minimize the effect of potential regret – which is a common argument in the literature to justify paternalistic interventions – is to ensure that the individual $I$ is a priori aware of such a potential disagreement, and that $I$ takes responsibility as an enduring individual for the actions of his different selves. In the restaurant, once $I$ is well aware of the tradeoff between an immediate enjoyment and a later productivity, $I$ can choose whether his current self should order the dessert or not, knowing the consequences for his afternoon self. As a result, none of his selves would be able to legitimately blame the other for taking the dessert or not. It is possible that $I$ ’s afternoon self later complains about the current self ordering the dessert, but this would not be a legitimate complaint if the dispute has already been settled. The problem of such a complaint is that the afternoon self wants the current self to skip dessert, while immediately enjoying the benefits in the afternoon. In this case, the afternoon self does not fully appreciate the opportunity cost that the current self supports.¹³ Managing the relationships between his different selves is the role of the enduring individual $I$ , whose identity defines the nature of these relationships.¹⁴ The adequate standpoint to form a judgement about one’s welfare is therefore the view from manywhere, since the relevant positions to judge what the ‘right’ choice is are the many positions from which one could see the choice problem.¹⁵

Rather than defining welfare with respect to an arbitrary criterion proposed by the theorist (that would tend to privilege the inner rational agent ${I_0}$ ), or systematically equating it with the individual’s observed preferences (meaning that such a criterion would always privilege the current self ${I_C}$ ), what matters in our approach is the process of confronting one’s different perspectives of a choice problem. It is questionable to state that either ${I_0}$ or ${I_C}$ should always have the final say on what is best for $I$ . On the one hand, imposing ${I_0}$ as the self who has the normative authority would be unacceptable for an individual who values spontaneity (which is hardly reconcilable with the consistent behaviour of ${I_0}$ ). On the other hand, imposing ${I_C}$ could be unacceptable for an individual who considers – upon reflection, from the perspective of another context – that some contexts are normatively unappealing. Rather than grounding normative authority on either ${I_0}$ or ${I_C}$ , our normative approach defends the idea that it belongs to $I$ to choose what is best for him.

As an illustration, consider the Asian disease experiment of Tversky and Kahneman (1981: 453). An unusual Asian disease is expected to kill 600 individuals, and you are asked to choose between two different health programmes. The first one is certain: you will save 200 individuals and let 400 die. The second one is risky, with 1 chance out of 3 to save 600 individuals (and no death) and 2 chances out of 3 to save nobody, with 600 deaths. The choice between the two programmes can be framed in terms of gain or loss, depending on whether we emphasize the number of lives saved of the number of deaths. The figures below correspond to the results of Tversky and Kahneman’s experiment – the % corresponds to the share of subjects who choose a certain programme and N corresponds to the total number of subjects per frame.

Frame ‘gain’ [N = 152]

A: 200 people will be saved [72%]

B: 1/3 probability that 600 people will be saved,

and 2/3 probability that no people will be saved [28%]

Frame ‘loss’ [N = 155]

C: 400 people will die [22%]

D: 1/3 probability that nobody will die,

and 2/3 probability that 600 people will die [78%]

This experiment is an example that preferences are likely to change depending on the way the choice problem is framed (here either in terms of gain or loss), even though a significant number of subjects may consistently choose A & C and B & D. The issue for the theorist is then to determine which programme is normatively preferred by individuals (if we can use this expression when facing such a stark choice), knowing that the preferences revealed by their choices are likely to depend on the way the choice problem is framed.¹⁶

According to the view from somewhere, there is not necessarily a problem in choosing inconsistently across frames: it belongs to ${I_C}$ in each choice problem to choose the best programme. According to the view from nowhere, ${I_0}$ ought to choose consistently across frames if the theorist imposes the principle of invariance, implying an equivalence between the choice problems A & C and B & D.¹⁷ According to the view from manywhere, what matters is that the individual embedded in a particular context (say, the gain frame) should also have the means to imagine herself in the other context (say, the loss frame). It is then only by becoming aware of these two frames – in terms of lives saved and lost – that $I$ will be able to avoid a genuine and unintentional framing effect. Eventually, this will lead him to determine the ‘best choice’, as judged by himself. What matters is however not his final choice (e.g. choosing A & D), but that he has the possibility to confront the different views prior to his final choice.

