Plan B

ABSTRACT

We sometimes strive to achieve difficult goals when our evidence suggests that success is unlikely—not just because it will require strength of will, but because we are targets of prejudice and discrimination or because success will require unusual ability. Optimism about one’s prospects can be useful for persevering in these cases. With that said, excessive optimism can be dangerous; when our evidence is unfavourable, we should be at most agnostic about whether we will succeed. This paper explores the nature and rational significance of agnostic practical commitments. Most importantly, the rationality of striving against the odds can depend on investing in a plan for failure—a plan B. I aim to make headway on an account of what backup plans are, how they are related to primary plans, and whether the standard norms of plan rationality apply to our agnostic commitments.

KEYWORDS:

• planning
• plan rationality
• commitment
• grit
• optimism
• agglomeration

1. Introduction

Consider Sonia, who is the child of Puerto Rican immigrants and lives in the Bronx. From an early age, she has had the goal of attending a top university, going to law school, becoming a successful attorney, and winning a federal judgeship. She organizes her teenage years around this goal, sacrificing hobbies and time with her family so that she can study. Despite her efforts, Sonia harbours significant doubt that she will achieve her ambitions. She is unsure that she has the ability needed to score highly on standardized tests and to earn the necessary grades for admission to a selective college. And even if she does matriculate, she notes that there are very few Latina women with positions at high-powered law firms or appointments to federal judgeships. This leads her to think that, for whatever reason, people like her generally do not succeed in this kind of ambition. Even as she gains admission to Princeton and then Yale Law, she continues to feel like an outsider who has succeeded primarily through hard work and luck rather than ability, and she doubts that her successes will be repeatable. While she remains devoted to her aspirations, when asked whether she will become a practicing attorney or federal judge one day, she sincerely replies ‘I don’t know.’ As it turns out, Sonia does ultimately achieve those goals and ends up with a seat on the Supreme Court, although she continues feel she does not belong there.¹

Sonia’s example is obviously not typical in some respects; few of us end up achieving anything comparable to sitting on the Supreme Court. Let us set aside that part of the story. In other respects, the case is a representative one. Many of us have long-term goals that will be difficult to achieve, and that will require us to exhibit ‘grit’: we will have to persevere in the face of obstacles and setbacks. It is not only that such goals take a long time to realize, but also that many who set out in pursuit of them eventually give up. Second, success in this kind of case is not dependent on grit alone. For one thing, agents like Sonia face a good deal of prejudice and discrimination, which can foreclose certain opportunities. For another, Sonia’s goals require an unusual level of ability, which many agents discover along the way that they lack, no matter how hard they work. Thus, Sonia has good reason to doubt that she will triumph, even though she is in fact highly talented.

Let us make two further stipulations to narrow down our topic. First, the goals that Sonia is pursuing are not morally obligatory or even of especially high moral priority. One can certainly do morally good things in the position of attorney and judge, but there are many other pursuits in life that offer the opportunity to do something that is morally good. Sonia therefore has no special moral reason to pursue these goals. Second, although she is not starting from a position of copious resources, we can imagine that she has other options that she considers to be valuable and that are significantly more feasible. To pursue the legal profession, perhaps she must forego the chance to be a young stay-at-home mother or to join the family business. Thus, it is not as though she aspires to this career because she takes herself to have no other good choices.

Although examples like Sonia’s are hardly unusual (again, setting aside the highly unusual ending), they raise difficult questions for our thinking about agency and rationality, both practical and epistemic. There is pressure to see her as making a rational mistake, either in her practical commitments or in her beliefs. On one hand, if Sonia’s pursuit of an ambitious legal career is rationally permissible, then we are tempted to think that she should be confident of success. Her problem, it is now fashionable to say, is that she has ‘impostor syndrome’ and should find a way to ‘believe in herself’. On the other hand, if she is correct to have significant doubts about her prospects, we are tempted to think that she ought to have chosen a different path that she has better reason to expect will succeed. She has other good options, and this choice opens her up to material risks and opportunity costs that might well be unaffordable for her. And a third kind of scepticism questions the very possibility of being sincerely committed to, or settled on, a goal while lacking the belief that one will succeed. At most, the sceptic thinks, what Sonia can commit to is something less ambitious that she believes that she can do—applying to law school, say—while, with respect to her further aspirations, she can commit only to trying to achieve them.

My own view is that, depending on the details, aspirations like Sonia’s are often irrational in light of their risks (although there are many other wonderful things that we can say about those who choose to strive against the odds) [Morton and Paul 2019]. We have stipulated that the goals in question are not morally required, and, from a prudential point of view, the cost of pursuing an option that is unlikely to work out is often too high to pay. With that said, prudence should not be confused with timidity. There is an important kind of value involved in attempting difficult things, even apart from the value of succeeding. And certain valuable ends in life—great accomplishments, long-term relationships—will be inaccessible to most of us without the capacity to persevere in the face of difficulty. Wealth and social privilege make such ends easier to achieve and the risks less costly to undertake, but we should hope to avoid the conclusion that they are rationally off-limits for all but the most fortunate (although this is in part an empirical matter, and might be true in the end). Therefore, I think that we should try to understand, as being rationally permissible, some choices like the ones that Sonia made.

