An agent deliberating about what to do may fail to think of some option available to her or actively refuse to entertain some option due to a failure of imagination or because she is in the grip of a groundless, stubbornly entrenched habit of thought. Cases like these look like instances of rational failing, and reflecting upon them, it is tempting to conclude that a disposition to consider all relevant options, so far as limits of time and cognitive resources permit, is a disposition constitutive of rationality in practical deliberation. Recent work has taught us that an agent need not deliberate at all in order to choose in a manner that is responsive to reasons (Railton Reference Railton, Sobel and Wall2009; Arpaly and Schroeder Reference Arpaly and Schroeder2012; Doris Reference Doris2015). Nonetheless, it is natural to suppose that if an agent is going to deliberate, she ought to consider her options to the extent she can—at least insofar as they are not obviously stupid or morally wrong. In what follows, I will argue that these tempting thoughts are mistaken. Indeed, I will argue that an agent may have good reasons not to consider some course of action, Φ-ing, in her deliberation even though Φ-ing is what there is most reason, all things considered, for her to do.
The reasons not to consider a course of action that I will point to are pro tanto reasons, which may be outweighed by competing reasons. In circumstances in which an agent has strong reason not to consider Φ-ing, but in which competing reasons grow stronger over time, the reason not to consider Φ-ing may be decisive at t0, but may come to be outweighed by competing reasons at t1. So, in the context of dynamically shifting reasons, the reason an agent has not to consider Φ-ing will often function, effectively, as a reason to wait before considering it. The reasons not to consider that I identify can thus give deliberation that is responsive to them a distinctive, temporally extended shape.
My argument will proceed as follows. In section 1, I will argue that an expected value framework for thinking about practical reasons can easily accommodate the idea that agents can have good reasons not to consider some courses of action. Considering a course of action is itself an action. Like other actions, it may have benefits and also costs—most obviously, opportunity costs. In some cases, the expected costs of considering some course of action are greater than the expected benefits, and it is reasonable of an agent not to consider it. In section 2, I will argue that in certain circumstances, considering a course of action may have distinctively epistemic opportunity costs, and as a result considering that action may make it more probable that an agent will arrive at a false belief about what she has most reason to do. Reasons stemming from these costs, too, can naturally be accommodated within the framework of expected value. In section 3, however, I will point to a more interesting class of reasons not to consider, which cannot so easily be accommodated within an expected value framework and which serve to put some pressure on it. These reasons stem from the facts that, apart from judging the goods at stake in some decision valuable, an agent may value some of these goods, and that often an agent cannot be compensated in kind for the loss of goods that she values. Where this is the case, I argue, an attitude of conservatism about value is warranted: it is reasonable of an agent to begin her deliberation by considering only ‘no-regrets’ courses of action, which do not involve the loss of important, valued goods for which she cannot be compensated in kind, and to proceed to considering other courses of action only when the cost of continuing to consider only no-regrets solutions has become intolerably high. In section 4, I will argue that the account I have offered of the reasons agents have not to consider some of their options has an important payoff for moral psychology: it can help to illuminate intuitions that play an important role in motivating deontological moral thought, and it can shed light on the complaint that there is something wrong with the way in which schematic thought experiments are frequently used in philosophy to drive moral theorizing.
1. Reasons Not to Consider: The Expected Value Account
As I will use the term, considering a course of action has, as its constitutive end, the discovery of reasons for and against that course of action. Considering courses of action is thus a constitutive part of deliberating about what to do. Merely daydreaming about a course of action, in a way that one takes to have no bearing upon decision or action, is not considering it in my slightly narrow use of the term. One might consider a course of action entirely ‘in one's head’—for instance, by trying to imagine what it would be like to take this course of action or by thinking through what its consequences or preconditions might be. Or one might consider it by talking it over with others or by drawing a diagram on a sheet of paper or by searching for actions of this type with Google. One might self-consciously choose to consider a course of action, or one may just find oneself doing so—and in either case, one may do so responsively to reasons one has to consider the course of action. (For a characterization of deliberation that is consonant with my characterization of considering and upon which my characterization leans, see Arpaly and Schroeder Reference Arpaly and Schroeder2012).
