Prudential Redemption and Its Significance

2.1 Failure Caused by One’s Mistake

While many things can causally contribute to an unsuccessful project or a broken relationship, the kind of failure under discussion mainly results from one’s own mistake — i.e., without the mistake, the relationship or the project would have continued smoothly.¹¹ In some cases, others might also suffer from one’s failure. For simplicity, I will focus on the harms that accrue to the person who made the mistake. What makes such a failure prudentially negative, then, is the fact that the person would have been better off had she not made the mistake.

The mediating factor for failures under discussion is most likely to be some lesson(s) the person has learned from the very experience.¹² A lesson that directly responds to the relevant mistakes meets the relevance condition because its content makes essential reference to the cause or the bad-making features of the negative event; the reference is essential when the lesson cannot be adequately described without referring to the latter. It is worth explaining more about what counts as a lesson. Typically, a lesson contains some backward-looking thoughts about what one could have done better, which can be translated into forward-looking maxims that safeguard against future failures. Also, a lesson often comes with information that is new to the person, whether it is a belief about some new subject matter or a novel manifestation of some existing belief. For the purpose of redemption, even if the person does not receive any new information, an existing belief becoming motivationally efficacious also qualifies as a lesson. In this case, the relevance condition is met because the explanation of how the person acquires this lesson essentially refers to the cause or the bad-making features of the negative event. A reasonable lesson does not have to be a correct diagnosis of one’s failure, though a correct diagnosis is more likely to meet the manifestation condition.

The manifestation condition guarantees that candidates for the mediating factor (i.e., lessons that have met the relevance condition) establish the right kind of relation with the later event. Here, it is helpful to make a distinction between the instrumental and the intrinsic value of the candidates. If a candidate only has instrumental value, it must bring about something positive to redeem the negative event. Imagine that Melissa, who has failed several midterms, learns she should never stay up late the night before an exam. This lesson has no value except that it might improve her performance in future exams. If the lesson does not make her do any better in the next exam than the previous ones, merely learning it does not suffice to redeem her past failures. But even in the successful case, if the lesson only plays a dispensable role in the process, her previous failures are still not redeemed. Suppose Melissa did well in her final exam, but her success was overdetermined. That is, even if she stayed up late the night before the exam, she would still get a good grade because the exam was too easy or she was over-prepared. It seems her previous failures cannot be redeemed via that specific lesson she learned.

What if a candidate for the mediating factor has some intrinsic values? Suppose Melissa gains from her academic failures some important self-knowledge — say, she realizes that her perfectionist tendencies have become a source of anxiety that prevents her from making progress. Some might think that merely acquiring this piece of self-knowledge without it leading to any further positive events can directly redeem previous failures; after all, learning something about oneself can be seen as a positive event. But the issue is more complicated than it appears, for even intrinsically valuable things might causally contribute to something bad. Imagine that Melissa adopts some strategies to counter her perfectionist tendencies — say, she stops polishing her essay or paying attention to details in her exams. Unsurprisingly, her academic performance gets even worse. If learning about her perfectionist tendencies causes all and only negative events in her life, we might hesitate to say that her previous failures are redeemed by the newly acquired self-knowledge, for it is incompatible with our intuition about redemption — namely that negative events can somehow influence one’s life positively. Thus, even if a mediating factor possesses intrinsic values, it needs to be manifested in a positive way. Suppose Melissa, upon learning the source of her anxiety, instead begins to meditate regularly, which allows her to ruminate on the newly acquired self-knowledge and eventually helps her overcome the unhealthy perfectionist tendencies. The mediating factor (i.e., self-knowledge) is distinct from the positive event (i.e., fruitful meditation); nevertheless, the former constitutes a good-making property of the latter.