Jennifer Morton and I have argued elsewhere [Morton and Paul 2019] that optimism will be an important part of such an account. The rational permissibility of pursuing difficult goals will be enhanced, other things being equal, to the extent that the agent is entitled to expect her efforts to succeed. However, the first part of this paper aims to emphasize the limits of that idea. Although recommendations abound from social scientists and self-help gurus to the effect that we should be hopeful and believe in ourselves when we undertake ambitious goals, agents in evidential contexts like Sonia’s should not in fact believe that they will succeed. The rationality of striving is undergirded not only by optimism, but also by prudent investment in backup plans for the case of failure. And we cannot make sense of investing in a plan B if the agent believes that her plan A will not fail.

Thus, the central claim of the paper’s second half is that agents in such contexts should be agnostic about success. They must not believe that they won’t succeed, but they should not believe that they will. The second task of the paper is to investigate the nature of agnostic practical commitments, and the implications that such commitments have for rational planning. Against the aforementioned sceptic, I will argue that agnosticism about success does not entail that one’s commitment is weak or outside the sphere of rational planning altogether. We cannot fully plan on such commitments, but we may plan much of our lives for and around them. With that said, we ought sometimes to form backup plans as well. An account is therefore needed of what such plans are, and of how they should interact with our other ‘primary’ plans.

2. Believing in Oneself

I will start by outlining the case for optimism about achieving our difficult goals, and then underscore the limits of this kind of argument. Begin with the observation that pessimism can be a major impediment to success. There are a variety of reasons why people abandon their difficult goals, but doubting that one’s efforts will be rewarded is often a significant factor. Sonia had substantial evidence, before she adopted her goals, that her chances were not favourable. But even one who starts out as sanguine, perhaps because there is little evidence either way, will tend to encounter setbacks, obstacles, rejections, and other kinds of disappointments as they proceed. For instance, Sonia’s first grade at Princeton was a C. The school newspaper there regularly published op-eds lamenting the affirmative-action policies that helped her to gain admission, claiming that recipients ‘could rightly be expected to crash into the gutter built of [their] own unrealistic expectations’ [Sotomayor 2013: 145]. Later, she concluded her summer internship at a top law firm without receiving the customary job offer. These kinds of experiences incline people to abandon their ends, not only because the experiences are emotionally painful, but also because they are evidentially significant: they suggest that you will not achieve your goals even if you continue to strive. A rational agent who takes herself to have good and more feasible alternatives will be motivated to avoid paying high opportunity costs and ending up with nothing. As the negative evidence mounts and her estimate of the likelihood of succeeding falls, the more she will be inclined to abandon that goal in favour of one where success is more probable.

In some cases, however, this would be a failure of grit. We should avoid wasting resources on a lost cause, but the tendency to give up too quickly when challenges arise is also problematic. As noted, the capacity to persevere in the face of difficulty gives us access to certain kinds of valuable long-term achievements and relationships that would otherwise be impossible for us (without quite a bit of luck, or enormous resources to clear our path of obstacles). Whether it is all-things-considered good for a particular agent to be gritty depends on the circumstances in which she normally finds herself; it can be better not to be disposed to persevere if the costs of doing so are generally too high, due, for example, to conditions of extreme poverty or severe discrimination. But, for most of us, the capacity for grit is beneficial on the whole. Depending on the further details of the case, Sonia might be seen as falling into the former camp, but I will assume that she is an agent for whom grit is generally beneficial.

The disposition to be optimistic that one’s efforts will pay off clearly contributes to the capacity to persevere. The claim is not that it is strictly necessary. There are myriad ways that one can avoid quitting; for instance, one can stubbornly refuse even to consider the possibility. Alternatively, one might have an unusually high tolerance for pain or risk, assign a high value to the display of grit itself, or be inclined to value a goal significantly more in proportion to the difficulty that one encounters in trying to achieve it (‘sweet grapes’). But most of us in need of grit do not have such values, tolerances, or dispositions; nor do they seem particularly rational to inculcate.

Moreover, those who do find ways to persevere in the face of significant doubt about their abilities and prospects can pay a high price from the perspective of mental health. Consider the aforementioned phenomenon of impostor syndrome—which is, roughly, a condition in which an agent engaged in a difficult pursuit entertains substantial and ongoing doubt about her ability to perform adequately, and in which she fears that her lack of ability will be discovered. Although the syndrome itself is not actually a recognized psychiatric disorder, it often involves symptoms that are. Sufferers frequently experience subclinical or even clinical levels of anxiety and depression. And, on the behavioural side, it tends to lead to wasteful and counterproductive practices like excessive preparation for a task or extreme procrastination [Sakulku 2011].² Ridding oneself of this condition is increasingly one of the imperatives of the self-help aisle. Of course, impostor syndrome is only one example of the psychic costs of pessimistic striving; one need not think of oneself as a fraud to suffer distress and despair in these circumstances.

