Temporal Partiality and the Veil of Ignorance

Todd Karhu; Todd Karhu

doi:10.16995/fe.23197

TEMPORAL PARTIALITY AND THE VEIL OF IGNORANCE

Todd Karhu

A prominent approach in moral and political philosophy holds that conclusions about distributive justice can be derived from what rational, self-interested individuals would want when situated behind a veil of ignorance that prevents them from knowing their personal characteristics or positions within society. One basic feature of human self-interested reasoning is the preference for our bad experiences to be in the past, even if they would be less bad if they were still to come, and our good ones to be in the future, even if they will be worse. This article demonstrates that individuals behind the veil who have that preference will unanimously favor distributive policies that are manifestly unjust. Since there is good reason to think that caring more about one’s future than one’s past is not irrational, we should be skeptical that distributive justice reflects rational self-interest from behind a veil of ignorance.

I. Introduction

How should the just society distribute its resources among its members? One prominent answer, which I shall call the Veil of Ignorance Thesis, says that it should aim to distribute its resources in the way that rational individuals, aiming to advance their own interests, would choose when deprived of knowledge of their personal characteristics or positions within the society. The Veil of Ignorance Thesis is attractive for many reasons, not least because it promises to inherit the virtues of earlier social contract theories while recommending social and economic institutions that align with the liberal egalitarian vision that no one should be disadvantaged on the basis of factors like their gender, social status, perceived race, or parentage.

Some skeptics of the veil of ignorance have questioned the very idea that what people would have agreed to under conditions of ignorance bears on what can justifiably be imposed on them when that ignorance is gone.¹ Even among those who accept its basic premise, however, there remains considerable disagreement about what arrangements the parties behind the veil would actually settle on. Some of it stems from prior divergence over what information it is appropriate to exclude with the veil, since what it is rational for the parties to agree to will in large part depend on what they can predict about their actual circumstances when the veil is lifted.² But philosophical dispute over what would be chosen from behind the veil can also result from disagreement about the structure of rationality itself.

Disagreement of this latter kind has usually centered on what attitudes it can be rational to take toward risk or uncertainty.³ By contrast, this article explores the prospects of the Veil of Ignorance Thesis in light of a very different aspect of self-interested reasoning: our disposition to want our good experiences to be in the present or future rather than the past, even when they will be less good for it, and our bad experiences to have already happened rather than be still to come, even if that means they will have been worse—a pair of preferences that have together become known as our future bias.⁴

Future bias is among the most universal aspects of human self-interested reasoning and reacting, firmly ingrained in our psychology and deeply rooted in our intuitive understanding of what makes something in or against our self-interest. Virtually everyone is relieved when their pain or discomfort is over and done with, and glad to reflect that a joyful experience is still to come—a commonplace that is by now well supported by empirical literature.⁵ Crucially, however, it isn’t just that we are future-biased. As Caspar Hare succinctly puts it, it also “seems to us that it makes sense to be future-biased.”⁶ One way to see the force of this latter point is to contrast the evaluative stance we would take toward our future bias with the one we take toward another temporal bias we exhibit: the bias toward the near over the further future.⁷ Thus imagine a person who always postpones her painful experiences for as long as she can, even when this means they will be much more painful when they arrive, and who tries to get the good things in her life as soon as possible, even when they would have been much better had she waited for them. Although this person’s tendencies are all too familiar, they strike most of us as less-than-ideally rational. Indeed, procrastination is something we work hard to combat, and delaying gratification is virtually the hallmark of rational prudence. In stark contrast, we would hardly tend to think that a person who is glad that some episode of agony is now in the past instead of still to come is making a rational mistake. That she has good reason to be glad seems almost too obvious to state; after all, it’s over and done with.⁸

Unlike the so-called “bias toward the near,” then, at least on the face of it, future bias strikes us as perfectly rational. If that assessment is correct, however, then the Veil of Ignorance Thesis needs to be rejected—or so I shall argue. In brief: because they strongly discount the past, future-biased individuals behind a veil of ignorance will tend to favor social and economic arrangements that impose heavy burdens on people in their youth in exchange for only minor benefits to people in old age, when they could instead have chosen to give everyone in society far more wellbeing and spread it evenly across age groups. Evaluated from outside of the veil, such a choice is manifestly unjust. It is equivalent to taking material goods from the young and investing those goods in a way that is sure to return to them in old age a fraction of the original value—a decision that no one would think justifiable on its own terms. Thus, if future bias is rational, truths about distributive justice cannot be derived from rational self-interest from behind a veil of ignorance.

This is a significant conclusion, which would seem to cast doubt on some important arguments about equality, the distribution of rights, and utility maximization. Nevertheless, my target in this article is not any particular author or theory. The veil of ignorance is probably best known from the work of Rawls and Harsanyi. They arguably conceive its role somewhat differently, and, when it comes to Rawls in particular, there is disagreement about how necessary the device even is for his arguments. What I am calling the Veil of Ignorance Thesis is fundamentally a view about justification; it is the proposal that a policy or principle can be assessed for its justness by asking whether it would have been chosen by rational individuals behind a veil.⁹ I take this to be an attractive idea in its own right, and one whose intuitive appeal I suspect accounts for much of the wider popularity of Rawls’s theory, whatever his actual view was. But if certain authors conceive the role of the veil differently, or turn out not to need it at all in order to establish their desired conclusions, all I can say is, so much the better for them.

Here is an outline of the article. Section II sets the stage for the main argument by sketching a simplified version of it, in which we ask what would follow from the rationality of future bias if the parties behind the veil were all in the middle of their lives. I show how, under that supposition, future-biased parties would favor distributive policies which are manifestly unjust when evaluated from outside the veil. Section III then does away with the illustrative assumption and generalizes the result, showing how it works when the parties have no knowledge of how old they are. I demonstrate how the rationality of future bias undermines the Veil of Ignorance Thesis given any plausible assumption about the parties’ attitudes toward risk and uncertainty, as well as given any reasonable proposal for what, if anything, they can predict about their age when the veil is lifted, or about the likelihood of surviving until old age if they are young.

Section IV discusses objections and additional issues. In Section IV.A, I consider alternative standpoints that the parties could take from behind the veil—namely, the standpoint of their whole life rather than that of particular times within it, and the standpoint of someone just born into their society. I show how neither of these alternative ways of constructing the parties’ task will save the Veil of Ignorance Thesis. Section IV.B explains how the argument works when the parties try to account for the interests they might have in how other people fare. Section IV.C discusses the moral irrelevance of hypothetical choice vis-à-vis hypothetical endorsement, thereby dispelling a worry that my arguments depend on the parties being able to affect the past. Section IV.D explains how my argument against the Veil of Ignorance Thesis goes through as long as it is even rationally permissible to care more about one’s future than one’s past. I conclude in Section V by asking what, if anything, can be preserved of the veil of ignorance.

Before proceeding, I should be explicit about two assumptions I will be making as I go along. The first is that the parties behind the veil are not biased toward the near: they do not prefer goods to be sooner rather than later, holding fixed their certainty and intrinsic value. Since near bias is widely suspected to be irrational, this assumption may not be very controversial, but, as I explain in Section IV.D, my arguments actually require only the far weaker premise that near bias is not rationally mandatory. Second, I will assume that the parties are deliberating on the basis of rational self-interest, evaluating policies in light of the positive and negative effects they can be predicted to have on their own lives, but not on the lives of others. As I will show in Section IV.B, however, my arguments go through even if the parties take into account concern they may have for other people’s welfare, in addition to their own.

II. Setting the Stage: Parties in the Middle of Their Lives

What kinds of policies would the parties behind the veil of ignorance endorse if they were biased toward the future? To answer this, it turns out that we need to first ask what they can know about how old they are. It might seem clear that the veil, properly conceived, should keep the parties ignorant of their age. Age, after all, like race, gender, or disability, is in many contexts thought to be an unjustifiable basis for political decision-making, discrimination on the grounds of which amounts to at least a prima facie injustice. Still, I want to start out by illustrating my argument with an assumption, only adopted for the sake of exposition, that the parties doing the deliberating are all in the middle of their lives and know this. Obviously enough, this is not a legitimate assumption to make in a full criticism of the veil of ignorance, but it helps to set the stage for the main argument. In the subsequent section, I’ll show how the argument in this one generalizes to the case where the parties don’t know their age.