Lastly, let us use Ding to test these two conditions. Ding’s self-reflection about his unhappy marriage yields a lesson whose content makes essential reference to the cause of the negative event — namely, his passive-aggressive way of handling conflict. This self-reflection is translated into a maxim that later helps him navigate through workplace conflicts, which is indispensable in bringing about his promotion and career success. Thus, the mediating factor account can provide an adequate explanation of Ding’s case.

2.2 Self-Sacrifice for Long-Term Goals

Self-sacrifice under discussion refers to the costs a person has paid for the sake of a goal whose achievement will benefit her in the future.¹³ The costs are broadly construed as whatever the person has given up in order to increase the likelihood of achieving the goal; they include, but are not limited to, the person’s material, financial, intellectual, emotional, and interpersonal resources. What makes this kind of self-sacrifice prudentially negative is that the person’s well-being at the moment is diminished, regardless of whether the goal will be achieved later.

Given the nature of self-sacrifice, the mediating factor will be whatever one obtains for the sake of the goal through the costs one has paid; they can be specific skills, habits, perspectives, products, etc. that one has cultivated, accumulated, or created along the way. Call these things assets. Assets satisfy the relevance condition because the explanation of how the person acquires them makes essential reference to the bad-making feature of self-sacrifice. Consider Ayumi, who spends time in L.A. receiving music training and putting on shows. An adequate explanation of her being able to obtain all the assets must refer to the fact that she separates from her family and friends in her hometown and works multiple part-time jobs to survive in L.A. Note that the following ways to obtain the assets are irrelevant to self-sacrifice. If the assets are obtained for a long-term goal but not through one’s own resources, there is no self-sacrifice involved; if they are obtained through one’s resources but not for the long-term goal, there is no self-sacrifice involved — for the assets are immediate compensation for one’s investment.

Relevant assets must also meet the manifestation condition to redeem past sacrifices. The most straightforward way to do so is to achieve the goal one sets out to pursue. Think about Alexa and Alexander. Alexa’s sacrifice for music is redeemed because the relevant assets play indispensable roles in bringing about her success. However, if the assets do not play an important role, if any role whatsoever, we tend to think that the past sacrifice is not redeemed. Alexander is a good example. His international reputation is brought by the lottery winnings, whose acquisition cannot be explained by his sacrifice for music. The relevant assets (e.g., his songwriting skills), however, do not make salient contributions to his success. Hence, my account can yield a result that accords with our judgments about both cases.

In the case where the person fails to achieve the initial goal or abandons it halfway through the pursuit, past sacrifices can still be redeemed if the relevant assets contribute to her other goals, either existing or newly adopted ones.¹⁴ For Ayumi, her other existing goals include maintaining a relationship with Sergio. Suppose Sergio begins to play some guitar, and they often practice together. The fact that music now becomes an important way to maintain their relationship somewhat redeems Ayumi’s earlier sacrifice, though to a much smaller degree than achieving her initial goal. This is because different goals often require different packages of assets, and it is not always easy to fully utilize the assets ready for one goal in another pursuit — unless the new pursuit is carefully tailored to the existing assets. Suppose Ayumi eventually gives up on songwriting and becomes a music teacher. Even though the initial goal is not achieved, her sacrifice for music can still be redeemed to a decent degree.

Now, a similar question arises: Can past sacrifices be redeemed directly by assets with intrinsic values? Assets are likely to contain intrinsic values, for activities that exercise rational capacities are valuable in themselves. But given that a package of assets is often tailored to a certain pursuit, I suspect that the whole package has more instrumental than intrinsic value in many cases. Even for the portion with intrinsic value, we should not forget the corresponding costs of these assets. If the costs outweigh the intrinsic value, obtaining the assets alone is insufficient to fully redeem the sacrifices; the assets must contribute to some future goal(s) to offset the remaining costs. Although the sacrifices, in this case, are partially redeemed by the assets as the mediating factor, the manifestation condition is still essential — for its purpose is to ensure that the assets exert positive influences on one’s life. Now, suppose the costs are outweighed by the intrinsic value of the assets. The person’s sacrifice, in this case, is immediately compensated, which means her momentary well-being is not diminished and the event overall is prudentially positive. If so, my account no longer applies because nothing needs to be redeemed in the first place.