The solution, it seems, is optimism. If we were entitled to be optimistic about achieving our difficult goals, this would help both to rationalize perseverance in some cases and to protect us from the psychic costs of pessimistic striving. We have stipulated that the evidence does not recommend optimism, and so the question is that of whether and to what extent these instrumental considerations can affect what we are rationally permitted to believe. To suppose that they can would be to adopt some form of pragmatism about, or pragmatic encroachment on, the justification of belief. There are a variety of forms that such pragmatism can take, some more radical than others and all of them controversial. But the point that I wish to make here is that, even if we assume the most radical form of pragmatism, the benefits of optimism are limited and frequently outweighed. Although I myself accept the pragmatic arguments just offered, they do not support the conclusion that we should all-out believe that we will succeed in such cases. A norm that required or permitted such confidence would be not only disadvantageous but ethically problematic.

First, quite simply, undertaking and persevering at a difficult goal is sometimes the wrong choice. It is to incur a risk, not only that we might fail and be disappointed, but also that we might pay steep opportunity costs along the way. And these risks will fall disproportionately on those who are already victims of injustice—those with scarce resources, and those who are targets of prejudice and discrimination. Those with copious resources and support can frequently recover from a failed attempt to achieve a difficult goal; they can simply switch to another pursuit. The ‘glass floor’ of wealth allows the offspring of tech billionaires to open failed startup after failed startup without significant consequences. For those without such resources however, a similar failure might be catastrophic. To encourage such agents to believe in their ability to succeed without regard for the evidence is not only dangerous but wrong [Morton ms]. Taking this kind of risk should only be done with eyes wide open. There are thorny issues here about what kinds of evidence the agent should consider in predicting what will happen. Perhaps it is a kind of bad faith for her to weigh the possibility that she will lack the strength of will to succeed [Marušić 2015]. But we can set aside that question, since it is clearly not a matter of bad faith to consider the possibility that she might lack the ability or the opportunity. Rather, to neglect to do so would be foolish, opening her up to potentially intolerable costs.

A somewhat less obvious point is that the proposal fails to make sense of investing resources to offset such risks. Suppose that Sonia sets herself the end of becoming a federal judge, and thereby comes to believe that she will not fail in her endeavour. It follows that, in planning for the future, she should treat scenarios in which she does not achieve her goal as being largely irrelevant. If she has settled the question of what will happen, why take costly measures to prepare for other scenarios? Why save money for retirement if her judgeship will qualify her for a handsome pension? Why freeze her eggs if she plans to devote herself to the legal career that she will have? Of course, she need not take herself to be infallible; she might well grant that she could be wrong. The bare acknowledgement of her own fallibility might make sense of making relatively cheap provisions for failure ‘just in case’. It will not suffice, however, for rationalizing large expenditures that she is committed to viewing as wasteful. Thus, this thought fails to vindicate the fact that it is sometimes rational to pay significant costs to provide for the possibility of failure. Once again, the harmful effects of a rational norm that effectively advises against doing so would be worst for the most disadvantaged agents. They are the ones who most need a plan B.

This argument assumes that we should plan on, and only on, our beliefs. Could the proponent of believing in oneself make room for backup planning by denying this assumption? The claim might be that we should plan on some other representation of what will happen in the future—our credences, our acceptances, or our worst fears. But this would be an odd move. For one thing, many philosophers think it essential to the role of beliefs that they serve as the premises in our practical reasoning. This is one feature often said to differentiate beliefs from credences. Second, on pain of undermining much of what was attractive about the proposal, the claim would have to be that we should normally plan on our beliefs, and should use our credences or worst fears only for certain precautionary purposes. If the latter states are relevant to the question of whether to persevere through difficult and discouraging periods, the belief that one will succeed might not help to rationalize perseverance. At the least, we would be owed a complicated and counterintuitive story about when we should plan on our beliefs and when we should not.

In the absence of such a story, we should reject the idea that committing to a goal entitles us to believe that we will achieve it. Even if advantage can matter to the justification of our beliefs, such a practice is not sufficiently advantageous. This does not mean that we must entirely reject the line of thought that we have been exploring. On my own view, agents who have committed themselves to a difficult long-term goal ought to reason about their evidence in a way that exhibits ‘epistemic resilience’, entitling them to a more optimistic outlook than a neutral observer might reasonably adopt. But when the evidence is relatively unfavourable, the rational permissibility of perseverance is secured—if at all—not only by optimism, but also by sober investment in backup plans. Thus, such agents are right to remain agnostic about success.

3. Agnostic Commitments

If Sonia lacks the belief that she will achieve her more ambitious goals, however, we might wonder whether she can be fully committed to those goals. After all, she does not take herself to be in a position to settle what will happen. And some philosophers have argued for a tight connection between deciding on a goal, planning on it, and thereby settling the future.

J. David Velleman, for instance, distinguishes between ‘goal states’ and ‘plan states’. The former are simply motivating desires, while only the latter states support the kind of coordination that is involved in planning for the future. For an attitude to count as a plan state, the agent must consider the object to be sufficiently under his control, such that, in deciding to pursue the goal, he is thereby deciding what will happen. This secures him knowledge of the future, and this knowledge is what governs our planning [2007: 208]:

Knowledge of one’s forthcoming actions is what provides the basis for coordination, and it is what constrains the options available to further practical reasoning.