Asking what policies future-biased people in the middle of their lives would endorse turns out to be relatively straightforward. Because the parties know they will not turn out to be children or young adults when the veil is lifted, for any given benefit, they will rationally want it to accrue to those who are no longer young, thereby ensuring that they receive the benefit in their present or future instead of in their past. Conversely, for any given burden, other things being equal they will prefer the burden to be borne by people in their youth, thereby ensuring that it is not in their present or future when the veil is lifted. Thus, as a general rule, future-biased parties who know that they are in the middle of their lives will favor a society which concentrates the benefits up for distribution later in life, and the burdens earlier in it.¹⁰

We can make this result concrete by comparing two hypothetical distributions between two different age groups, which we can label the Young and the Old (Table 1). Numbers will represent the average amount of resources or wellbeing that the members of a given age group enjoy. (I will use “resources” and “wellbeing” as placeholders; readers are free to replace them with their preferred distribuendum, be it utility, primary social goods, or what have you.)

Table 1

	Young	Old
D1	50	50
D2	20	80

Future-biased parties behind a veil of ignorance who know that they are in the middle of their lives will prefer D2 to D1 because in D2, the 30 units of wellbeing at stake are in their future rather than their past.

To some, a society resembling D2 will already be unjust by comparison with one like D1, which contains the same total wellbeing but spreads it uniformly across age groups.¹¹ But this is not obviously so; although under D2 people are worse off early in their lives than they would be under D1, they are to that extent compensated by being better off when they are older.¹² So advocates of the Veil of Ignorance Thesis may not be too worried even if they end up forced to conclude that rational self-interest behind the veil favors distributions like D2 over those like D1.

Crucially, however, our bias toward the future does not apply only to the temporal location of benefits or burdens when we hold constant their magnitudes. More than that, we would rather have much greater past burdens than far smaller future ones, and would rather have a much smaller benefit in our future than a far greater one in our past.¹³

Just how strongly future-biased we are is up for debate. Some report and endorse absolute future bias,¹⁴ meaning that they completely prudentially discount their past benefits and burdens.¹⁵ But friends and enemies of future bias alike agree that the asymmetry between our preferences over the past and the future is at least a very stark one. This is evidenced by a convergence on Derek Parfit’s response to his classic illustration of the phenomenon:

Your Past or Future Operation You have been admitted to the hospital to undergo an operation that is intensely painful but otherwise safe. So painful is this operation, in fact, that patients are afterwards administered a drug that erases their memory of undergoing it. Now you wake up, unable to recall whether you have had the operation yet. The nurse on duty can only give you the following information: If the operation happened yesterday, it lasted for ten hours; if it is scheduled for tomorrow, it will last for one hour.¹⁶

In this predicament, most of us would hope to have already had the operation, even though that means it will have been ten times as long.¹⁷ Extrapolating a bit, we might reasonably take as a working assumption that we discount the past at a rate exceeding 10:1. If so, then future-biased parties behind the veil who know that they are in the middle of their lives would prefer an arrangement like D3 to D1 in Table 2.

Table 2

	Young	Old
D1	50	50
D3	10	55

For future-biased people in the middle of their lives, D3 would involve losing 40 units of past wellbeing for a gain of 5 units of future wellbeing, which they would view as a worthwhile trade-off. However, evaluated from outside of the veil of ignorance, the choice of D3 over D1 is manifestly unjust. It imposes an enormous loss on people early in life in order to confer a far smaller benefit later in life. Virtually no one would think that a society’s decision to implement D3 instead of D1 was defensible. If rational parties behind the veil of ignorance really would make a choice like that, we should be very skeptical of the Veil of Ignorance Thesis.¹⁸

III. Selecting Distributions in Ignorance of One’s Age

That is how the argument might work if we stipulated that the parties could know they are in the middle of their lives. Obviously enough, a more reasonable construal of the Veil of Ignorance Thesis has the parties deprived of knowledge of how old they are. And there are straightforward rationales for construing it that way. For one thing, age might be something arbitrary that should be excluded in the name of impartiality; for another, we may think that everyone who is affected by a distribution—or at least everyone who contributes to the benefits of social cooperation—should be represented in the decision procedure. The next task, then, is to show how the form of argument in the preceding section goes through even when we assume that the parties do not know what age they are.

Perhaps unsurprisingly, it turns out that what distributions such parties would endorse depends on what, if anything, they can predict about how old they will be when the veil is lifted. The two plausible suggestions for what they can predict seem to be: (a) they should assign an equal probability to being any age in their society, and (b) they should not be able to assign any probabilities at all.¹⁹ I will argue that on either proposal, future-biased parties behind the veil will endorse policies which are manifestly unjust, placing enormous burdens on the young in exchange for only minor benefits to the old.

III.A RISK-NEUTRALITY

The first proposal is that the parties should assign an equal probability to being any age in their society. This proposal may have an important principled rationale. Perhaps, as Rawls proposed, the veil should exclude not only knowledge of who one is, but also knowledge of the likelihood of being one person or another, of having one set of characteristics rather than some other.²⁰ In Section III.B, we will examine uncertainty-aversion, including the maximin rule. But a well-known alternative is the principle of indifference, according to which when one does not know the probabilities of the different possible outcomes one faces, rationality requires having the same credence in each of them.²¹ If the veil should prevent the parties from knowing the probability of being a given age, applying the principle of indifference can instruct them to assign the same probability to being any age in their society.

To know how the parties would choose once they assign an equal probability to being any age, we next need to ask what attitudes they would take toward risk. In this subsection, we will consider future-biased parties who are risk-neutral. For simplicity, let’s once again divide society into two age groups: the Young and the Old. It will be easiest to begin by considering parties who discount their past wellbeing absolutely. With those stipulations in place, the expected value of a given distribution for an individual who is equally likely to be young or old is to be calculated as follows. We take the average of (i) her future wellbeing conditional on her being young (arrived at by adding together the Young and Old columns in a given row) and (ii) her future wellbeing conditional on her being old, which just corresponds to the value in the Old column of that row. Thus, for absolute past-discounters who are risk-neutral and equally likely to be young or old, the expected values of the distributions in Table 2 run as shown in Table 3.

Table 3: Expected values of Table 2 with absolute past-discounting.

	Young	Old	Expected value
D1	50	50	$100 + 50 2$ = 75
D3	10	55	$65 + 55 2$ = 60

As we can see, D1 now has the greater expected value. Between these two distributions, the parties would correctly endorse the just option, which distributes more total wellbeing uniformly across the groups, rather than the manifestly unjust alternative of D3 that was chosen by parties who knew they were in the middle of their lives.

As it turns out, however, the preference for D1 obtains only because the trade-off involved in moving from D1 to D3 is so extreme. Relative to D1, D3 imposes a loss of 40 units of wellbeing on the young while conferring a gain of only 5 units on the old. Provided we just make that trade-off somewhat less extreme, the parties will opt to impose large burdens on the young for the sake of much smaller benefits to the old, as Table 4 shows.

Table 4

	Young	Old	Expected value
D1	50	50	$100 + 50 2$ = 75
D4	10	71	$81 + 71 2$ = 76

To be sure, the choice of D4 over D1 is not as seriously unjust as the choice of D3 over D1. While both D4 and D3 impose a loss of 40 units of wellbeing on the young by comparison with D1, under D4 the young can expect an additional 21 units in old age compared to D1, as opposed to the mere 5 additional units they could expect under D3. Nevertheless, from the standpoint of justice, D4 remains obviously inferior to D1. By choosing D4, it is as though society takes material goods from the young and invests them in a way that is sure to return to them in old age barely half of what those goods were worth. Yet this is precisely the sort of trade-off that future-biased parties behind the veil tend to favor. For risk-neutral, absolutely future-biased parties in particular, each unit given to the Old is worth twice as much in expectation as a unit given to the Young, since benefits to the Old matter regardless of one’s age, while benefits to the Young matter only to those who are young. This means that such parties will endorse any distribution that sacrifices up to two units of the Young’s wellbeing for each unit gained by the Old, as D4 does relative to D1. If it really is rational to choose that way behind the veil, then we should be skeptical of the Veil of Ignorance Thesis indeed.