3.1 Mediating Factor and the Effect of Redemption

Let us begin by distinguishing an improvement sequence, in which the negative and the positive events only satisfy condition (a) in my account, from a redemption sequence, in which both (a) and (b) are satisfied. Put differently, there is a mediating factor between the negative and the positive events in a redemption sequence, one that does not necessarily appear in an improvement sequence. While this distinction highlights how a redemptive trajectory is more than a bare upward trajectory, why the distinction must be drawn in the way I have described might not be self-explanatory. After all, the negative event in many cases does not just precede the positive event. Ayumi’s sacrifices for music and her relationship with Sergio bear a causal relation, which is stronger than temporal adjacency but does not count as redemption on my account. Without articulating the explanatory value of the mediating factor, it is unclear why the distinction between redemption and improvement is not arbitrary.

According to our pre-theoretical understanding, the central idea of redemption is that some negative things in the past are somehow made less bad.¹⁵ The mediating factor, indeed, is defined for the sake of stably producing this effect; its purpose is to guarantee the right kind of relation between the negative and the positive events such that the evaluation of the former can be improved by the latter. In the case of failure, the past mistake is overcome or rectified; in the case of self-sacrifice, the costs one has paid are justified or rationalized without going to waste. Of course, one’s well-being in the past cannot be improved by the later event, unless we endorse some highly controversial claims about backward causation. But this does not prevent us from reevaluating the significance of past events to our life. Once a redemption sequence is complete, a diachronic perspective can be constructed based on the mediating factor. When being evaluated from this perspective, a redeemed negative event becomes less bad to one’s life as a whole than it otherwise would.

By contrast, for improvement sequences where the negative event merely precedes the positive one, the negative event cannot be reevaluated in a similar way due to the lack of mediating factors. It is undeniable that the positive event contributes to the person’s current well-being; without it, the negative event might persist indefinitely. But however good it is, this positive event can only be seen as a stand-alone event. At most, it might make the person feel glad that the negative event is prevented from making her life even worse; if two events are similar in kind, the positive event might even make the person feel less painful when she looks back. Still, the evaluation of the negative event remains the same.

However, it is reasonable to doubt whether my claim is also true for improvement sequences where the negative event causes the positive one. Suppose Andy hits several cars on the highway while rushing to the airport and thus misses a flight that ends up crashing. Some might have the intuition that avoiding the plane crash makes the car accident less bad. Yet the bare causal relation between the two events seems sufficient to produce this effect. My response to this objection is threefold. First, cases like Andy are not counterexamples to my account, for my account does not claim that the mediating factor is necessary for making the past less bad. Second, there is good reason to think that mere causation is insufficient to produce the said effect in general, considering why we move away from the bare causal account at the outset. Third, even if causing a positive event, in some cases, can compensate the loss incurred in a negative event, the redemption sequence still involves a distinct way of making the past less bad — namely, by improving the evaluation of the negative event. In Andy’s case, avoiding the plane crash merely mitigates how bad the car accident is, but it does not rectify what makes the accident bad — i.e., his reckless driving.

Some might complain that focusing on redemption distracts us from more natural ways to explain why overcoming failures and getting sacrifices compensated are good for our life. According to Dunkle (2022, 588–589), overcoming failures has more to do with the person’s growth in motivation or ability, which is itself a valuable feature of life. Compensated sacrifices tend to be discussed together with achievement (Portmore 2007) or the completion of projects (Dorsey 2021, 158–176). Along these lines, Clark (2018) appeals to self-realization to identify the value of upward trajectories more generally. What these views have in common is the idea that the positive present deserves more attention than the negative past.