3.2. The self-determination criterion

We propose a formulation of a normative criterion that accounts for the confrontation of different possible perspectives. As stressed in section 2.2, we make a distinction between (i) motivational properties $P \in {{\cal M}_I} \subseteq {\cal P}$ , (ii) known properties $P \in {{\cal K}_I} \subseteq {\cal P}$ and (iii) relevant properties $P \in {{\cal R}_I} \subseteq {\cal P}$ . We also defined context properties as the properties which are motivational but not relevant: $P \in {{\cal C}_I} = {{\cal M}_I}\backslash {{\cal R}_I}$ . We also assumed, for the sake of simplicity, that the theorist correctly identifies the entire set of motivational properties ${{\cal M}_I}$ – i.e. she precisely knows to which properties the individual is behaviourally sensitive. The whole problem is that, from the perspective of the theorist, we cannot be certain that the properties in ${{\cal R}_I}$ are the correct ones (it is indeed the theorist’s own representation of the problem).

According to the view from nowhere, we have to imagine the behaviour of ${I_0}$ , who is supposed to exhibit context-independent preferences. Here since ${{\cal C}_{{I_0}}} = \emptyset $ , we know that the set of relevant properties precisely corresponds to the set of motivational properties. In this approach, the behaviour of ${I_0}$ gives direct evidence of $I$ ’s welfare. The problem is however that reaching a context-independent judgement on the counterfactual choice of ${I_0}$ may not be possible for an actual theorist. We will very likely see theorists stating their own value judgement on the problem at stake, even though some may genuinely believe that they provide a context-independent judgement. This implies that the view from nowhere cannot be implemented in practice.

According to the view from somewhere, the correct perspective to look at the problem is the individual himself. This means, in practical terms, that for any context $C$ , ${{\cal C}_{{I_C}}} = \emptyset $ . If a property is motivational, the theorist has no reason to consider that it is not relevant to $I$ . The solution offered here in order to identify the set ${{\cal R}_I}$ is (we judge) quite extreme. We have to reject the possibility of context properties, and include all motivational properties in the set of relevant properties (allowing the theorist to preserve the consumer sovereignty principle). The difficulty is that some individuals may still consider, when looking at a choice problem from another perspective, upon reflection, that their behaviour may have been unduly influenced. This means that ${I_C}$ may consider that ${{\cal C}_{{I_{C{\rm{'}}}}}} \ne \emptyset $ , implying that he is aware that – when $I$ will have to choose in context $C{\rm{'}}$ – $I$ as an enduring individual may regret his choice (or he has made a mistake).

Our proposition is therefore the following. It should be the responsibility of $I$ himself to judge whether he should choose according to the preferences of ${I_C}$ or ${I_{C{\rm{'}}}}$ . Because of the impossibility to observe directly the set of relevant properties of $I$ (properties which are constitutive of his welfare), yet knowing that some contexts could be perceived as problematic by some ${I_C}$ themselves, we propose to extend the informational basis from which $I$ forms his normative judgements.

Self-Determination Criterion. Consider the triplets of sets of properties ${\cal S} = \left\{ {{\cal M},{\cal K},{\cal R}} \right\}$ and ${\cal S}{\rm{'}} = \left\{ {{\cal M}{\rm{'}},{\cal K}{\rm{'}},{\cal R}{\rm{'}}} \right\}$ . ${\cal S}$ is strictly normatively preferred to ${\cal S}{\rm{'}}$ if ${\cal M}\backslash {\cal K} \subset {\cal M}{\rm{'}}\backslash {\cal K}{\rm{'}}$ .

Our normative criterion can then rank a triplet ${\cal S} = \left\{ {{\cal M},{\cal K},{\cal R}} \right\}$ without reference to the set of relevant properties ${\cal R}$ , i.e. the properties that the theorist considers as a priori relevant for the individual. This means that the judgement of the theorist on ${\cal R}$ , and therefore on the set of context properties ${\cal C} = {\cal M}\backslash {\cal R}$ , is irrelevant for this normative criterion. What matters is the set ${\cal M}\backslash {\cal K}$ , which corresponds to the set of properties which are motivational and not known. That is to say, it is preferable to be aware of more properties that are motivational than less.