Similarly, Berislav Marušić and John Schwenkler characterize the distinction between desire and settled practical commitment in terms of truth [2018: 321]:

When we desire something, nothing is settled. But when we intend to do something, just as when we believe something, we have made a commitment: We have settled a question or reached a conclusion … It is in light of having settled this question that we make plans and settle further questions. We think it is plausible that the commitments that constitute intentions are truth-commitments—the truth-commitments of an agent.

Like Velleman, they hold that to fully settle the practical question of what to do is to conclude that one will do something and thereby to form a kind of belief. Marušić also links this belief to the sincerity of the commitment [2015: 137]:

Yet when we decide to do something, we should engage in belief-exhibiting behaviour; we should plan for doing what we have decided to do, and we typically should tell others, when they ask, that we will do what we have decided to do. Our unwillingness to proceed in this way would indicate a lack of sincerity in our decision. That is why, again, it seems plausible to think that in deciding to do something, we can come to believe that we will do it.

Further, in a different but related vein, Kieran Setiya argues that we are not entitled to decide to ϕ unless we know how to ϕ [2008: 406–7]:

Can’t I decide to dance the tango at my wedding, one might ask, even if I don’t yet know how? The answer is that this decision would not be justified. Rather, I must decide to learn how to dance the tango and to exercise this knowledge at my wedding, once it has been acquired. … I cannot decide to dance the tango at my wedding without an unjustified leap of faith.

All of these views imply that Sonia cannot decide or plan on attaining, or intend to attain, a seat on the federal bench.

I think it is clear, however, that Sonia’s orientation toward her ambitions is not that of mere desire. She is actively engaged in pursuing a legal career: it is her end, or her will. It poses problems for practical reasoning to solve, and, most importantly, it is governed by some version of the Instrumental norm. Sonia is practically irrational if she remains committed to her end without, at the necessary junctures, pursuing those means believed to be required for realizing her end. There need be nothing rationally amiss if she neglects to learn all of the two-letter words in the Scrabble dictionary, even if playing well at Scrabble is a motivating desire of hers and she knows that two-letter words are the key. But Sonia’s commitment to a legal career is rationally incompatible with failing to sign up for the Law School Admission Test (LSAT).

In my view, this already suffices to show that her attitude is a plan-state. It may be infelicitous to say that she is planning on a successful legal career, in the sense of expecting it to happen, but she is planning for it. Those who wish to deny this may argue that Sonia is only planning for a legal career in the sense that she intends to try to achieve this goal. Strictly speaking, the goal is not part of her intentions or plans, since she does not believe that she will achieve it; rather, she believes that she will try, and that is what she plans to do.

What does it mean to carry out an intention to try? The concept of trying, I submit, is elusive: it admits of a wide spectrum of readings, and thus does little on its own to define the satisfaction conditions of an intention to try. On one hand, there is a trivial reading on which an agent counts as having tried to ϕ as soon as any action at all is initiated with the ultimate aim of doing ϕ. This sense of trying is omnipresent in intentional action, can be virtually instantaneous, and is compatible with nothing at all happening as a result [O’Shaughnessy 1973; Yaffe 2010]. On this interpretation, Sonia counts as having tried to become a lawyer just in virtue of having watched a Perry Mason episode with the aim of furthering that goal (she wants to learn how it is done). On the other hand, there is a substantive reading on which we mean that she is committed to really trying: there are success conditions that she might fail to meet even if she exerts some minimal effort. Roughly, she is committed to exerting significant effort, making sacrifices, seeking out whatever acceptable means are available, and so forth. ‘You did not really try’, we might say to Sonia if she gives up after one law-school rejection or declines to take the LSAT because it conflicts with a Perry Mason marathon. And, of course, the extent of her commitment might lie somewhere in between.

If the concept of trying leaves open exactly what the agent is committed to, some other way is needed of articulating the satisfaction conditions of an intention to try. The claim to be defended by my opponent is that these conditions must go in lockstep with one’s beliefs: the intention can include only those things that one believes one will do as part of trying to ϕ. Perhaps Sonia’s intention is specifically to take the LSAT and to apply to law school, since this is what she believes that she can do. The problem with this strategy is that it fails to capture the sense in which Sonia is planning around the goal of a legal career. It is possible to have the intention to take the LSAT and to apply to law school in the spirit of buying a lottery ticket: one hopes that it will work out, but one does not structure one’s other commitments around the possibility of this outcome. This non-committal form of trying to ϕ does not rationally obligate one to avoid pursuing other courses of action that would prevent one from ϕ-ing if successful. It may be rationally permissible to try at the same time to work one’s way into a position in a crime syndicate, knowing that, if one does so, membership at the bar will be ruled out.