These seemingly competing explanations, however, do not necessarily exclude each other and might well coexist. Ding’s unhappy marriage is made less bad by his personal growth and subsequently his career success. If Alexa redeems her sacrifices for music, it sounds reasonable to say that her journey is also one of self-realization. Indeed, competing explanations can be seen as different ways to describe the same process. I mainly focus on the negative events here because my goal is to identify the structural features shared by cases of redemption. Although the idea of growth might capture the value of redeeming failure, it does not offer a natural explanation for redeeming self-sacrifice. After all, having to make self-sacrifice does not mean that our motivation or ability is weak, and redeeming them does not require us to grow in either way. By contrast, appealing to achievements might account for the value of redeeming self-sacrifice but not for that of redeeming mistakes, since not every mistake is redeemed by completing a valuable project. While self-realization has the potential to accommodate both, it is too broad to pick out what is unique about redemption — for growth, achievement, etc. might also involve self-realization. Thus, making the past less bad is a better candidate than the alternatives.

3.2 The Weight Reduction Proposal

With the contrast between redemption and improvement, it is easier to see how a redemptive trajectory is prudentially better than a bare upward trajectory. Because of the mediating factor, a redeemed negative event is made less bad and hence contributes less to one’s lifetime well-being than its unredeemed equivalent. The adjustment of contribution, as I will argue, is fulfilled by reducing the weight of the momentary well-being associated with this event in the lifetime well-being calculation.¹⁶ The weight reduction proposal has theoretical advantages over two competitors. Before discussing them, I will first unpack this proposal.

The weight reduction proposal presupposes that various life events can have different weights in one’s lifetime well-being. Just like calculating course grades, while students get a score for each exam, these scores are not always weighted equally in the final grade — e.g., all midterms have equal weight, but they are less weighty than the final exam. In the case of well-being, all life events are assumed to have equal weight in the mere sum of momentary well-being. However, the overall structure can exert influence on one’s life by adjusting the weight of various events in the calculation of lifetime well-being. This framework clarifies how lifetime well-being can be separate from the mere sum of momentary well-being without turning it into a mysterious whole that cannot be computed from its parts. There is no lack of intuitive support for the idea of weight and the possibility of weight adjustment. We often think that some life events are more (or less) significant than the others, and their significance can be altered by later events.¹⁷ For example, friendship is usually considered an important prudential good. However, the later betrayal of one’s friend can drastically decrease the significance of this friendship, compared to a similar friendship that naturally faded away.

Indeed, the idea of weight — especially in relation to lifetime well-being — is not entirely new in the literature. When introducing the initial example of the Shape-of-a-Life phenomenon, Slote (1982) contends that prudential goods obtained in one’s prime bear greater significance than those obtained in childhood or dotage. On this view, the exact same achievement has more weight in one’s life if it is achieved, say, in one’s thirties rather than seventies. The framework I introduced earlier differs from Slote’s view in one crucial respect. On my view, the adjustment of weight is based on the relation between events rather than the event’s temporal location in a life.¹⁸ When a redemptive relation holds between a negative and a positive event, the weight of the negative event in one’s lifetime well-being will be reduced in proportion to the degree of redemption. Other relations besides redemption might possess a similar ability to adjust an event’s weight. A good example is corruption, a relation in which an earlier positive event is made less good by a later event.¹⁹ While the weight reduction proposal has the potential to also explain how corrupted positive events might contribute less to one’s lifetime well-being, for the purpose of this paper, I will only discuss its application to redeemed negative events.