This normative criterion can be interpreted in a ‘negative’ sense, i.e. it is preferable to limit the number of factors that influence $I$ without him being aware of this influence. When ${\cal M} = {\cal M}{\rm{'}}$ , a more ‘positive’ interpretation is that it allows the individual to consider the choice problem under additional new perspectives – such as the loss frame in the Asian disease experiment, which may considerably decrease the appeal of programme A. What matters here is $I$ ’s ability to accumulate and confront many views from different perspectives. Note that we do not expect individuals to be aware of all motivational properties and similarly, to look at every possible perspective on a given problem, which would require the extraordinary cognitive capacities of the inner rational agent. Our normative criterion states that it is better for individuals to be aware of more and more factors amongst all the factors that influence their preferences. If behavioural economists find a new result about how individuals behave, they ought to convey the information to these individuals rather than to benevolent nudgers who could use such knowledge as a way to promote what they think is best for these individuals.

Consider again the dessert problem. The only context property is the relative location of the cake, and it is unknown to the individual. Suppose that on a given day, the individual would choose the cake if and only if it is sufficiently ahead of others (e.g. the value of ${P_\ell }\left( {Cake} \right)$ exceeds a certain threshold ${\overline p_\ell }$ ), which means that the nudge of the restaurant owner is quite efficient. According to the view from nowhere, we have to imagine how the individual would choose in a context-free situation, e.g. if ${P_\ell }$ was not motivational. The problem is that this counterfactual scenario is hardly imaginable. There will necessarily be a value for ${P_\ell }\left( {Cake} \right)$ , meaning that we cannot conceive such a ‘context-free’ situation. The temptation for the theorist would then be to imagine that the correct choice of the individual corresponds in fine to her own conception of the good choice (e.g. to skip dessert and work harder later), and that the choice made in a counterfactual context-free situation corresponds to the choice made in contexts such that ${P_\ell }\left( {Cake} \right) \lt {\overline p_\ell }$ , precisely because it leads in fine to what the theorist considers as the good choice, i.e. skipping dessert.

This is however largely arbitrary. We could similarly argue that the correct choice for the individual consists in choosing the cake, and that he is occasionally refrained from doing so when the cake is in the back, e.g. because he fears the social stigma of voluntarily taking a slightly less accessible dessert. Our criterion states that the theorist should not try to identify a ‘correct’ choice. She should rather contribute to inform the individual that the relative location of items matter in his choice ( ${P_\ell }$ would then be a known property), and that the right approach to choose in such a situation is for the individual to imagine himself in the contexts where ${P_\ell }\left( {Cake} \right) \lt {\overline p_\ell }$ and ${P_\ell }\left( {Cke} \right) \geqslant {\overline p_\ell }$ , prior to making his final choice. Once the individual becomes aware of the different perspectives of the choice problem (when the cake is ahead of others or not), then we can consider that $I$ , as an enduring individual, can take full responsibility for his actions, and is likely to avoid an unintentional framing effect. Given the choice faced by the individual of an immediate enjoyment implying a delayed cost, we could also imagine that $I$ willingly chooses to use the context as a convenient heuristic to order the dessert or not. Using context-dependence as a strategy to settle complex choices is here seen as a perfectly acceptable choice process, even though the individual’s final choice may not be consistent – which is, in our approach, not the concern of the theorist.

A possible limitation of our normative criterion is that we only allow for comparisons between nested sets of ‘motivational and unknown’ properties, and are unable to rank overlapping sets according to the criterion.¹⁸ This, however, only poses a genuine issue in cases where we would need a precise measurement of property sets. For instance, imagine we have to choose between two programmes to fund: a prevention campaign against obesity based on messages highlighting either the addictive power of added sugars, or the effects of marketing on a given population. Both of them are properties which are motivational and probably unknown to many. Apart from those situations, which require additional elements to rank property sets, the main added value of our normative criterion is to shift the focus from outcomes and preference satisfaction to the factors that shape individual choices and the dynamics of preference formation. A solution would be to measure the ‘quality’ of different properties in terms of behavioural impact. That is, if a property has a significant effect on behaviour when unknown (such as the loss frame) compared with another (such as the relative location of the cake on the buffet), then it is a priori preferable to be aware of the former property. If we have at our disposal a metric for the degree to which properties impact individual behaviour when not known, we could also limit the risk of choice overload by targeting only the properties which have a relatively significant behavioural impact.