Relaxing the assumption of absolute past-discounting will attenuate the result, but only barely. If the parties discount their past wellbeing at a rate of 90%, D1 will win out again, but only just, as Table 5 shows.

Table 5: Expected values of Table 4 with 90% past-discounting.

	Young	Old	Expected value
D1	50	50	$100 + 50 + (0 .1)50 2$ = 77.5
D4	10	71	$81 + 71 + (0 .1)10 2$ = 76.5

As it turns out, however, we only need to increase the average wellbeing of the young in D4 from 10 units to 12 for D4 to win out.²² That would still involve imposing a loss of 38 units of wellbeing on the young for the sake of conferring a benefit of merely 21 units on them in their old age. There can be no justification for such a choice. Indeed, even if we lower the rate at which the parties discount their past wellbeing to 80%, we only need to increase the wellbeing of the young in D4 to 16 units for that distribution to be preferred.²³ Viewed from outside the veil, these decisions are manifestly unjust.²⁴

At this point, one might object that a society structured like D4 would have certain extrinsic defects that make choosing it contrary to the parties’ overall interests, even if it seems like it would be desirable to them when we are focused more narrowly on their personal benefits and burdens. For instance, if they anticipate that the young would refuse to comply with an arrangement that burdened them so heavily, the parties might not choose D4 because of worries about its instability. But for the argument against the Veil of Ignorance Thesis to work, we need not imagine that the young are being enslaved or treated like human guinea pigs. The distribution can be very unjust without being so intolerable as to realistically threaten instability, since people routinely tolerate serious distributive injustice even when better alternatives exist.²⁵ A similar response can assuage the worry that the parties would not choose a distribution like D4 since they would simply find it unappealing to live in a society that burdens its young so heavily, whatever their own age turns out to be. The central injustice does not depend on the young being miserable in absolute terms but on the indefensible shortfall between how they fare under D4 and how they could have fared under D1. Alternatively, one might think the parties would not pick the unjust distribution because in a society which so overburdened the young for the sake of the old, young people would lack the social bases of self-respect. But being treated unfairly is not the same as being treated as inferior. Age-based inequality is different from a caste system: everyone is generally expected to pass through the stages of life, so the young are not thereby being marked out as fundamentally subordinate to others.

III.B RISK- AND UNCERTAINTY-AVERSION

We have seen that future-biased parties behind the veil would endorse manifestly unjust distributions of benefits and burdens across age groups, if those parties are risk-neutral. Yet it is plausible that rationality permits or even requires being averse to risk, at least when the stakes are high. A risk-averse agent is one who, when faced with a gamble, tends to prefer to mitigate her potential losses, even if this means accepting a lower average payoff. (For example, someone who is risk-averse with respect to her own wellbeing might prefer having 40 units of wellbeing for certain to having a fifty-fifty chance of receiving either 0 or 100 units.)

It may seem that the rationality of risk-aversion can help shield the Veil of Ignorance Thesis from the challenge raised by future bias. For it may seem that risk-averse parties behind the veil would be unwilling to take the chance of a policy which imposes huge burdens on the young for the sake of the old, lest they turn out to be young.

In fact, however, the opposite is true. Because the old typically have far less to look forward to, future-biased individuals prefer to be young. To get a sense of how this preference might look to someone behind the veil, imagine waking up temporarily unable to recall whether you are forty or sixty years old, or anything else specific about your life. Virtually everyone would hope to be forty. This judgment would be all the more pronounced for a person behind the veil who might be approaching the end of their life or in the prime of their youth with many decades to come.

Far from viewing being badly off while young as the predicament to hedge against, then, risk-averse future-biased parties would take themselves to have all the more reason to favor principles which confer additional benefits on people in old age, at the cost of even greater sacrifices to the young. Doing so mitigates the risk they face, by shrinking the gap between the bad and good outcomes.

We can also see why risk-aversion only makes things worse for the Veil of Ignorance Thesis by situating it within our model. From the perspective of a future-biased individual behind the veil, being Old is the worse outcome: whatever resources the Old receive, the Young receive those too, plus whatever they receive while young. Since the gap between the better and worse outcomes is driven by what the Young receive, distributions favoring the Old will shrink it directly. Thus, whenever a risk-neutral party would be willing to place a given burden on the Young in order to secure a given benefit for the Old, a risk-averse party would be willing to place an even greater burden on the Young to secure the very same benefit for the Old.

It remains to examine a formulation of the veil in which the parties face not just risk but uncertainty: they lack any objective grounds for assigning probabilities to being one age or another. As we noted earlier, when someone lacks knowledge about the likelihood of competing possible outcomes, one strategy is simply to assign an equal probability to every outcome. We have seen that future-biased individuals who proceed this way, whether risk-neutral or risk-averse, will select distributions which are clearly unjust when viewed from outside the veil. In the absence of any probabilities, the remaining decision procedures all involve some degree of uncertainty-aversion, according to which one prefers prospects under which the value one stands to receive is less dependent on which of the possibilities obtains.²⁶ The most extreme version is the maximin strategy, on which this concern is pushed to its limit: one favors the option whose worst possible outcome one regards as better than the worst possible outcome of any alternative option, whatever happens above it. Applying a maximin rule would seem to give absolute priority to benefiting the oldest members of society, since future-biased parties regard an outcome as worse the less life the individual has left. And giving the parties weaker degrees of uncertainty-aversion will not help either, for the same reason that giving them risk-aversion did not. Because they want their benefits in the future, the parties will regard as worse those outcomes in which they find themselves nearer to the end of life, and so they will be even more willing than they would be otherwise to concentrate small benefits on the old at greater expense to the young, thereby reducing how much turns on which possibility obtains.

III.C THE POSSIBILITY OF PREMATURE DEATH

The arguments I have given so far may seem unduly idealized in the following way: they assume that the parties who turn out to be young will all live long enough to actually enjoy whatever benefits get allocated to the old. In the real world, many people are unfortunate enough to die before ever reaching old age. Perhaps this fact of life is something the parties should be taking into account. In fact, however, building in the possibility of premature death does not stop the parties from choosing manifestly unjust policies like the ones we have already seen.

If the parties are to account for the realistic possibility of premature mortality, we must ask: realistic for which societies and at which times? Since advocates of the veil of ignorance think it can be used to evaluate the institutions of contemporary affluent democracies, it seems reasonable to look there first. Mortality statistics are reported for specific ages, so, in order to model how premature mortality affects our results, the “Young” and “Old” categories need to be operationalized as ages. Let us use age 25 for the Young and age 70 for the Old. In the United Kingdom, approximately 85% of 25-year-olds live until at least 70.²⁷ Conditional on being young, then, the parties will now assign themselves an 85% chance of actually receiving the benefits that accrue to the Old. By contrast, whatever benefits accrue to people when young, the parties can treat as assuredly theirs.

As before, we will begin by considering parties who are risk-neutral and equally likely to be in either age group. In Section III.A, when I first explained why risk-neutral, future-biased parties would prefer manifestly unjust arrangements over their just alternatives, we saw the following result: parties with absolute future bias will unanimously endorse a policy which gives people 81 units of wellbeing in total, allocating 10 to them while young and 71 to them while old, when they could instead have had a policy that gives people a full 100 units divided evenly across their ages (see Table 4). But even when the Young have only an 85% chance of ever seeing the benefits that accrue to the Old, this barely affects the result. Indeed, we only need to increase what the Young receive under the old-favoring distribution from 10 to 12 units for the parties to prefer it. Table 6 gives the calculations.

Table 6: Expected values with an 85% survival probability.

	Young	Old	Expected value
D1	50	50	$50 + (0 .85)50 + 50 2$ = 71.3
D5	12	71	$12 + (0 .85)71 + 71 2$ = 71.7

Viewed from outside the veil, the decision to adopt D5 over D1 would seem clearly unjust. Doing so takes 38 units of wellbeing from the young in exchange for only 21 units of wellbeing for the old. As before, making the parties’ future bias less extreme tempers the result but only barely.²⁸ If this is how rational self-interested parties behind the veil would choose, the Veil of Ignorance Thesis is in serious doubt.