It is easy to offer an informal argument that goes some way toward supporting the idea that a disposition to consider one's options—so far as time, cognitive bandwidth, resources, etc. allow—is a constitutive part of rationality in deliberation.
The Argument for Considering: For some course of action available to an agent, Φ-ing,
(1) Considering Φ-ing is a way for the agent to discover expected costs and expected benefits of Φ-ing.
(2) So, considering Φ-ing is way for the agent to improve her epistemic position with respect to the question of what course of action maximizes expected value (or what courses of action satisfice with respect to expected value).Footnote 1
(3) So, considering Φ-ing increases the probability that the agent will choose a course of action that maximizes (or satisfices with respect to) expected value. This is an expected benefit of considering Φ-ing.
(4) (3) is a good, practical reason for the agent to consider Φ-ing.
The argument is agnostic with regard to the question whether ‘value’ should be understood as something subjective (e.g., individual preference satisfaction, as it figures in rational choice theory) or objective. As I have formulated it, the argument does presuppose that preferences or judgments of value conform to the basic axioms of expected utility theory so that it is possible to construct an index of value, the mathematical expectation of which an agent may maximize or may satisfice with respect to. (For a treatment according to which judgments of objective value conform to these axioms, see Broome [Reference Broome1991].) The argument could easily be formulated without this presupposition, simply by replacing talk of ‘expected costs and benefits’ with talk of ‘reasons for and against’ and by replacing ‘what maximizes (or satisfices with respect to) expected value’ with ‘what there is most reason (or sufficient reason), all things considered, to do’.
One way to resist the conclusion of The Argument for Considering is to put pressure on the generalizations that are its premises, by pointing to kinds of cases in which they fail. I will do this in section 2. For now, it is important to note that even if the argument goes through, the conclusion it warrants is, at most, that agents have good reasons to consider their options, not that this is what they have most reason, all things considered, to do. Considering a given option, whether by inquiring and investigating or merely imagining it ‘in one's head’, is an action, which a person may do intentionally. (Arpaly and Schroeder [Reference Arpaly and Schroeder2012] emphasize an analogous point about deliberation.) Like any action, considering may have not only benefits (in advance, expected benefits), such as the one that The Argument for Considering identifies, but also costs (in advance, expected costs), which count as countervailing reasons not to do it. Most obviously, it may have an (expected) opportunity cost: time and other resources spent considering any given option are time and resources that are not spent doing something else. And so, even where The Argument for Considering goes through and considering a course of action has a positive expected benefit, this may be outweighed by the expected costs of considering.
Here is an easy case:
Cleveland: Akeela and Kim have decided to splurge on a once-in-a-lifetime, bucket-list vacation, and are trying to decide where to go. They are tempted to hike the Incan trail to Machu Pichu, but exploring Rome is also enticing, as is traveling around the historical sites of India. They never consider visiting Cleveland, Ohio.
A partisan of Cleveland might point out that it has many charms of which Akeela and Kim have no inkling, but which they would discover if only they would read up on it. Suppose he is right: if they were to read about Cleveland, they would discover something to interest them there. Be that as it may, it seems clear that Akeela and Kim are reasonable in having ignored Cleveland thus far—and also in only pretending to read the websites and brochures to which their enthusiastic friend now points them. And it seems clear why this is reasonable. Although, we may imagine, the opportunity cost of considering Cleveland is small (they are not short on time), the expected benefit is even smaller. The probability that the attractions of Cleveland they might discover would compare favorably to those of a trip to Rome is miniscule. And so the probability that discovering these attractions would improve their epistemic position with regard to the question of what course of action would maximize expected value, and thus that it would improve the probability that they will take the course of action that maximizes expected value, are comparably miniscule. If Akeela and Kim are open-minded, the subjective probabilities they assign to these possibilities may not quite be zero. Nonetheless, they are likely low enough that, even if The Argument for Considering goes through, the expected benefit it identifies of considering Cleveland will be even smaller than the very small expected cost.
This example suggests a straightforward, initial proposal:
Expected Value: it is reasonable of a person not to consider some course of action, Φ-ing, if the expected benefit of considering Φ-ing is less than or equal to the expected cost of considering Φ-ing.