4. Opportunity, Positional Views and Diversity

In this section, we propose two justifications of our normative criterion in terms of Sugden’s (2004, 2018) opportunity criterion, and Sen’s (2009) ‘positional views’ in his theory of justice. We also relate our approach to the contemporary theories of justice of Muldoon (2016) and Gaus (2016), which account for the diversity of perspectives in their characterization of a just society.

4.1. Opportunity extended

We can first interpret our normative criterion as an extension of Sugden’s (2018: Ch. 5) individual opportunity criterion to the process of preference formation. Sugden emphasizes that the theorist should not try to answer for others the question ‘how should one live?’ (which is, according to him, a question that should be left to ethicists) but rather the question ‘how do we live together?’ (which is a question that can be addressed by economists). In his approach, societies must be such that the opportunity of individuals to act on any preference should be maximized. A point that is however not discussed by Sugden (2018) is whether individuals have the ability to merely conceive other preferences. We follow Davis (2011), who argues that individuals have the capacity to critically reflect upon their own evolving preferences and to continually redefine their identity through self-reflection. If we accept such a dynamic view of personal identity, we should consider that individuals’ preferences do not pre-exist the choice situation, but are rather progressively determined by the process of choice. As Nozick (1981) puts it,

The reasons [considered in deliberation] do not come with previously given precisely specified weights; the decision process is not one of discovering such precise weights but of assigning them. (Nozick 1981: 294)

If what matters is that individuals can satisfy any preference they might have, this necessarily requires ensuring that social institutions allow individuals to experience different preferences. This means fostering (or at least not deterring) the creativity of individuals. A good way to ensure that individuals do not stay stuck in a single conception of their life is to make sure that they have the opportunity to learn. The process of preference formation is likely to be path-dependent, i.e. it depends on the individual’s initial sets of representation of the world. If the individual is only aware of one way to look at the world (e.g. he always chooses the cheapest good without considering some counterparts, such as the working conditions of the employees who have produced the good), then his opportunities to learn new preferences are very likely to be reduced (see Schubert’s (2015) ‘opportunity to learn’ criterion). Yet if we value opportunity with choice sets and accept that one’s identity is the result of an evolving process and critical reflection upon one’s experiences, then opportunity also seems to be valuable when considering the sets of possible future identities (Buchanan 1999; Dold 2018). In line with Dold and Schubert (2018), our normative criterion emphasizes that economists should contribute to improve the process through which $I$ forms his own preferences, and not merely focus on the satisfaction of his preferences.

The problem one may have with an approach that promotes ‘teaching’ new preferences or properties to individuals is that the theorist can be suspected of teaching only what she thinks are valuable preferences (e.g. preferring cycling over taking one’s car, or saving more for one’s retirement). Our aim is not to promote a normative approach in which we should tell individuals what ‘good’ preferences are, but rather to make sure that they are well informed, i.e. that they are aware of what makes them prefer some options over others. In this regard, it is preferable for an individual to have an unhealthy behaviour (smoking, overeating, drinking too much alcohol) while being perfectly aware and fine with the reasons that led him to have such behaviours, than for the same individual to have a much healthier behaviour while ignoring why he behaves like that (e.g. because of some norms of fat shaming, and/or overvaluing physical appearance).¹⁹

4.2. Positional views

Another justification of our normative criterion is the parallel we see with Sen’s (2009) concept of ‘positional views’ in his theory of justice. Baujard and Gilardone (2017, 2019) emphasize that Sen’s theory of justice is ‘poorly understood’, which is why it continues to raise many debates, such as the proper place of capabilities in his theory of justice. We suggest that one of the reasons of these misunderstandings is that Sen endorses the second-person standpoint in his theory of justice. Sen explicitly rejects the third-person standpoint, which consists of defining a normative criterion in a transcendental perspective. At the same time, he considers that individual preferences are not necessarily informative about individual welfare. This is because of the phenomenon of adaptive preferences: first-person judgements about one’s welfare can depend on one’s current state, e.g. deprivation. Our argument is based on the same concerns. We consider that the theorist should refrain from imposing what she thinks individuals should prefer (because of the arbitrariness of such a standpoint), but at the same time, she should acknowledge that there might be problematic situations in which satisfying individuals’ preferences may not be in their best interests.