One might wonder what follows with lower survival rates. While the amount of wellbeing that the parties are willing to sacrifice in order to shift benefits toward the old decreases as their survival probability diminishes, it does so slowly. Even at a survival rate as low as 70%, we only need to boost the wellbeing of the Young in D5 to 15 units for it to be preferred to D1.²⁹

In fact, it can be shown, as a general result, that risk-neutral parties will sacrifice total wellbeing in order to shift benefits toward the old whenever the survival probability exceeds the weight they place on their past wellbeing.³⁰ This means that for absolutely future-biased parties, at any survival probability above zero they would still accept some reduction in total wellbeing to shift benefits toward the old. (Parties who discounted the past by 90% would accept a sacrifice at any survival probability above 10%.)

Thus, the argument from future bias goes through, even with the inclusion of premature mortality, at least when the parties are assumed to be risk-neutral and equally likely to be Old or Young. How does the possibility of premature mortality influence parties who give additional priority to avoiding worse outcomes, either because they are averse to risk, or because they do not know what probabilities they face and are averse to uncertainty? As noted in Section III.B, while it may at first seem like such parties would resist overburdening the young lest they turn out to be so, in fact they will regard their prospect as worse the nearer to death they are. However, incorporating the possibility of premature death means it no longer follows from this fact that they would regard turning out to be old as the worst outcome. Indeed, as long as they place any value at all on having past goods, they will generally regard the very worst state of affairs to be dying when young.

That means that a strict maximin rule would lead the parties to give absolute priority to the moribund young. The idea that justice demands policies that give such extreme priority is implausible, however. No one thinks that the just society would direct its available resources toward very minor improvements to the quality of life of young people who will die soon anyway, at any expense to anybody else, no matter how old or healthy they are.

Strict maximin does not require the parties to know any probabilities, since it attends only to the single worst outcome. But for any rule that weights worse outcomes more heavily without giving absolute priority to the very worst outcome, one must look at multiple outcomes, and to weigh them one must have some idea about which are bad and how bad. For that reason, for any rule short of maximin, the argument as it applies to risk- or uncertainty-averse parties will go through as it did before. Granting that dying young would still be the very worst outcome even to a future-biased individual, the vast majority of their very bad outcomes will still cluster in old age, since it is overwhelmingly the elderly who find themselves nearer to death. So risk- and uncertainty-averse individuals behind the veil of ignorance would still be motivated to hedge against turning out to be old, giving them all the more reason to place large burdens on those early in life for the sake of smaller benefits to those later in it.

It would seem, then, that incorporating the realistic possibility of premature death cannot save the Veil of Ignorance Thesis from the threat posed by the rationality of future bias. Perhaps there are other real-world facts that, if the parties knew them, could discourage them from choosing distributions that are manifestly unjust outside the veil. However, even if there are such facts, arguably they cannot save the Veil of Ignorance Thesis anyway. The veil of ignorance may be best understood as a construction at the level of moral theory, as a general view about what makes social arrangements justifiable. Unlike mortality rates, moral theory is not contingent. So a rescue of the thesis that succeeded only because of high premature mortality, or some other highly contingent demographic fact, might not be a rescue at all.

We can summarize the overall findings of Section III as follows. Even when future-biased parties do not know their age, they will still favor distributions that impose heavy burdens on the young in order to benefit the old in ways that do not even approach compensating them for that sacrifice. This result persists across the range of plausible assumptions about their attitudes toward risk and uncertainty, about what they can predict regarding their age when the veil is lifted, and even about their likelihood of reaching old age at all. When we evaluate these decisions from outside of the veil, they are plainly unjust. So, provided that people can rationally care considerably more about benefits and burdens in the future than in the past, we have reason to doubt that the veil of ignorance can be used to identify just social arrangements.

IV. Further Considerations

IV.A WHICH PERSPECTIVE SHOULD THE PARTIES ADOPT FROM BEHIND THE VEIL?

Throughout, I have been assuming that the parties behind the veil of ignorance are tasked with advancing their self-interest once the veil is lifted and their knowledge is restored. In assuming this, I have taken for granted that each of them has a determinate age, even if they don’t know what it is, and is concerned to advance their interests conditional on that age turning out to be one thing or another. A defender of the Veil of Ignorance Thesis might try to respond that I have misconstrued the standpoint that the parties should take toward their lives when reasoning about which arrangements would best advance their self-interest.

One alternative construal is that when the parties are deliberating about how to allocate benefits and burdens throughout their lives, each should be aiming to promote the value of their life as a whole, considered atemporally. On this picture, the parties are not thought of as comparing how different policies would advance their interests conditional on having some possible past behind them and some possible future ahead of them, but just conditional on turning out to have one whole life or another. From the atemporal perspective, events in our lives are earlier or later than others, but none is in the past or the future. Accordingly, even if rationality directs us to be biased toward the future as we lead our lives from one moment to the next, this fact would be irrelevant to the Veil of Ignorance Thesis, since the latter draws on the preferences of individuals who, unlike us, are only concerned with the value of their lives as a whole.

Taking whole lives to be the basic unit of the parties’ prudential concern may not avoid counterintuitive results. Consider a person who is right now in the midst of horrible suffering and the deepest despair. Many of us would believe that our society is at least permitted to use its resources to help alleviate their suffering, instead of using those resources to benefit someone else who is currently happy and content—even if the happy person had a harder childhood such that the overall quality of their entire life, even with those additional resources, would still be lower than the suffering person’s. Yet parties behind the veil who were tasked with promoting the value of their whole lives would reject policies which allowed this sort of trade-off. Hence, even if the whole-life proposal could avoid the threat posed by future bias, doing so casts doubt on the Veil of Ignorance Thesis from a different direction.

Even if adopting a whole-lives perspective could avoid any counterintuitive results, however, it is hard to square that perspective with what makes the veil appealing in the first place. If it is rational to be future-biased, the life one ought rationally to want for oneself often bears little resemblance to one’s “best life” considered as a whole. The whole life that includes a painful operation in 2025 is in that respect more desirable to me now, in 2026, than it was in 2024. We never experience ourselves as whole lives but always have some particular past behind us and some particular future ahead of us. So unless the parties are very different creatures from us, the whole-life proposal amounts to placing an external normative constraint on how they are allowed to reason, forcing them to do so in a way that would be irrational for us to reason if we were deprived of the same information that they are.

Indeed, while philosophers do talk of the prudential value of whole lives, I myself find the idea somewhat obscure. I can certainly understand the idea of one whole life being better than another axiologically, contributing more impersonal value to the world. But, perhaps because I only ever experience life from a particular time within it, it is difficult to grasp what is supposed to be the subject that is benefited by my having a better life, if not myself at a time. To be clear, this skepticism has nothing whatever to do with experientialism about wellbeing, nor with any other view which ties a person’s momentary wellbeing to what is happening to them at that same moment. I can accept that it is not only better, but better for me, to live a life dedicated to fighting injustice instead of a more complacent but much happier life. And I can certainly see how it could be better for me to live a more successful or accomplished life, even if those accomplishments are tied to events after my death. What I worry is that the only meaningful sense in which it is better for me to live a more virtuous or accomplished life is the sense in which it is better for me, at determinate times within it, to have such a life.³¹

A second alternative construal of the parties’ standpoint behind the veil agrees that they should be conceived as trying to advance their interests conditional on having some particular age, but takes the age to be that of a person just born into their society.³² If that is how we should construe the Veil of Ignorance Thesis, then the challenge from future bias might seem easily answered. At birth, all the benefits and burdens in a person’s life are in their future. Assuming that near bias is irrational, it would seem irrational for even a future-biased agent to sacrifice more wellbeing early in life for less wellbeing later. For somebody just born, this only means that they have less wellbeing to look forward to.

One problem with this approach is that if we include the empirical reality of premature mortality, parties who take the perspective of newborns would seemingly give implausibly great weight to preventing very early deaths, and moreover be tempted to massively front-load benefits where premature death cannot be prevented. A more abstract problem is just that, as far as I can see, this approach artificially privileges the beginning of life as the correct prudential standpoint to take from behind the veil. Why should the parties be reasoning about the life they would want for themselves if just born, rather than what life they would want to have had if they were at the end of their lives and looking back? And for that matter, why should they not be tasked with reasoning about what kind of life they would want if they found themselves squarely in its middle? If future bias is rational, whether a given pattern of benefits and burdens across a person’s life is in their interest changes depending on what stage in that life we are asking the question about. I see no principled reason why the parties should concern themselves exclusively with the interests they have at birth, rather than with the balance of interests they have at other times in their lives.