In contrast, Sonia’s commitment does rationally exclude pursuing other options that would actively render her legal aspirations impossible or highly unlikely to succeed. Continuing to have this commitment requires her to sacrifice other goals that strongly conflict with her ability to succeed; it is not rationally permissible to pursue all of them and to wait to see how things work out. In other words, Sonia’s attitude is subject to a norm of consistency that rules out the pursuit of other options that she believes would prevent her from succeeding at her end. And, crucially, the extent of the options that are ruled out cannot be identified only by reference to an intention to take the means about which one is confident. There will be many actions that would prevent Sonia from succeeding at her end but that are consistent with successfully taking some of the means. To understand the scope of what is rationally excluded by her commitment, we must suppose that it includes the end itself and not merely some of the steps along the way.

Thus, we should understand Sonia’s attitude toward her end as a genuine plan-state, subject to rational norms that do not apply to either desire or minimal trying. Whether we should characterize it as an intention is beyond the scope of this paper. Some have argued that, when full belief is lacking, the attitude adopted must be one of ‘partial’ intention or ‘aspiration’.³ I am content with this for my purposes here, provided it is accepted that such agnostic commitments lie squarely within the sphere of rational planning. Importantly, it need not be the case that such attitudes are merely partial commitments. Sam Shpall has defended a notion of partial intention according to which an intention is more partial to the extent that the agent is less committed to, or less inclined toward, realizing the goal [2016]. I think that these two senses of partial intention are largely orthogonal to one another, since the strength of a commitment can vary independently of the agent’s expectations of success. One can be very weakly committed with a high expectation of pulling it off, or very strongly committed with dim expectations. A teenager might nonchalantly throw a rock at a balloon, thinking it likely that he will pop it, while having no disposition to devote further effort if he does not pop it. And a political activist might be devoted to the cause of ending racial discrimination while considering the likelihood of success to be vanishingly small. The difference in their commitments lies not in their confidence, but rather in the way that they are disposed to sacrifice and shape their other plans around the goal.

4. Agnostic Planning

I have argued that it is sometimes permissible to strive against the odds, but that our beliefs, even if relatively sanguine, should reflect the significant chance of failure. Moreover, agnostic commitments of this type do have implications for practical reasoning and rational planning—both in that requirements of instrumental rationality and consistency are engaged, and in that we should sometimes invest in backup plans for the case of failure. But in what ways does agnostic planning differ from confident planning? This is a huge topic, and so I aim only to make some headway here.

Let us start with three widely accepted norms of plan rationality, influentially formulated by Michael Bratman in terms of intention [1987]. First, there is the norm of Means-End Coherence, which forbids us from intending to ϕ without, at the necessary junctures, intending to take means that are sufficient for ϕ-ing. Second, the norm of Consistency forbids us from having intentions that are inconsistent with one another, or with our beliefs. Third, the Agglomeration norm forbids us from having intentions that cannot rationally be combined into one single overarching plan. On Bratman’s view, these norms apply regardless of whether we believe that we will carry out our intentions, and so agnosticism is no obstacle to supposing that they also apply to the kind of practical commitment under discussion—perhaps because it simply is intention.

But Velleman trenchantly questions why agnostic commitments would rationally obligate us to make plans that are consistent and that agglomerate with one another, rather than maximizing our chances of bringing off at least one plan by making several incompatible ones. ‘If I can have a plan without believing that it will be carried out in this world’, he writes, ‘why should I confine myself to planning for a world in which it is carried out?’ [2007: 206]. In particular, he suggests that Agglomeration and Consistency apply to our plans only in virtue of believing that we will realize them, and that backup plans are thereby excluded by these norms. Velleman imagines planning to fly to Chicago tonight while doubting that one will actually arrive. He points out that [ibid.: 205]

[i]f I form two intentions—to fly to Chicago and to have dinner with you in New York—then I will have a better chance of doing something that I intend, and a better chance of enjoying myself on Tuesday.

He submits that it is not in fact rational to make both of these plans, and argues that the best explanation of the irrationality is that we must believe that we will do as we plan.

If Velleman is right that backup plans are excluded by the norms of plan rationality, then these norms must not apply in the same way to agnostic planning—for surely we are sometimes rationally permitted to make backup plans in the face of uncertainty. Is he right? Let us take a closer look at the relevant norms. Bratman states the Agglomeration principle as follows: ‘If at one and the same time I rationally intend to A and rationally intend to B, then it should be both possible and rational for me, at the same time, to intend both A and B’ [1999: 220]. The question is that of what would make it either impossible or irrational to agglomerate multiple plans. Here, the requirement of Consistency does much of the work. If an agent believes that it is impossible to both A and B, then agglomerating those plans will be inconsistent with her beliefs. Further, even if it is possible to do both, she cannot consistently plan to do both and believe that she will not do both.⁴ And one reason why she might believe this is that she has no good reason to do both, or a strong preference not to do both: perhaps she takes herself to have reason to do B only if she does not do A.

At first glance, this is exactly the situation that the backup planner is in. She believes that she will not carry out the entire agglomerated plan because she will complete the contingency plan only if her primary plan fails.⁵ In some cases, like the situation of Velleman’s traveller, carrying out both plans would actually be impossible. If he does make it to Chicago, then he cannot have dinner in New York that night. In other cases, the agent simply views the two plans as in competition rather than as a single complex goal. Thus, it looks as though Velleman is correct to point out that our contingency plans might well violate both Consistency and Agglomeration.