Discussing the case of normatively problematic preferences in Sen’s work (1985, 1992, 2001), Sugden (2006: 34) argues that the position of Sen implies ‘that “we”, as ethical theorists, can claim to know better than some particular individual what is good for [him]’. According to Sugden’s reading of Sen, the viewpoint from which this normative judgement is formulated is ‘the viewpoint of a moral observer, making “judgements”, “assessments” or “evaluations” of a person’s state from outside’ (Sugden 2006: 36). Theorists would then impose as the social standard ‘the kinds of lives that a majority of [our] fellow citizens, after reflective deliberation and open debate, judge to be valuable’ (Sugden 2006: 40). However, Sen (2006: 89) rejects this ‘monstrous political philosophy’. He argues that the emphasis should be put on the process through which individuals’ desires are formed and confronted, while guaranteeing in fine the respect of their own preferences. It is true that the theorist has predefined ideas about what counts as a good life, but such ideas should only serve as inputs among other inputs in the collective discussion.

Our proposition turns out to be very similar to Sen’s normative approach. In our view, the role of the theorist should not be to impose her own conception of a good life, nor to convince individuals to endorse what is collectively considered as a good life. Instead, her role is to promote their ability to enter into a phase of self-reflection, so that they can have the opportunity to choose what is best for themselves. Promoting individual self-reflection does not mean that individuals ought to endorse a ‘rational’ behaviour – such as in related approaches assuming ‘true’ preferences (Camerer et al. 2003; Thaler and Sunstein 2003, 2009). Instead, it means ensuring that individuals do not stay trapped in a behaviour they could consider as problematic if they became aware of the reasons leading to such a behaviour. In the dessert problem, ‘being within sight’ may be a necessary condition for many individuals to actually order the dessert – implying that $I$ never orders the dessert when it is not within sight. The self-determination criterion means that it is better for $I$ to be able to imagine what he would do in another context (when the dessert is within sight). Whether $I$ turns out to order the dessert much more frequently (because he considers that the location of the dessert does not matter that much) is however not relevant with respect to the self-determination criterion.

According to Baujard and Gilardone (2019), the concept of ‘positional views’ in Sen’s (2009) The Idea of Justice accommodates the two positions of (i) not imposing the preferences of the theorist and (ii) acknowledging that some preferences might be problematic from individuals’ own perspective. A positional view is defined as ‘an individual judgement towards any social state, considering objectively the context from which she or he is able to assess this social state’ (Sen 2009: 3). One’s positional view can evolve if some information from different positions is communicated. For example, the individual’s judgement on a policy debate (like the implementation of a universal basic income) is likely to depend on various elements that characterize his current position (positional parameters), such as being in a situation of poverty or not, being in a health condition that limits his opportunities of employment, etc. Positional views can be ‘objectivized’ because they can be described by the individual, e.g. being poor or not, or being handicapped or not. By ‘objectivized’ we do not presuppose that there is one ‘correct’ way to describe such positional views, or that such positional views are objectively true or have an independent existence, but that they can be made understandable to others. Since such positional views ‘(1) may influence observation and (2) can apply to different persons’ (Sen 1993: 127), they can constitute a relevant input for collective choice. The confrontation of such positional views in a phase of public deliberation then contributes to the emergence of an intersubjective view on the question. We advance that it is only by confronting many views from different positions that individuals can collectively form an enlightened judgement about a specific state of affairs. The relevant inputs for normative analysis are therefore the views of all the various individuals that constitute society. Confronting different positional views offers a way to widen the informational basis of all participants.