Lastly, leaving aside any worry about its arbitrariness, the proposal that the parties must take a newborn’s perspective has counterintuitive implications once we notice that many policy decisions shape the very interests that people will come to have as they grow up. To start with some more quotidian cases, if there were no skateparks or football stadiums, there would be few skateboarders or football fans. So from the perspective of someone just born into society, who has no attachment to those activities over others, there is no reason to want society to support them, yet the parties might well turn out to value them when the veil is lifted. Arguably, the proposal to take a newborn’s perspective fails even to show that banning certain religions would be unjust—and that, if anything, seems like a result that proponents of the veil of ignorance want it to yield. For, if a particular religion is banned before we are born, we may not care much, since we won’t be raised into that religion or even exposed to it. On the competing proposal that the parties should reason as if they might be any age in their society, it is easy to see why the parties would not want to ban a particular religion. For the religion might already be theirs. I doubt, then, that a defender of the Veil of Ignorance Thesis can convincingly respond to the challenge from future bias by maintaining that the parties behind the veil should take the perspective of someone who was just born.

IV.B CAN OTHER-REGARDING INTERESTS RESCUE THE VEIL?

Throughout our discussion, we have focused on how future-biased parties will evaluate policies based on the benefits and burdens that they themselves might have when the veil is lifted. In doing so, we have been assuming that their evaluation of those policies does not also depend on the benefits and burdens that the policies would be expected to give to others outside the veil. This may give rise to the following objection to my argument. In reality, people care a great deal about others’ wellbeing in addition to their own. Isn’t it natural, therefore, for the parties to evaluate a policy in part based on how it is expected to affect people besides themselves outside the veil, since those people may be ones they turn out to care about? Importantly, this objection does not say that the parties should be concerned with each other behind the veil—in Rawls’s phrase they can remain “mutually disinterested,” with each concerned only to advance their own interests when the veil is lifted. Rather, it says something about the scope of those interests themselves. Even if the parties should be concerned only with their own interests, perhaps they should not be exclusively what we might label self-regarding: concerned only to promote their personal welfare, to the exclusion of anyone else’s.³³ Factoring in an anticipated concern for others might seem to push the parties away from distributions that heavily burden the young for the sake of small benefits to the old, by providing a counterweight to their self-regarding future bias.

One worry with this approach is that it would lead the parties to give weight to policies that favor the already well-loved over the friendless or orphaned, since the parties would know that whomever they end up caring about will be someone who is cared about (and presumably the more cared about a person is, the more likely they are to be someone the parties will care about). We can leave this aside, however, because taking their anticipated concern for others into account does not obviously provide a counterweight to self-regarding future bias, anyway. For one thing, we are not neutral between the past and future of the people we care about, either. (That is well confirmed by empirical literature,³⁴ and arguably required by consistency given the rationality of future bias in the first-personal case.³⁵) By way of illustration, imagine that you visit your spouse in the hospital only to find them already asleep in their bed; you do not know whether their excruciatingly painful operation already happened yesterday or is still to come tomorrow.³⁶ Most people would hope it was over and done with, and they would hope so out of concern for their spouse. In the same way, parties behind the veil could reason that, whomever else they do turn out to care about, they want those people’s burdens to be in the past and benefits to be in the future. Whether their anticipated concern for others will lead them to endorse policies that place heavy burdens on the young in exchange for small benefits to the old thus depends on what they can predict about the ages of the people they anticipate caring about. Insofar as they predict having concern for children or grandchildren who are young, they will be inclined away from such policies, since most of a young person’s wellbeing is expected to be in the future where they do not discount it behind the veil. Insofar as they care about parents, spouses and friends in maturity, or adult children, however, the parties will be pulled back toward burdening the young in order to shift benefits into old age, just as self-regarding parties are pulled that way in light of the fact that they themselves may turn out older.³⁷

It is difficult to say how the veil of ignorance should incorporate these competing pulls, which depend on what the parties can predict about the ages of the people they care about. But I see no obvious argument for presupposing that, on balance, their concern for others would run counter to the self-regarding future bias that leads them to favor the old over the young.³⁸ In any case, even if there were such an argument—for example, one showing that the parties could predict that outside the veil they will be mainly concerned for children—it would still not be clear that taking the parties’ other-regarding interests into account can rescue the Veil of Ignorance Thesis from the challenge posed by future bias. For while most people care greatly about the welfare of others, they also place rather significant weight on their own welfare. This means that even if the other-regarding portion of the parties’ concern pushed entirely toward more equal and efficient distributions, the self-regarding portion of it would still pull toward the unjustly old-favoring distributions we saw in Section III. Recall that the distributions endorsed by self-regarding, future-biased parties are not only unjust, but massively so. They involve dramatically reducing the total wellbeing of every person in society and distributing the remainder very unequally across ages. Thus, even if the burdens on the young were lessened until parties who have both self- and other-regarding interests would prefer them without question, it is likely that the resultant distributions would still be pretty obviously unjust when considered from outside of the veil.

IV.C THE UNIMPORTANCE OF (HYPOTHETICAL) CHOICE

We have been conceiving the individuals behind the veil as comparing policies that have different implications about when in people’s lives their benefits and burdens are located. Depending on what their actual age is, many of those benefits and burdens might already be in their past. Since we cannot change the past, however, my arguments might seem incompatible with the common description of the veil of ignorance as generating a “hypothetical choice situation.” After all, the individuals behind the veil either do or do not have a certain amount of wellbeing behind them, even if they don’t know what it is. There is no genuine possibility of reducing it now in the hopes of securing a more desirable future. So it might be said that my arguments fail on the grounds that they depend on the parties’ undertaking an impossible task.

This attempt to save the Veil of Ignorance Thesis is unconvincing for a few reasons. First, there is already independent reason to think that the veil of ignorance scenario should ultimately be about hypothetical endorsement, acceptance, preference, or what you would choose if you could, rather than about hypothetical choice, literally understood. The insistence on choice simply rules out too many policies that plainly come under the purview of justice. Surely a policy like “Give every citizen a college savings fund at birth” is susceptible to justification from behind the veil if anything is. But evaluating this policy from the perspective of rational self-interest requires deliberating about something that either is or is not already in place at the time of deliberation.

More generally, though, it is simply unclear what additional justification is conferred on a policy if it would have been hypothetically chosen, over and above the justification conferred on it by hypothetical endorsement, for example. Certainly, the fact that a person actually chose an outcome seems able to matter beyond whether they merely wanted it to occur (sexual consent may be an important example).³⁹ One’s having actually chosen something seems able to alter the normative situation in a way that merely desiring it or privately endorsing it does not. But the fact that a person would have chosen something, had her knowledge or circumstances been otherwise, does not seem to have additional special normative significance in the same fashion. Pointing out that someone would have voted for some policy if given the chance under fairer circumstances might help justify subjecting her to it now, but not in a way that surpasses the justification that would be conferred by the fact that under those circumstances she would have endorsed, preferred, or accepted it.

Lastly, insisting that the veil of ignorance must involve literal choosing just seems contrived. Consider that in theory, we could at any point in time take any individual in society, place them behind a veil of ignorance which deprived them of knowledge of their circumstances including their age, and ask them whether they endorse their society’s rule of placing heavy burdens on people when they are young to provide benefits when they are older that fall far short of compensating them. If my arguments are sound, then every member of society would rationally reply that they do, indeed, endorse that policy over an alternative one that gave everyone much more wellbeing spread evenly throughout their lives. Anyone moved by the veil of ignorance as a test for justification should, I believe, be equally moved by the idea that a policy is justified when and because it would pass a test like that one.

IV.D WHY RATIONAL PERMISSIBILITY IS ALREADY ENOUGH

In making my arguments, I have asked what emerges from the veil of ignorance if we take for granted that the parties behind it are biased toward the future. Since most of us probably believe we have good reason to care more about the future than the past, the conclusion that future-biased parties would choose unjust policies may already be enough to undermine the Veil of Ignorance Thesis. In fact, however, that common-sense stance toward future bias is a much more demanding premise than the argument requires.