While Expected Value is formulated as a claim about the reasonableness of failing to consider some course of action, it does not treat considering as a distinctive kind of action, with distinctive kinds of costs and benefits. It would apply just as well to some other kind of action—such as performing an experiment, conducting research, or getting an education—that improves one's epistemic position and has practical value as a result. And it would apply just as well to some kind of action that did not improve one's epistemic position at all, but that was valuable in some entirely different way. In section 2, however, I will identify some costs that are distinctive to considering certain options and that constitute distinctive reasons not to do so.
2. Reasons Not to Consider: Epistemic Considerations
In cases like Cleveland, Expected Value is easy to interpret. The expected benefit of considering a course of action is what The Argument for Considering identifies: considering a course of action increases the probability that an agent will choose the course of action that maximizes expected value. The expected cost is the expected opportunity cost of time spent considering, which might instead have been spent doing something else.
Not all cases, however, are so straightforward. In what follows, I will focus on cases marked by two features: (1) the appropriateness of epistemic humility because of the inaccessibility to consideration of important facts; and (2) the existence of epistemic opportunity costs. The following case has both of these features:
Moving On: When Sandra and Jean got married, they left the city where they lived and moved to a small town. They had recognized that the town needed a grocery store and judged, correctly, that they could make a better livelihood there than in the city. They invested all of their savings and set up a small grocery. Now, a couple of years later, a Walmart has just opened up outside of town, and their business is suddenly losing money. A friend tells them about the booming oil towns of North Dakota and suggests that they could close their current business, move there, and open a new business before their savings dwindle to nothing. Sandra and Jean refuse to consider doing this, for the time being. They recognize, however, that if they cannot find a way to make things work where they are, they might have to move on eventually.
As I have described the case, Sandra and Jean not only refuse to move to North Dakota; they refuse, for the time being, to consider doing so. Intuitively, this refusal does not appear unreasonable. But if we understand the cost of considering moving as a straightforward opportunity cost, as in Cleveland, it is hard to explain this intuition. Given the very real probability that they could lose their savings if they stay, the possible benefit of considering moving on may be quite large. And as I have described the case, the opportunity cost of considering moving on appears small. However, we can identify two sorts of costs to considering moving on, beyond those noted in Cleveland, that explain why this appearance is misleading.
2.1 Appropriate Epistemic Humility and the Mere Exposure Effect
We can stipulate that by considering the possibility of moving on, Sandra and Jean could change their epistemic position with regard to a range of questions about moving on. What is the current business climate in North Dakota? Are banks there offering credit to new businesses? Would they be able to sell their current home for a good price? Nonetheless, Sandra and Jean are cognitively and epistemically limited in ways that considering moving on may do little to remedy. They are limited in what they know about their present circumstances. Perhaps, for instance, there is currently sufficient demand, of which they are unaware, to support other sorts of business in their town, such as a bed-and-breakfast for weekenders from the nearest large city, and they could do well by opening such a business. They are limited in their ability to predict how their future circumstances will unfold. Perhaps their town will soon receive an influx of new people to support their shop, because the town will be discovered as a desirable place for second homes or because an initiative by the state government will succeed in attracting a new industry to the area. Or perhaps the price of oil will plunge, pulling down North Dakota boom towns with it, so that if they move there, they will end up losing their shirts. Thus some of the most important considerations bearing on their choice may be inaccessible to consideration (or inaccessible to consideration at the present time). Even as considering the possibility of moving on would improve their epistemic position with regard to some relevant questions about this course of action, therefore, doing so might not improve at all—or might improve only very little—their ability to form a well-founded estimate of its expected value. And so, while premise (1) of The Argument for Considering might come out true when ‘moving on’ is substituted for Φ-ing, the transition from (1) to (2) might be blocked, or its force might be attenuated: considering moving on might not improve at all (or might improve only a little) their epistemic position with respect to the question of what course of action would maximize expected value. And so the expected benefit of considering it might be zero (or small). Recognizing that this is the case requires the virtue of epistemic humility.