The weight reduction proposal, however, is not the only game in town when it comes to the mechanism by which a redeemed event is made less bad to one’s life. An obvious alternative is what I call the direct deduction proposal, according to which the negative value of a redeemed event is directly deducted. Although both proposals yield a similar result (i.e., less contribution of the negative event), only my preferred proposal allows us to capture the intuition that a redeemed event is made less bad without also making one’s well-being in the past less bad. After all, directly deducting the negative value of a past event amounts to increasing one’s momentary well-being associated with that event. Yet, my preferred proposal does not have this troubling implication, for what is adjusted on this proposal is not the momentary well-being but merely its weight.²⁰ The more a negative event is redeemed, the lower its weight will be in the calculation of lifetime well-being. In the case of full redemption, the negative event might cease to have any weight, even though the corresponding momentary well-being remains unchanged. Conversely, if a negative event is unredeemed, like the ones in bare upward trajectories, its weight cannot be reduced in the same way.

The weight reduction proposal also has advantages over what I call the bonus proposal, which treats the value of redemption as a separate value that emerges from a specific combination of negative and positive events. On this proposal, prudential redemption enhances one’s lifetime well-being by adding extra values on top of the unadjusted values contained in all events by virtue of the redemptive relation. This proposal, however, has a different troubling implication. Note that adding bonuses in such a way amounts to claiming that the redemptive relation has intrinsic value, which gives rise to at least a pro tanto reason for us to pursue it. If so, we have reason not only to redeem negative events that have already happened but also to create a few on purpose merely for the sake of redeeming them later. While the first half of this recommendation sounds reasonable, the second half is clearly absurd. My preferred proposal, however, can keep the first half but leave the second. The possibility of weight reduction gives us a reason to redeem negative events, but it does not presuppose that the redemptive relation is valuable as such.

The remaining task is to explain in virtue of what the redemptive relation can reduce the weight of a negative event in the lifetime well-being calculation. It is tempting to conceive this relation as one of instrumentality, namely that something related to the negative event brings about the positive event. But this is the wrong way to look at it. Note that the redemptive relation has a certain direction, which goes from the positive to the negative event instead of the other way around. This is because the mediating factor, strictly speaking, is a relational property of the positive rather than the negative event; after all, it only makes essential reference to some specific things about the negative event, but it does not belong to this event. Thus, the redemptive relation, when conceived in the right way, is one in which the positive event can transform the past event. The redeeming event is not just something good as such; rather, it has the kind of good that also overcomes or transfigures the badness of the redeemed event by virtue of the mediating factor. To illustrate with Ding, the career success built on his open-mindedness is an explicit repudiation of his passive-aggressiveness that leads to the divorce. As the redeeming event, Ding’s success carries an evaluative import that goes beyond merely compensating for the loss of momentary well-being in his unhappy marriage.

While the redemptive relation derives its value from positive events with intrinsic value, this does not mean that, by reducing the weight of the negative event, the value of the positive event is double counted. Note that the mediating factor does not play any role in calculating the momentary well-being associated with the positive event; after all, it is not one of the relevant properties that determine the intrinsic value of any prudential goods.²¹ Although the mediating factor does not have a say in how good the positive event is, this relational property is tracking to what extent the badness of the negative event is transformed by this positive event. Accordingly, the weight of the negative event will be reduced in proportion to the degree of transformation. Even if some might insist that there is still a subtle and indirect form of double counting involved, I believe the move I am making here is not entirely unwarranted.

3.3 Dunkle’s Dilemma

Recently, Dunkle (2022, 596–600) posed a challenge to theorists of redemption who rely on a distinction between synchronic and diachronic perspectives on a person’s well-being. According to Dunkle, these theorists face a dilemma when evaluating whether a redeemed event is good or bad for the person. If they think that the negative event, when being redeemed, is not (that) bad for the person diachronically but bad for her synchronically, they are committed to two-answer redemptionism. If they only hold that the redeemed event is not (that) bad for the person, either by prioritizing the diachronic over the synchronic perspective or by reconciling two perspectives to yield a single result, they are committed to one-answer redemptionism. Dunkle claims that one-answer redemptionists give the wrong answer to the question because it downplays the hardship in a person’s life, and in worse cases, even apologizes for the oppression she has suffered from. By contrast, two-answer redemptionists have other problems, for they claim that the synchronic and diachronic perspectives are not to be reconciled. If they drop this claim and try to reconcile two perspectives, they will become one-answer redemptionists and face the trouble mentioned above. But if they leave two perspectives unreconciled, there is no reason why the person should adopt the diachronic perspective (which Dunkle considers to be delusional) rather than the synchronic perspective when evaluating the negative event.