Even if we have no more prudential reason to care about what will happen to us in the future than the past, having future-biased preferences seems at least rationally permissible. Perhaps, as McTaggart was prepared to concede, our preference for benefits to be in the future instead of the past is no more a requirement of rationality than a preference of “burgundy to claret, or of claret to burgundy.”⁴⁰ The Veil of Ignorance Thesis is undermined by my arguments even if future bias is simply rationally optional, for the following reason. As a completely general observation, it must sometimes be possible for rational parties behind the veil to find themselves facing a choice between different ways of structuring their society, neither of which is favored over the other by rational self-interest. When that happens, the Veil of Ignorance Thesis would likewise have to be neutral between them, by implying that neither policy is more just than the other. Thus, if future bias is rationally optional, the Veil of Ignorance Thesis lacks the resources to evaluate the policies that future-biased parties would endorse differently to the ones that non-future-biased parties would endorse. As we have seen, future-biased, otherwise rational parties would endorse a policy which yields far less wellbeing, distributed unequally across age groups, over one which yields far more wellbeing and distributes it evenly. Since these policies are plainly not equivalently just, the Veil of Ignorance Thesis must be rejected as long as the bias toward the future is rationally permissible.

In addition to stipulating that the parties are future-biased, I have assumed throughout that the parties are not biased toward the near over the further future. Disallowing near bias behind the veil may not be very controversial, since it is already widely believed to be irrational.⁴¹ Just as with future bias, however, a much weaker premise will do. As long as it is not rationally mandatory to prefer smaller goods in the near future to larger ones in the further future (holding fixed their certainty), the veil of ignorance lacks the resources to morally differentiate the policies that would be chosen by parties who lack near bias from those that would be chosen by parties who have it. We have seen that future-biased parties who are not near-biased will select manifestly unjust distributions over just ones. As before, however, if the Veil of Ignorance Thesis grants equal justification to policies which are nowhere near being equally justifiable outside the veil, that alone is sufficient reason to reject it.

In sum, the Veil of Ignorance Thesis would already be in serious doubt if it were incompatible with the claims that future bias is rationally required and near bias rationally impermissible. But the thesis fails so long as future bias is permissible and near bias is not required. Neither of those premises is beyond dispute, of course, but rejecting them would strain, if not downright invert, our ordinary understanding of prudential rationality.

V. Conclusion

There is a lot of attraction to the view that social justice mirrors the preferences of rational, self-interested individuals behind a veil of ignorance. I have examined how the Veil of Ignorance Thesis fares in light of the widespread prudential bias toward the future. Future-biased individuals behind the veil would favor distributive policies which impose substantial burdens on people when they are young in exchange for much smaller benefits to them in old age, over policies which differ only in that they afford significantly more total wellbeing to everyone and spread it evenly throughout their lives. Such policy decisions are clearly unjust when considered outside of the veil.

Does this mean that we should reject the Veil of Ignorance Thesis? Only if future bias is not irrational, and while that premise is very plausible, it is certainly not without its challengers. Since the Veil of Ignorance Thesis is independently compelling, opponents of future bias could always take my arguments as evidence that it is irrational.

Another option is to quarantine the bad result, accepting that the Veil of Ignorance Thesis fails, but only as applied to issues involving the distribution of resources across different ages. That is hardly a minor concession, since these are among the most important areas of social policy, including healthcare, social security, and education funding. But it is also unclear what the principled basis for a quarantine would be. What independent feature of age-based policies makes them relevantly different from other distributive ones, such that it makes sense to apply the veil of ignorance to the one but not the other? Lacking an answer, if the veil gets age-related policies wrong, we need to ask what grounds we have to think it is getting other policies right.

Finally, one might take the upshot to be that the veil simply needs to be reconfigured so that it doesn’t yield implausible results. Even this move is not as easy as it may seem. Other objectionable forms of partiality, like those toward a gender or race, can be neutralized by depriving the parties of more information about themselves. But partiality toward the old cannot be averted in that way, because endorsing it is rational even when you do not know how old you are. Reformulating the veil to avoid counterintuitive results can therefore only be done by intervening in how the parties are allowed to reason (and as Section IV.A showed, even these interventions carry intuitive costs of their own). More importantly, if we have to work backwards from its outputs to formulate the veil, then it does not justify anything we were not already committed to. Some authors are content to see the veil as an epiphenomenon like that. My target has been the idea that when people would have chosen some policy in ignorance of who they are, this helps justify that policy. That is the conception of the veil that would warrant its pride of place in moral and political theory. So if the only way to save the veil of ignorance is to abandon that conception of it, I will be satisfied to have shown that much.