If considering pulling up stakes has little expected benefit because it does not help Sandra and Jean to improve their epistemic position with respect to the question of what course of action would maximize expected value, it might nonetheless seem that the costs of considering this course of action are similarly low. I have already said that the opportunity cost, understood in a normal way (time and energy spent considering could be spent doing something entirely different), seems likely to be low. I will argue in section 2.2 that considering moving on does, in fact, have a distinctive, epistemic opportunity cost. But first I want to note that findings in empirical psychology suggest another possible cost to considering in cases where epistemic humility is appropriate. One of the most well-replicated psychological findings of recent decades is the mere exposure effect, first identified by Zajonc (Reference Zajonc1968). Repeated mere exposure to an arbitrary stimulus (such as a word, a Chinese character, or a picture of a face) has been shown to generate liking for this stimulus. This has been confirmed not only in laboratory contexts, but also ‘in the field’, including in social and interpersonal contexts, and it is a cornerstone of modern advertising practice. These findings suggest the following hypothesis, which is just a specification of the more general ‘familiarity-leads-to-liking hypothesis’ (Harrison Reference Harrison and Berkowitz1977): attending to a course of action over time by repeatedly considering it, its possible consequences, and so on, might bias an agent in favor of that course of action—even in cases where considering it does not, in fact, give the agent good reasons to increase her estimation of its expected value.
We can see how this might work in a case like Moving On. I have stipulated that the most important considerations bearing on this course of action are inaccessible to consideration, so that considering this course of action will not improve Sandra's and Jean's epistemic position with regard to the question: ‘what course of action maximizes expected value?’ or will improve it only a little. Nonetheless, the familiarity-leads-to-liking hypothesis supports an expectation that I think common experience already makes plausible: in thinking through those aspects of moving on that they can think through, finding the information that they can find, doing their best to imagine what it would be like to move on, Sandra and Jean might, as we say, accustom themselves to the idea. The idea might come to have, for them, the false feel of familiarity, so that choosing it might come to feel like choosing a known option—even as its most significant features remain unknown. This is a form of irrationality, and it can attend any deliberation. But it seems most likely to pose a danger in circumstances where epistemic humility is appropriate: a bias like this is most likely to prove decisive precisely where decisive reasons are absent.
The moral is: in circumstances where the most important considerations bearing on some option are inaccessible to consideration, so that considering that option cannot improve an agent's epistemic position with regard to its expected value, considering it may nonetheless increase the probability that the agent will choose that option. Whether considering has a positive or negative expected value will then depend on whether the course of action itself happens to have positive or negative value. If the course of action in fact has less expected value than other courses of action available, then considering it will increase the probability that the agent will do something other than what maximizes expected value—contrary to (3) of The Argument for Considering.
2.2 Epistemic Opportunity Cost
So far, I have argued that considering moving may not improve Jean's and Sandra's epistemic position with respect to the question ‘what course of action maximizes expected value?’ and may decrease the probability that they will do what maximizes expected value. I want now to argue that considering moving on may, in fact, worsen their epistemic position with respect to the question of what maximizes expected value because considering this possibility has an epistemic opportunity cost.
Desirable courses of action, such as supplementing their income by starting a bed-and-breakfast to cater to weekenders from a nearby city, may initially be invisible to Sandra and Jean. Discovering them may take work—not only time spent reading, inquiring, talking things over with others, and otherwise investigating, but creative, imaginative effort as well. Time, attention, cognitive bandwidth, and emotional bandwidth are scarce resources. In trying to imagine, investigate, and talk and think through what might happen and what things might be like if they were to move on, Sandra and Jean are, of necessity, using resources that might be deployed in other epistemic efforts. This can be stated as a positive point: treating moving on as ineligible for consideration and hence ‘off the table’, for the time being, can serve to force their attention to dwell on what is still on the table and to spur their imagination, so that they find there novel, creative, hitherto invisible solutions to their problem—such as opening a bed and breakfast. And so considering moving on may prevent Sandra and Jean from improving their epistemic position with respect to the question of what course of action maximizes expected value in a way that they would be able to improve their position if they did not consider this option. (In this respect, the epistemic opportunity cost discussed here is a converse of the opportunity cost discussed with respect to Cleveland. In that case, there was a very small epistemic benefit to considering Cleveland, but it was outweighed by the nonepistemic opportunity cost of doing so.)