At first glance, I seem to be a quintessential two-answer redemptionist in Dunkle’s taxonomy, for my view appeals to both synchronic and diachronic perspectives on the redeemed event. However, I do not endorse the further claim that both perspectives are not to be reconciled. Although my project is inspired by Velleman (1991), I do not share his view that lifetime well-being and momentary well-being are two discrete axes of value nor that there is no algorithm for us to compute one from another (63–68). Indeed, the weight reduction proposal elaborated earlier can be considered a rudimentary algorithm to compute one’s lifetime well-being from momentary well-being. It allows us to coherently hold that the redeemed event was bad for the person during the period it occurred and that it is not (that) bad for her life as a whole. Thus, even if I deny the irreconcilability of the two perspectives, it does not follow that I will end up as a one-answer redemptionist. But for the sake of argument, I will grant that two-answer redemptionists must accept the further claim attributed to them by Dunkle. Still, it is unclear why there is a problem to leave two perspectives unreconciled. After all, being a two-answer redemptionist is just to offer two answers of equal importance to the question “Is a redeemed event good or bad for the person?” Forcing them to choose or prioritize one over another is begging the question against two-answer redemptionists.

Yet, Dunkle might rephrase his challenge by asking why the person must adopt the (delusional) diachronic perspective in addition to the (real) synchronic perspective. To respond, I do not think the diachronic perspective is less real than the synchronic one, for the mediating factor account aims to identify the objective relation between the redeemed and the redeeming events. Even if no one adopts the diachronic perspective, the redemptive relation still holds. However, one reason for adopting such a perspective is to enjoy the psychological benefits derived from it. For the protagonist, knowing that a negative event in her life has been redeemed is likely to help her find closure from the past or effectively cope with negative emotions. For the audience of her story, they might also feel inspired and hence rekindle their hope in life. Even if Dunkle is not impressed by these psychological benefits, the reason against adopting the diachronic perspective should not be that it is delusional.

4.1 Decision-Making in Face of Equally Good Options

Earlier, I established the claim that a redemptive trajectory is prudentially better than a bare upward trajectory when the sum of momentary well-being is held fixed. This claim has a practical implication for deciding between options that are equally good regarding their prudential values determined by non-relational properties.

Imagine that Ben has a history of alcoholism but later begins a new life as a law student. Now, he is choosing between two jobs, one in a law firm as an associate attorney, and another in a non-profit dedicated to helping juvenile delinquents with substance abuse problems as a legal consultant.²² Assuming that both jobs are equally good, the consideration of completing a redemption sequence can tip the scales in favor of one option over another. For Ben, being a legal consultant in the non-profit can redeem his dark history but working in a normal law firm cannot. Indeed, even if the redemptive option is in itself less good than the alternative, at least in some cases it is still prudent to choose it if doing so can fully redeem a very negative past event. Suppose Ben gets an offer from a prestigious law firm, which is also his dream job. When considered independently, this job might be prudentially better than the non-profit one. But given his past, the non-profit job indeed has huge redemptive value, and hence it is possible that the overall prudential value is still higher than his dream job.²³

Of course, the consideration of redemption can be outweighed in many ways. If the non-redemptive option has much greater prudential value or the redemptive option can only redeem a trivial negative event, it is imprudent to choose the redemptive option. I only mean that the consideration of redemption is a pro tanto reason for a certain decision or course of action. It might be the case that whenever there is a negative event in our life, we have a pro tanto reason to redeem it. But at the end of the day, the strength of this reason depends on how negative the past event is, how much and how likely it can be redeemed, and how valuable the non-redemptive options are.