However, this argument seems to have overlooked the possibility of making conditional plans.⁶ Conditional plans have the form ‘ϕ if C’, where the agent is currently uncertain about whether circumstances C will obtain but expects to find out before the need to ϕ arises [Davidson 1978; Bratman 1987; Ferrero 2009; Ludwig 2015]. Bratman suggests that, when we are agnostic about succeeding in one of our plans, the subsequent plans that we make should be conditional [1987]. His well-known example concerns a plan to remove a heavy log from his driveway, where he is uncertain about whether he can move it himself and so is not in a position to plan on the belief that he will have moved it. In this case, Bratman says, ‘I will likely form two conditional intentions: to go to work if I have moved it, and to have the tree company move it if I haven’t’ [ibid.: 39]. Likewise, Velleman’s traveller can plan to have dinner with friend A if he is in Chicago tonight and have dinner with friend B if he is in New York tonight.

These conditional plans are not inconsistent with the traveller’s beliefs, since he believes that he might be in either place. There is also a sense in which they will agglomerate, since, as Luca Ferrero points out, we should not think of conditional plans as commitments to making a conditional true. Rather, when the conditions specified in the antecedent do not obtain, the conditional plan becomes ‘moot’ [2009: 705]. And, as Kirk Ludwig observes, there is an uninteresting sense in which we can count a conditional plan that has become moot as satisfied—at least, it was not unsatisfied [2015: 48]. When one has multiple conditional plans with antecedent conditions that are mutually exclusive, at most one of them can be realized while the others will be rendered moot and thus satisfied in the trivial sense. Therefore, we can consistently agglomerate such plans.

The problem is that this solution mischaracterizes the structure of what we might call the agent’s ‘primary plan’. Again, the thought is that, as soon as doubt is introduced, all further planning must be conditional in form. In Sonia’s case, this will be quite early on in the plan—at the point of scoring highly enough on the SAT to gain admission to a top university, say. Each of the rest of her plans concerning her legal career will be conditional on an ever-more complicated set of circumstances, including the success of her prior plans. Any backup plans that she has will also be conditional plans, although conditional on the case of failure. Let us imagine that if she can have a successful legal career, then she does not want to have children, but that if she does not have that career, then she would like to have biological children and be their primary caretaker for the first few years.⁷ The latter scenario is her plan B, and for this reason she spends a significant amount of money and endures invasive physical procedures to freeze her eggs. If Bratman’s suggestion is right, then we should describe her as having two conditional plans—to have a successful legal career if all goes well, and to have biological children and stay home with them if she cannot have a successful legal career.

This incorrectly depicts the two plans as on a par, as if they are both contingency plans and incur the same kinds of instrumental obligations. When we apply the norm of Means-End Coherence to a merely conditional plan to ϕ-if-C, it requires us only to keep open a sufficiently good version of the option: we must take whatever steps we believe necessary, and some that we believe to be jointly sufficient, to preserve the possibility of ϕ–ing should circumstances C arise. We are not obligated to do more than this, because we have directly committed ourselves, not to ϕ-ing, but only to ϕ-ing-should-C-obtain. This is plausible when it comes to one’s plan B, but not when it comes to one’s primary plan. As we saw in section 3, it is not as though Sonia’s commitment to a legal career will only become active once she discovers that all of the preconditions are satisfied, or that her instrumental obligation is merely to ensure that this possibility remains open. Rather, her commitment to making the career happen is what generates instrumental pressure downwards to the more immediate steps and explains why she takes those steps as means to that end.

Further normative differences distinguish between the two kinds of commitments. Even if Sonia thinks that the chances of either scenario are relatively even, for instance, there is a difference in how she is entitled to involve other people in her plans. Other things being equal, one is less entitled to ask others to take risks, make sacrifices, or devote resources to one’s backup scenario. It might be appropriate for Sonia to ask her parents to set aside money to help her through Yale Law even before she has been admitted, and even if they themselves could use the money. It would, ceteris paribus, be less appropriate to ask them to relocate for the same amount of sacrifice, just in case she will need them to help her with babysitting.

Something must ground these asymmetries. An initial thought is that the difference lies in the agent’s preferences: Sonia prefers realizing her plan A to realizing her plan B [Bratman 1979]. This is surely true, and might well suffice to distinguish the category of contingency plan from the category of primary plan, together with the fact that, under some description, they are both plans to realize the same end. More than this must be said to explain the distinctive rational and ethical implications of backup planning, however. Backup plans are in some sense ‘subordinated’ to one’s primary plans in a way that is unlike ordinary cases of preferring broccoli to Brussels sprouts. The challenge is to articulate in more concrete terms the ideas of plan subordination and priority.

Ferrero suggests that, if one plan is subordinated to another, the unsubordinated plan is given ‘lexical priority’ [2009: 736–7n49]:

When ϕ-ing is subordinated to ψ-ing, we give lexical priority to the latter, primary pursuit. As far as the ϕ-ing is concerned, we are expected to comply with the norms of intention only in the space left, if any, once we have taken care of the demands of ψ-ing. In many cases, it might be that we never get around doing anything to promote the subordinated pursuit, since all our resources are absorbed by the primary plans. Once the fate of the primary plan is settled, we might be required to pursue the subordinated plan, provided that this is still possible.