Notes

For example, Ronald Dworkin, “The Original Position,” The University of Chicago Law Review 40, no. 3 (1973): 500–533, https://doi.org/10.2307/1599246. ⮭
John Rawls, A Theory of Justice, rev. ed. (Belknap Press of Harvard University Press, 1999), https://doi.org/10.4159/9780674042582; Thomas Nagel, “Rawls on Justice,” The Philosophical Review 82, no. 2 (1973): 220–234, https://doi.org/10.2307/2183770; John Harsanyi, “Cardinal Utility in Welfare Economics and in the Theory of Risk-Taking,” Journal of Political Economy 61, no. 5 (1953): 434–435, https://doi.org/10.1086/257416; John Harsanyi, “Can the Maximin Principle Serve as a Basis for Morality? A Critique of John Rawls’s Theory,” The American Political Science Review 69, no. 2 (1975): 594–606, https://doi.org/10.2307/1959090; Derek Parfit, On What Matters, vol. 1 (Oxford University Press, 2011), 350–351, https://doi.org/10.1093/acprof:osobl/9780199572809.001.0001. Even those who agree (if only for the sake of argument) on what the parties should and should not be prevented from knowing might still disagree over what they would rationally settle on given their knowledge. Compare, for instance, Rawls, Theory of Justice; John Tomasi, Free Market Fairness (Princeton University Press, 2012), https://doi.org/10.23943/princeton/9780691144467.001.0001; and Johan E. Gustafsson, “The Difference Principle Would Not Be Chosen behind the Veil of Ignorance,” Journal of Philosophy 115, no. 11 (2018): 588–604, https://doi.org/10.5840/jphil20181151134. ⮭
For example, John Rawls, “Some Reasons for the Maximin Criterion,” The American Economic Review 64, no. 2 (1974): 141–146, https://www.jstor.org/stable/pdf/1816033; Rawls, Theory of Justice; David Lyons, “Rawls versus Utilitarianism,” Journal of Philosophy 69, no. 18 (1972): 535–545, https://doi.org/10.2307/2025370; Harsanyi, “Can the Maximin Principle Serve as a Basis for Morality?”; Lara Buchak, “Taking Risks behind the Veil of Ignorance,” Ethics 127, no. 3 (2017): 610–644, https://doi.org/10.1086/690070; Jacob Nebel, “Rank-Weighted Utilitarianism and the Veil of Ignorance,” Ethics 131, no. 1 (2020): 87–106, https://doi.org/10.1086/709140. ⮭
Terminology which derives from Derek Parfit, Reasons and Persons (Clarendon Press, 1984), 160, https://doi.org/10.1093/019824908X.001.0001. Despite the label “bias,” “future bias” is a descriptive term. “Bias” does not suggest anything about whether the phenomenon is irrational or in any other way defective; indeed, many who use the term think future bias is rationally permissible or required. So even though the veil is meant to model impartial reasoning, in which the parties cannot rely on their personal “biases” while deliberating, future bias is not the kind of thing that can be pre-emptively excluded by the veil. ⮭
Eugene Caruso, Daniel Gilbert, and Timothy Wilson, “A Wrinkle in Time: Asymmetric Valuation of Past and Future Events,” Psychological Science 19, no. 8 (2008): 796–801, https://doi.org/10.1111/j.1467-9280.2008.02159.x; Ruth Lee et al., “Pain in the Past and Pleasure in the Future: The Development of Past–Future Preferences for Hedonic Goods,” Cognitive Science 44, no. 9 (2020): e12887, https://doi.org/10.1111/cogs.12887; Preston Greene et al., “Hedonic and Non-Hedonic Bias toward the Future,” Australasian Journal of Philosophy 99, no. 1 (2021): 148–163, https://doi.org/10.1080/00048402.2019.1703017; Preston Greene et al., “On Preferring that Overall, Things Are Worse: Future-Bias and Unequal Payoffs,” Philosophy and Phenomenological Research 105, no. 1 (2022): 181–194, https://doi.org/10.1111/phpr.12819; Preston Greene et al., “How Much Do We Discount Past Pleasures?,” American Philosophical Quarterly 59, no. 4 (2022): 367–376, https://doi.org/10.5406/21521123.59.4.03. ⮭
Caspar Hare, “Time—The Emotional Asymmetry,” in A Companion to the Philosophy of Time, ed. Adrian Bardon and Heather Dyke (Wiley-Blackwell, 2013), 519, https://doi.org/10.1002/9781118522097.ch29. ⮭
See Parfit, Reasons and Persons, 124. ⮭
For some defenses of future bias, see Chris Heathwood, “Fitting Attitudes and Welfare,” in Oxford Studies in Metaethics, vol. 3, ed. Russ Shafer-Landau (Oxford University Press, 2008), 47–73, https://doi.org/10.1093/oso/9780199542062.003.0003; Hare, “Time—The Emotional Asymmetry”; Dale Dorsey, “Future-Bias: A (Qualified) Defense,” Pacific Philosophical Quarterly 98, no. 1 (2017): 351–373, https://doi.org/10.1111/papq.12176; Christian Tarsney, “Thank Goodness That’s Newcomb: The Practical Relevance of the Temporal Value Asymmetry,” Analysis 77, no. 4 (2017): 750–759, https://doi.org/10.1093/analys/anx108; Sayid Bnefsi, “B-Theory and Time Biases,” in Logic and Philosophy of Time: Further Themes from Prior, ed. Patrick Blackburn, Per Hasle, and Peter Øhrstrøm (Aalborg University Press, 2019), 49–50, https://openresearchlibrary.org/content/092cf308-9e54-4906-8d4a-439e7c3e03e5; Sayid Bnefsi, “Future Bias and Regret,” in Logic and Philosophy of Time: The History and Philosophy of Tense-Logic, ed. David Jakobsen, Peter Øhrstrøm, and Per Hasle (Aalborg University Press, 2023), 1–13, https://doi.org/10.54337/lpt.v5i1.7857; Samuel Scheffler, “Temporal Neutrality and the Bias toward the Future,” in Principles and Persons: The Legacy of Derek Parfit, ed. Jeff McMahan et al. (Oxford University Press, 2021), 85–114, https://doi.org/10.1093/oso/9780192893994.003.0005; Todd Karhu, “What Justifies Our Bias toward the Future?,” Australasian Journal of Philosophy 101, no. 4 (2023): 876–889, https://doi.org/10.1080/00048402.2022.2047747. For challenges to future bias, see David Brink, “Prospects for Temporal Neutrality,” in The Oxford Handbook of Philosophy of Time, ed. Craig Callender (Oxford University Press, 2011), 353–381, https://doi.org/10.1093/oxfordhb/9780199298204.003.0012; Tom Dougherty, “On Whether to Prefer Pain to Pass,” Ethics 121, no. 3 (2011): 521–537, https://doi.org/10.1086/658896; Preston Greene and Meghan Sullivan, “Against Time Bias,” Ethics 125, no. 4 (2015): 947–970, https://doi.org/10.1086/680910. ⮭
This reference to both policies and principles points to one further question about the relationship between my project and Rawls’s. I will usually frame the veil of ignorance as a procedure for justifying policies, social arrangements, or patterns of distribution, whereas the parties in Rawls’s Original Position are tasked with selecting coarse-grained principles of justice. I do not think the Veil of Ignorance Thesis can escape my arguments by insisting that the parties’ task is to select principles, however. The arguments I will give for why future-biased parties would endorse policy decisions that unjustly favor the old also establish that such parties would arrive at implausible judgments at the level of general principles, since they would favor the principles that licensed those decisions. (For example, they might well endorse a principle like “Inequalities across age groups are permissible when they improve the prospects of citizens in old age.”) Moreover, my arguments can be understood as showing why future-biased parties would be led to define the terms in their principles in ways that lead to the counterintuitive results, for instance by defining the worst-off as those with the least future wellbeing. ⮭
Note that sometimes we might want to have had a benefit in the past because of its beneficial effects in the present or future, as a happy childhood or a good education might yield. Insofar as we want a past benefit for that reason, we are future-biased with respect to the benefit. ⮭
For arguments to that effect, among other reasons because old and young would not relate to one another as equals, see Juliana Uhuru Bidadanure, Justice Across Ages: Treating Young and Old as Equals (Oxford University Press, 2021), https://doi.org/10.1093/oso/9780198792185.001.0001. ⮭
See Norman Daniels, Am I My Parents’ Keeper? An Essay on Justice Between the Young and the Old (Oxford University Press, 1988). ⮭
For empirical studies to this effect, see Andrew Latham et al., “Future Bias in Action: Does the Past Matter More When You Can Affect It?,” Synthese 198, no. 12 (2021): 11327–11349, https://doi.org/10.1007/s11229-020-02791-0; Greene et al., “On Preferring that Overall, Things Are Worse”; Greene et al., “How Much Do We Discount Past Pleasures?” ⮭
The term first appears in Greene and Sullivan, “Against Time Bias,” 961. ⮭
Parfit, Reasons and Persons, 173; Heathwood, “Fitting Attitudes and Welfare.” Heathwood succinctly articulates what is probably the intuition behind absolute future bias: of someone whose pain has passed, he writes, “[w]hy should he care about it now? No reason—it’s over and done with” (57). ⮭
Parfit, Reasons and Persons, 165–166. ⮭
See Greene et al., “On Preferring that Overall, Things Are Worse.” ⮭
The preceding claim that the choice of D3 over D1 is unjust may invite two questions. First, since within either distribution everyone fares equally well over a complete life, one might ask whether injustice is really the right charge to level against the selection of D3. The disagreement here may be partly terminological, but in any case, in line with prominent proponents of the veil of ignorance, I will take distributive justice to involve more than interpersonal equality. (A society that gave every member 1 unit of wellbeing when it could instead have given some people 2 and the rest 3 would be choosing unjustly, despite plumping for equality.) Second, some may prefer to cast the objection to D3 in terms of unacceptability to the young. That reading is congenial, since an arrangement’s being reasonably rejectable is plausibly evidence of its injustice, and on some views it is part of what injustice consists in. ⮭
A third possibility is that the parties should apportion their credences about their age to the demographics of the population they are deliberating about. In societies that skew younger, this approach would somewhat attenuate the results to come; in societies that skew older, the parties would be all the more inclined to burden the young for the sake of the old. Since such skews would merely modulate the results of the equal-probability assumption in one direction or the other, I won’t consider them separately, though there is also a more principled reason not to do so. As I explain at the end of Section III.C, the Veil of Ignorance Thesis is not compelling if it delivers plausible verdicts only because of contingent demographic facts, such that it would cease to deliver them were those facts different. ⮭
Rawls, A Theory of Justice. ⮭
John Maynard Keynes gave the principle its name in his influential early critique of it: A Treatise on Probability (Macmillan, 1921), 41. For contemporary defenses of the principle of indifference, see Richard Pettigrew, “Accuracy, Risk, and the Principle of Indifference,” Philosophy and Phenomenological Research 92, no. 1 (2016): 35–59, https://doi.org/10.1111/phpr.12097; Jon Williamson, “Justifying the Principle of Indifference,” European Journal for the Philosophy of Science 8 (2018): 559–586, https://doi.org/10.1007/s13194-018-0201-0; and Benjamin Eva, “Principles of Indifference,” Journal of Philosophy 116, no. 7 (2019): 390–411, https://doi.org/10.5840/jphil2019116724. ⮭
Individuals would then enjoy 12 units when young and 71 units when old, yielding an expected value of $83 + 71 + (0 .1)12 2$ = 77.6. ⮭
The expected value of D1 with an 80% past-discount rate is $100 + 50 + (0 .2)50 2$ = 80, while the expected value of the revised D4—in which the young enjoy an average wellbeing of 16 units—would be $87 + 71 + (0 .2)16 2$ = 80.6. ⮭
True, Harsanyi and others have argued that the Veil of Ignorance Thesis, properly fleshed out, yields average utilitarianism—and many people think that the practical implications of utilitarianism show it to be unjust. However, utilitarianism is not manifestly unjust. At a minimum, the abstract notion that society should maximize wellbeing has a certain ring of truth. In contrast, there is no prima facie appeal, abstract or concrete, to a social arrangement which predictably reduces the total amount of wellbeing in society while simultaneously spreading that lower amount of wellbeing unevenly across age groups. ⮭
For similar reasons, those who hold that the “strains of commitment” forbid the parties from agreeing to principles they could not later willingly endorse will find no rescue here: that constraint rules out the intolerable, not the unjust. See Rawls, A Theory of Justice, 153–154; see also Jeremy Waldron, “The Strains of Commitment,” in The Original Position, ed. Timothy Hinton (Cambridge University Press, 2015), 64, https://doi.org/10.1017/CBO9781107375321.004. We can readily imagine societies that opt for deeply unjust, old-favoring distributions without thereby placing the young in positions that they cannot abide. ⮭
For further issues that arise given uncertainty-aversion behind the veil (also called aversion to “ambiguity,” to mark it off terminologically from risk-aversion, where probabilities can be assigned), see H. Orri Stefánsson, “Ambiguity Aversion behind the Veil of Ignorance,” Synthese 198, no. 7 (2021): 6159–6182, https://doi.org/10.1007/s11229-019-02455-8. ⮭
Based on the UK National Life Tables for 2022–2024: Office for National Statistics, “National Life Tables, United Kingdom, Period Expectation of Life, Based on Data for the Years 2022-2024,” released December 10, 2025, https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/lifeexpectancies/datasets/nationallifetablesunitedkingdomreferencetables. ⮭
With a 90% past-discount rate, parties who assumed that conditional on being young they would have an 85% chance of seeing the benefits given to the Old would still only require 15 total units of wellbeing when young to prefer the resulting distribution over the superior D1. The value these parties assign to the new distribution is $15 + (0 .85)71 + 71 + (0 .1)15 2$ = 73.9, whereas the value to them of D1 would be $50 + (0 .85)50 + 50 + (0 .1)50 2$ = 73.8. ⮭
With a 70% survival rate, the expected value of D1 becomes $50 + (0 .70)50 + 50 2$ = 67.5, while a distribution of 15:71 yields $15 + (0 .70)71 + 71 2$ = 67.9. ⮭
Let p be the survival probability and w be the weight the parties place on past wellbeing. From behind the veil, old-age benefits are weighted by 1 (if old) and p (if young), while benefits to the young are weighted by 1 (if young) and w (if old). The asymmetry favoring old age disappears only once p falls to w. ⮭
For a different set of arguments that prudential rationality is always time-relative, based on considerations about personal identity and epistemic rationality, see Brian Hedden, Reasons without Persons: Rationality, Identity, and Time (Oxford University Press, 2015), https://doi.org/10.1093/acprof:oso/9780198732594.001.0001. ⮭
Thus Nagel claimed that “the subject of an egalitarian principle is not the distribution of particular rewards to individuals at some time, but the prospective quality of their lives as a whole, from birth to death.” Thomas Nagel, Equality and Partiality (Oxford University Press, 1991), 69, https://doi.org/10.1093/0195098390.001.0001. Rawls describes the basic structure as one we “enter only by birth and exit only by death” but by this he means to stipulate that the society is closed. John Rawls, Justice as Fairness: A Restatement (Belknap Press of Harvard University Press, 2001), 182, https://doi.org/10.2307/j.ctv31xf5v0. ⮭
I use the term “self-regarding” because Rawls reserves “egoistic” to describe parties with particular kinds of interests, such as wealth, prestige, and domination. Rawls, Theory of Justice, 12. ⮭
Greene et al., “Hedonic and Non-Hedonic Bias toward the Future”; Greene et al., “On Preferring that Overall, Things Are Worse”; Greene et al., “How Much Do We Discount Past Pleasures?” ⮭
As Caspar Hare argues. See “A Puzzle about Other-Directed Time-Bias,” Australasian Journal of Philosophy 86, no. 2 (2008): 269–277, https://doi.org/10.1080/00048400801886348; “Time—The Emotional Asymmetry.” ⮭
Based on Hare, “A Puzzle about Other-Directed Time-Bias.” ⮭
It is relevant that we seem to care a good deal about the people we love not having had heavy burdens in their past. But this is unlikely to matter much for three reasons. First, our concern seems to be primarily that they not have undergone seriously negative experiences. But at least in relatively affluent societies, distributions which are heavily skewed toward the old do not have to involve horrors or deep deprivations for the young in order to be unjust. Second, our concern for other people’s pasts seems not to manifest itself in a way that competes with benefits in their future. Suppose you learn two things about an elderly person you love: you were wrong that they had a nasty childhood, and they will die half a year sooner than you thought. I doubt this would strike many people as good news on balance. Finally, our concern for other people’s pasts is still heavily outweighed by our concern for their futures, so the structure of the argument is unlikely to change much. ⮭
One such argument might start from the interesting observation that our other-regarding future bias seems weakest with respect to people who are unidentified, or whose condition is not salient to us. If that is right, then insofar as the parties’ concern for others is impartial rather than directed at particular people, it will be comparatively free of future bias, and might therefore be thought to tip the scales away from the unjust, age-skewed distributions. Two things can be said in reply. First, it is highly questionable whether this contrast between our attitudes toward the identified and the unidentified is rationally defensible. Hare, “A Puzzle about Other-Directed Time-Bias”; “Time—The Emotional Asymmetry.” Second, and perhaps more importantly, impartial concern of this kind is precisely what must be excluded from the parties’ motivation behind the veil of ignorance, on pain of building into the device the very moral commitments it is meant to deliver. See Rawls, A Theory of Justice, 513. ⮭
See Tom Dougherty, “Yes Means Yes: Consent as Communication,” Philosophy & Public Affairs 43, no. 3 (2015): 224–253, https://doi.org/10.1111/papa.12059. ⮭
J. M. E. McTaggart, The Nature of Existence, vol. 2, ed. C. D. Broad (Cambridge University Press, 1927), 350. ⮭
Dale Dorsey explores a Parfitian argument that something like near bias could be rationally required, on the grounds that we have stronger prudential reason to care about those who are psychologically more connected to us, which our near-future selves typically are. “A Near-Term Bias Reconsidered,” Philosophy and Phenomenological Research 99, no. 2 (2019): 461–477, https://doi.org/10.1111/phpr.12496. It is perhaps noteworthy that even Dorsey ultimately concludes that near bias is irrational. ⮭