Note that this implication is limited to comparing redemptive and non-redemptive options. For options that have the same non-redemptive and redemptive values, my account remains neutral between them. There might be reasons for redeeming one rather than another negative event in such cases — e.g., one redemptive option is more coherent with the person’s other values or commitments than another. But these reasons are beyond the sphere of redemption.

4.2 The Rationality of Honoring Sunk Costs

A person honors sunk costs when some unrecoverable costs she has paid for a project influence her current or future decision. Steele considers this to be the “main form” of honoring sunk costs discussed by economists (1996, 608). But Steele also points out what he calls the “Concorde” form, according to which historical costs that have not yet been recouped call for special attempts to recoup them in the future (1996, 609). Honoring sunk costs in either way is irrational because it leads the person to adopt inferior options.

Consider Michael, an undergraduate who has not declared his major and is choosing between engineering and biology. While he likes biology more, he decides on engineering solely because, in his first two years, he has taken many general education courses that count toward an engineering rather than a biology degree.²⁴ Had he taken a different set of courses for general education, he would in fact major in biology. But considering the time and energy he has spent on studying engineering, he thinks it is better to declare that major — even if this means he will invest more in the next two years in a major he does not prefer. Michael is thought to commit the sunk cost fallacy because he regards the costs incurred in the past as the only reason to invest more in finishing an engineering degree.

Portmore (2007, 25) and Kauppinen (2020, 659) attempt to rationalize the behavior of honoring sunk costs by questioning whether the costs are genuinely sunk in some cases.²⁵ Honoring sunk costs has many similarities with redeeming self-sacrifices. If there are assets that bear a proper relation to past costs, and the assets can contribute to the person’s existing or newly adopted goals, the costs should not be considered unrecoverable. For Michael, the costs incurred in the past can be recovered by majoring in engineering instead of biology, since the relevant assets (i.e., the courses) can contribute to the former but not to the latter. Thus, there is at least a pro tanto reason for Michael to take the recoverable costs into consideration, though the strength of this reason depends on how much the costs can be recovered.

But someone can still commit the “Concorde” form of sunk cost fallacy if she holds on to the relevant assets that might redeem the recoverable costs. In these cases, questioning whether the costs are genuinely sunk misfires. Suppose Michael tries very hard to find a major that can maximally utilize the assets obtained from previous courses. Although he is not fixated on genuinely sunk costs, it still seems irrational for him to do so for two reasons. First, he ignores further costs he might have to pay (i.e., two more years doing engineering) to recover the past costs. Second, he fails to make a comparative cost-benefit analysis with other available options (i.e., biology). Suppose both majors require the same costs and Michael can benefit equally from both in the future. It is indeed rational to choose engineering because of its redemptive value. Things get complicated when majoring in engineering is more costly or less beneficial than majoring in biology. However, it is still rational for Michael to choose engineering if the recovered costs are greater than the extra benefits offered by biology.

So far, we have stipulated that majoring in biology cannot recover the past costs. But what if it can? Suppose some courses Michael has taken also count toward a biology degree. Is there still a way to say that choosing engineering can be somewhat rational? I think we can appeal to the idea that a package of assets is often specific to a certain goal. However well the new goal is tailored to the assets, the degree of fit cannot be compared to how the package fits the initial goal. For Michael, majoring in engineering rather than biology can fully utilize all previous courses. Thus, the degree of redemption will be higher, for the whole package of assets efficiently contributes to achieving the goal. Indeed, if the assets are unique — e.g., if Michael took archeology instead of engineering courses, sticking to one’s initial goal is rather rational because finding another goal that utilizes the assets at all can be very difficult. Nevertheless, such decisions might not be all-things-considered rational. We should also consider further costs of continuing the initial pursuit and the potential benefits of changing the goal.