The approach we propose – the view from manywhere – is based on a similar reasoning. Each self ${I_C}$ can ‘objectivize’ his current position/context $C$ based on the various properties $P$ that characterizes $C$ . If ${I_C}$ has the ability to imagine the context $C{\rm{'}}$ of ${I_{C{\rm{'}}}}$ that is described by different values of the properties $P$ , then ${I_C}$ can enter into this phase of self-reflection and form in fine an intrasubjective judgement. The normative criterion of self-determination means that we should widen the number of properties the individual is aware of. This means increasing $I$ ’s ability to characterize the various contexts $C$ and then widening the informational basis of his intrasubjective judgement. Just as in our proposition, Sen’s (2009) theory of justice shifts normative appraisal from outcomes to the process of choice, and does not make any presumption about the outcome of public deliberation. This means that there is no ready-made theory of what a ‘good’ society is (or what ‘good’ preferences are) but that there is a general approach (confronting the different views on the same question) that contributes to form collective judgements. If the theorist turns out to be aware of some positional parameters (in our approach, the properties $P$ ) then she ought to inform individuals about these properties. Whether individuals eventually take them into consideration is however not relevant to the theorist. What matters is that each self ${I_C}$ has the ability to imagine what his preferences in other contexts would be, and possibly choose to act on the basis of the preferences of another self ${I_{C{\rm{'}}}}$ , depending on how $I$ – as an enduring individual – defines his identity and relationship between his different selves.

4.3. Diversity in non-ideal worlds

We now compare the ‘view from manywhere’ to two approaches in social contract theory that specifically account for the diversity of perspectives in public deliberation: Muldoon’s (2016) Social Contract Theory for a Diverse World and Gaus’ (2016) The Tyranny of the Ideal. In a nutshell, we argue that our proposition is consistent with these theories of justice, although there are some contrasts to be emphasized.²⁰

In Social Contract Theory for a Diverse World, Muldoon (2016) accounts for an important aspect of social contract theory that, in his terms, has not received sufficient attention for addressing the contours of a just society in our contemporary societies: the increasing diversity (or heterogeneity) of individual actors resulting from for example international mobility or intercultural exchanges. According to Muldoon, in a world that contains more and more diversity of perspectives, not only is it necessary to have a theory of justice that accounts for disagreements of the desirability of some outcomes (which is already captured by standard approaches of social contract theory, such as Rawls’ (1999) A Theory of Justice) but also for disagreements about the definition of the available options. Muldoon proposes a typology which contrasts the views from somewhere, nowhere and everywhere. He argues that the ‘view from everywhere’ is epistemically superior to the two others because it combines the ‘neutrality’ (in his words) of the view from nowhere with the importance of particular interests and desires of the view from somewhere. Our overall argument is fundamentally consistent with Muldoon’s defence of the methodological superiority of confronting different perspectives to form normative judgements.

However, our approach departs from Muldoon in at least one aspect: the term ‘everywhere’, if taken prima facie, requires to consider every perspective of a choice problem. Yet, in practice, one may doubt about such a possibility. In fact, such a problem seems less salient when one accounts for interpersonal normative evaluation – which is Muldoon’s account, as well as the standard account on public reason – rather than intrapersonal relations. Indeed, while interpersonal deliberation requires to take into consideration the perspective of every individual belonging to a group (e.g. voting for a presidential election), intrapersonal deliberation requires to take into consideration every possible perspective of a single individual (e.g. all selves voting for eating the cake or not in the dessert problem). In other words, intrapersonal deliberation requires to account for the psychology of individuals. Psychological accounts, however, are given little consideration in theories of justice (an exception is Gaus (2016), to be discussed below). Yet, in our view, the empirical findings of behavioural economics specifically emphasize the need to account for the limited cognitive abilities of individuals when they make (intra- or interpersonal) deliberations. Also, neuroscience provides significant evidence that our perceptions as human beings can be very different from one person to another.²¹ As previously stated, considering every perspective of a choice problem is an exclusive capacity that only the inner rational agent could have, hence the use of the alternative term manywhere – which remains compatible with a more realistic account of individual psychology.²²

In The Tyranny of the Ideal: Justice in a Diverse Society, Gaus (2016) takes the same path of Muldoon by giving importance to the diversity of perspectives in non-ideal worlds. He argues that such a diversity of perspectives is actually more beneficial under certain conditions to solve a problem than looking at it from a single perspective – this, according to Gaus, even if such a single perspective is judged to be the ‘best way’ to look at the problem (Ch. 3). In contrast with Muldoon, Gaus discusses at length the cognitive aspects of reflexivity and deliberation, with a specific focus on the work of Hong and Page (2001) about the processes by which individuals with different representations of some problems and limited abilities identify optimal solutions. In particular, this refers to the formation of the internal representations of the problems by the individual and the heuristics they use to generate solutions based on those representations. Gaus offers a more in-depth analysis of the characterization of an evaluative perspective, which requires the following three elements (Ch. 2: 43).