Although this proposal is something like what we want, I think that it is too strong. We ought sometimes to take steps to further our backup plans even though we know that this will decrease the probability of success in our primary plans. Perhaps spending the money to freeze her eggs means that Sonia cannot afford other things that she believes would further her legal career. Likewise, many people believe (accurately or not) that asking for and getting a prenuptial agreement will lower the probability of their marriages succeeding—not drastically, but enough to prevent them from risking it. Even supposing that this is true, it can be rational to go ahead with the egg-retrieval or the prenuptial agreement. This shows that we need not give absolute lexical priority to our primary pursuits; the dynamics of primary and contingency plans are more complicated than that.

Return to the idea that the means-end obligation incurred by a conditional plan is limited to keeping open a sufficiently good version of the option, whereas our primary plans at least sometimes commit us to doing our best to actually realize the end, consistently with our other commitments and values. This suggests that, even when we entertain significant doubt about success, the content of our practical commitments can be unconditional: we are committed simply to ϕ-ing. They are of course conditioned on one’s beliefs about what the future will be like, as nearly all plans are, in that one will be disposed to alter them if there is a relevant divergence from what one expects or if strong new reasons present themselves. One is not committed to ϕ-ing no matter what, come hell or high water. But we need not understand these eventualities as being part of the content of the plan. The advantage of this is that the normative consequences of the commitment attach directly to ϕ-ing and not to the mere possibility of discharging a conditional. This is why one is obligated to take sufficient means for ϕ-ing (or to give up that end), and not merely to keep the option open and see how things play out. Further, it is why it can be more appropriate to enlist the participation of others in this plan than in mere contingency plans. When there is a risk that it will all come to nothing, it matters that you yourself are unconditionally committed to finding a way if you can.⁸

This proposal can make sense of prioritization without lexical priority. The backup plan ‘B if I do not A’ is subordinate in virtue of the fact that the agent has an unconditional intention to make the antecedent of the conditional intention false. On the other hand, even unconditionally planning to A does not obligate one to make it happen by hook or by crook. In committing oneself to traveling to Chicago, one is not committed to murdering anyone in order to get there, or to spending all of one’s child’s college fund on hiring a private jet. We are likewise permitted to omit the most effective means to our ends if they are in conflict with other important goals, or to further one of our goals at some expense to another. I argued in section 3 that we cannot rationally act in a way that we believe will render our primary plans impossible or very unlikely to succeed. But we can rationally make more limited trade-offs, such as taking means to promote the B plan that to some extent decrease the likelihood of A–ing.

This proposal also does better than a third alternative, according to which primary plans and contingency plans are both conditioned rather than conditional. The idea here is to think of both primary and contingency plans as unconditional plans aimed at accomplishing the same end, but as conditioned on different, mutually exclusive, representations of the future. This view has the same problem as does the view that they are both conditional plans, however: it treats them symmetrically, and thus fails to capture the way in which the plan B is subordinated to the plan A. Further, it seems to require a certain amount of disunity or compartmentalization. It is unclear how two plans that are conditioned on conflicting assumptions about what will happen can be brought together into a single practical perspective, as would have to occur when practical questions arise about trade-offs.

Can’t some primary plans be conditional, however? Surely nothing prevents us from planning to ϕ-if-C—‘go to Bali this summer if the cost of the ticket drops low enough’—while also having a backup plan—‘go to the lake house’—if not-C. I argued earlier that if both plans are conditional, they do not exhibit the relevant kind of asymmetry. But the intuition might remain that one of these conditional plans is a backup to the other. I suggest instead that this case is better described as a matter of having a disjunctive plan [Ferrero 2016]. Calling one disjunct a ‘backup’ is merely a way of saying that the agent hopes that C obtains because she prefers ϕ-ing-in-C to ψ-ing-in-not-C. She prefers Bali if it is reasonably cheap and otherwise prefers the lake house, and her instrumental obligation is to keep each option open until she learns the price of the ticket. So, while there is a preference ranking here, such cases are relevantly distinct from the kind of subordination that we see in Sonia’s example. There is an important difference between waiting to find out whether a condition is satisfied before the full commitment kicks in and being committed to the end all along.

Finally, what about a plan C? The sketch of a solution that I have offered can distinguish between a plan B and a plan C in the sense that the plan C is explicitly conditional on the failure of the plan B, while the B-plan aims to falsify the antecedent of the C-plan. But it does not have the consequence that the plan C is subordinated to the plan B, in so far as both are conditional. As long as the plan A is viable, the other plans are subordinated to it but not to each other. However, note that, once the A-plan is abandoned, the B-plan becomes unconditional and the subordination of the C-plan is thereby established. I think that this result is acceptable, since, as long as the A-plan is viable, the agent is committed to keeping both B and C open in a way that is relatively symmetrical (although investments in C will be further constrained by the need to avoid seriously undermining both A and B).