Acknowledgments

For valuable comments and discussion, I am indebted to Matthew Adams, Nicolas Côté, Tomi Francis, Helen Frowe, Joe Horton, Nannerl O. Keohane, Kacper Kowalczyk, Henrik Kugelberg, Michal Masny, Kieran Oberman, Michael Otsuka, Aidan Penn, Massimo Renzo, Thomas Rowe, Bastian Steuwer, two Associate Editors of this journal, and audiences at King’s College London, the Stanford Center for Ethics in Society, the Early Career Ethics Network, and the Institute of Philosophy PPE series.

Competing Interests

The author has no competing interests to declare.

Temporal Partiality and the Veil of Ignorance

Abstract

Keywords:

I. Introduction

II. Setting the Stage: Parties in the Middle of Their Lives

III. Selecting Distributions in Ignorance of One’s Age

III.A RISK-NEUTRALITY

III.B RISK- AND UNCERTAINTY-AVERSION

III.C THE POSSIBILITY OF PREMATURE DEATH

IV. Further Considerations

IV.A WHICH PERSPECTIVE SHOULD THE PARTIES ADOPT FROM BEHIND THE VEIL?

IV.B CAN OTHER-REGARDING INTERESTS RESCUE THE VEIL?

IV.C THE UNIMPORTANCE OF (HYPOTHETICAL) CHOICE

IV.D WHY RATIONAL PERMISSIBILITY IS ALREADY ENOUGH

V. Conclusion

Notes

Acknowledgments

Competing Interests

Share

Author details

Downloads

Information

Citation

File Checksums

Table of Contents