1. 1. A set of evaluative standards or criteria by which alternative social worlds in a domain are to be evaluated.

2. 2. A specification of the world features that are relevant to evaluation. Those include external properties of the world such as institutions, but also internal properties such as psychological facts.

3. 3. A mapping function that takes the evaluative standards and applies them to the world features.

This specification is related to our formal definition of context-dependence, since we consider each self ${I_C}$ as offering a possible evaluation of an alternative $x$ . Each alternative is defined by properties $P$ which may be used in normative evaluation, and the evaluation is realized by confronting the evaluations of the different ${I_C}$ by aggregating their judgements based on the various properties. The ‘mapping function’ could thus be approximated by the function ruling the aggregation of intrapersonal views. Similarly to our proposition, Gaus emphasizes that there is no ‘ideal’ society, and that the social contract must be based on minimal agreement for public deliberation, which relates to our criterion of minimal capacity of self-reflection for individual deliberation.

In short, our approach is consistent with Muldoon (2016) and Gaus (2016), and more particularly with Gaus’s psychological account of deliberation and public reasoning. We therefore highlight possible cross-fertilization between, on the one hand, the literature of behavioural normative economics – with the issue of intrapersonal aggregation of preferences – and, on the other hand, the literature of political philosophy and social contract theory – raising the importance of diversity of perspectives in interpersonal aggregation of preferences.²³

5. Concluding Remarks

We have proposed a novel approach to normative evaluation when preferences are context-dependent. Based on the distinction between the different standpoints one could endorse when offering a normative judgement, our proposition is to base normative evaluations on the second-person standpoint, according to which the definition of one’s welfare should be the result of an intrapersonal discussion and fruitful confrontation of different perspectives of a choice problem. This approach – that we label the ‘view from manywhere’ – states that our aim, as theorists, should be to widen the informational basis of individuals so that they can have the opportunity to adapt their own preferences, as judged by themselves. This can solve some of the issues of the first-person and third-person standpoints, since the view from manywhere (i) remains agnostic about how individuals should choose and (ii) ensures that individuals have the means to understand why they choose as they do, and the possibility of changing their choices if they want to. We see promising applications of our theoretical/philosophical framework. An existing experimental design that can be seen as a direct application of our proposition is the one of Benjamin et al. (2020). The authors ask subjects to make identical retirement investment choices in different frames then give them the opportunity to revise their choices, while also questioning those who changed their choices about the reasons they have actually done so, for example ‘I made a mistake when I first chose’, or ‘answering all of these questions made me change what I want’. Related to the literature of behavioural normative economics, our proposition also bears similarity with the methodology proposed by Ferreira (2023), who accounts for what he calls ‘confirmed choices’ as a ‘proxy of subjective welfare’.

By conferring a significant role to the external observer (the theorist being the only judge of the normative authority of individual preferences), the third-person standpoint may easily give a justification for paternalistic regulation such as nudging, in order to steer individuals’ behaviour into what they think is the ‘right’ direction (Thaler and Sunstein 2009). The first-person standpoint, on the other hand, takes individuals’ preferences as their protected sphere of liberty and rejects any policy intervention aiming at interfering with the expression of individuals’ freedom of choice. In this approach, whatever individuals’ preferences turn out to be, individuals are considered as responsible persons – i.e. they are accountable for their own choices. The implication is that there is no need for paternalistic regulation. The aim of the theorist is here only to ensure that society is organized in such a way that individuals can express their freedom of choice. In this approach, the exclusive aim of public policy should be to guarantee some rules that allow to enhance individuals’ opportunity, such as rules of fair competition. By focusing on the internal process of preference formation, the second-person approach offers a rationale for more ‘educational’ policies. The aim is to foster individuals’ abilities to critically reflect on their own preferences and increase their opportunities to learn new preferences. In our view, theorists should ensure that society guarantees that each individual can learn new properties and accumulate complementary views on different choice situations. This is typically the philosophy of the boost agenda (Grüne-Yanoff and Hertwig 2016) inspired by Gigerenzer and Todd’s (1999) simple heuristics programme, and by Gigerenzer’s (2015) call for reforming school curriculums in order to improve individuals’ health, financial and risk literacy.