Much more argument would be needed in order to draw any firm conclusions. But if it is right to think of our primary plans in these cases as unconditional and our contingency plans as conditional, then there is a sense in which rational contingency planning is incompatible with the spirit of the Agglomeration norm. Importantly, this is not because it violates Consistency. There is one way in which the whole agglomerated plan ‘A and (B if I do not A)’ could be satisfied—if the agent succeeds in A-ing, then the conditional conjunct becomes moot. But if she carries out her contingency plan, this will be because she did not A, and her commitment to A-ing is not thereby rendered moot. This means that her agglomerated plan can only be fully satisfied if she does not carry out the contingency part. The contingency plan, therefore, becomes an idle wheel once it is agglomerated with the primary plan: it contributes nothing to the success conditions of the plan when taken as a whole.

Although the problem is not with consistency, I suggest that there is a rational defect in this kind of agglomerated plan. Plans are inputs into further practical reasoning. But if one reasons about how to achieve the agglomerated plan, the conclusion will favour taking means only to furthering the A part of the plan. After all, succeeding at plan A is the only way to realize the plan as a whole. It would not be strictly irrational to further the plan B, but there would be no rational pressure in favour of doing so until one has given up on A. I argued earlier that this is the wrong result. It is only if one reasons about plan B in an atomistic disaggregated way, in competition with rather than in conjunction with A, that we will get the right conclusion about how to proceed. Thus, applying the Agglomeration norm to such plans generates an amalgam that we should not use in practical reasoning. Since it is rational to make contingency plans and to reason about how to promote them, we should conclude that we are not required to agglomerate our primary plans with the associated backup plans.

It is possible that a more restricted version of the norm does apply.⁹ The motivation for agglomerating plans is clearest when we are planning for a single future. Velleman suggests that beliefs, and thus intentions, should agglomerate because they ‘aim to fit the world, of which there is just one’ [2007:206]. Bratman tentatively agrees with this characterization, and argues that it is just as plausible that each intention ‘aims to make its content true as a part of a coordinated realization of one’s planning system, in the world as one believes it to be’ [2009: 52]. The goal of coordination, in turn, is to ensure that we are not working at cross-purposes with ourselves. But I hope to have shown that this does not mean that we must plan for only one future; sometimes, we should simultaneously invest in multiple, mutually exclusive futures. This is an important way in which we broaden the limits of that for which we are rationally permitted to strive.¹⁰

Notes

1 Many details of the example are taken from Sonia Sotomayor’s autobiography [2013]. But some are rather presumptuously imagined. So, I intend that the case throughout be treated as a fictionalized example that draws heavily on real life.

2 Sotomayor characterizes her own experiences as paradigmatic of the syndrome. I think that the phenomenon is a useful data point about the psychology of perseverance, but elsewhere [Paul 2019] I express scepticism about the benefits of conceiving of self-doubt in this pathological way.

3 This move is potentially troublesome for those who aim to link intention with belief for the purpose of explaining the norms of practical rationality in terms of the norms of theoretical rationality [Harman 1976]. If ‘cognitivists’ about intention must grant that there are attitudes that are akin to intention and subject to practical requirements like the instrumental norm, but that do not involve a corresponding belief, the reduction of intention-rationality to belief-rationality leaves unexplained much of the wider sphere of practical or plan rationality.

4 Although, if a set of plans is large enough, it seems perfectly rational to plan on each of them while believing that you will not carry out all of them [Shpall 2016].

5 There might be elements of the plan that she will pursue even if her primary plan does not fail. It is only a plan B if conceived of as an alternative to one’s plan A, but a variation of the same activity could also be part of an A plan if undertaken for a different purpose (as a hobby rather than a career, say).

6 Some have thought that any doubt about success necessarily renders the plan conditional: it’s a plan to ϕ ‘if I can’ or ‘If I am not prevented’. But this is not a meaningful sense of conditional planning; these generic conditions could be added to the expression of virtually any intention [Anscombe 1963; Davidson 1978].

7 I am emphatically not suggesting that one cannot do both; I am simply imagining that this person might only want to do one or the other.

8 An alternative picture defended by Ferrero [2009] might counter that the content of these commitments is in fact conditional, but that the conditions are assumed to have been satisfied and therefore pushed into what he calls the ‘cognitive background’. In such circumstances ‘the agent might thus proceed to reason as if the intention is unconditional’ [ibid.: 709], which would suffice for my purposes. However, I doubt that the agent will be best understood as assuming that the conditions of succeeding are satisfied in these cases. Sonia is painfully aware that they are not, and her reasoning about how to proceed and what precautions to take is shaped accordingly.

9 For instance, it might be that all plans that are conditioned or conditional on the same circumstances, such that they would be carried out in the same future, should be agglomerable with one another.

10 My thanks to Jennifer Morton, Antonia Peacocke, Matthew Silverstein, the members of the ORGiE reading group, and audiences at Northern Illinois University, the University of Chicago, the University of Vienna, the 2020 Royal Ethics Conference at UT-Austin, the University of Leeds, and the Washington University in St. Louis Ethics and Epistemology group for very helpful conversations and comments. I am especially grateful to two generous and incisive referees for this journal, whose constructive advice permeates the paper.