It is easy to imagine Sandra and Jean using such phrases as ‘we'll find a way’ and ‘it will all work out’ to explain their refusal to consider pulling up stakes. And it is easy to hear in this explanation a kind of double irrationality: not only do they impoverish themselves epistemically by wearing self-imposed blinders, but they also justify doing so by appeal to an optimistic assessment of their predicament that looks suspiciously like wishful thinking. If what I have argued is correct, however, then ‘we'll find a way’ may express not just an epistemic stance but a practical one. In putting on blinders, they direct their (finite) attention to one region of the space of possible solutions rather than another—rather than ceasing to attend to their problem and its possible solutions, as the charge of irrationality presupposes.
Even if ‘we'll find a way’ is partly a practical stance, however, it expresses an epistemic attitude as well, and that attitude may, with time, prove to be unjustified—and with it so may the practical stance. At some point, as time progresses and their savings dwindle, as the prospect of being unable to keep the store open becomes imminent, and as their best, most strenuous efforts to find an alternative way to make money in town continue to come up empty, holding on to the idea that a solution will magically present itself may really begin to look like a denial of reality. As it becomes less and less likely that further efforts will yield an alternative solution, the epistemic opportunity cost of considering moving on diminishes, and the risk that they will be led astray by exposure effects may become comparatively less significant. At the same time, as their savings diminish, and the probability increases that by doing nothing they will lose their livelihood and their savings entirely, the reasons to consider moving on may grow stronger—until this is what they have most reason, all things considered, to do.
3. Complicating the Picture: Valued Goods and Non-compensable Losses
In Moving On, as I presented it, the expected costs and expected benefits of moving to North Dakota and of waiting are monetary, and hence they are clearly commensurable with one another; and so the expected costs and benefits of considering this course of action are also commensurable. Considerations about the appropriateness of epistemic humility (section 2.1) and about epistemic opportunity costs (section 2.2) complicate the assessment of the costs and benefits of considering moving on, but they do not challenge the framework laid out in Expected Value for assessing whether it is reasonable to consider this course of action. Here is a case that is structurally similar to Moving On in obvious ways, but in which there are reasons not to consider a course of action that do not fit comfortably within Expected Value.
Pulling Up Stakes: Sarah and James own a grocery store in an economically struggling small town where their families have lived for generations. They are pillars of the community, and their lives are tightly woven into the social and civic fabric of the town. A Walmart has recently opened outside of town, and their business is suddenly losing money. A friend tells them about the booming oil towns of North Dakota and suggests that they could close their current business, move there, and open a new business before their savings dwindle to nothing. James and Sarah refuse to consider doing this, for the time being. They recognize, however, that if they cannot find a way to make things work in their town, they might have to pull up stakes eventually.
I will argue that the nature of the goods that are salient in Pulling Up Stakes and of the agents’ relations to them mean that these goods provide reasons, different from the costs of considering evident in Moving On, not to consider pulling up stakes and moving to North Dakota. In order to see how these goods constitute reasons not to consider pulling up stakes, it will help first to examine the reasons the goods provide not to pull up stakes in fact.
3.1 Valued Goods as Reasons Not to Pull up Stakes
Consider some of the goods that Sarah and James stand to lose if they decide to pull up stakes. Sarah and James have many close, personal relationships in the town, forged over decades from childhood on. They each play key roles in various civic organizations (the library board, the local Kiwanis, their church, the town board of supervisors,…). They know that if they close their grocery, it is unlikely to be replaced, and this will make it much more difficult for the remaining businesses in the town's struggling downtown to stay afloat. And they enjoy the respect, earned over a lifetime, of others in their town.
These goods are, on their face, heterogenous. Some of them, such as the respect they enjoy, are goods only for Sarah and James, whereas others are also goods for others or for the town. Perhaps all of these are things that everyone ought to recognize as in some way valuable. Nonetheless, it seems clear that at least some of these goods, and perhaps all of them, are things that James and Sarah are not required, rationally or morally, to value Footnote 2 or to care about.Footnote 3 Others in a similar position who found that they did not care about these things, but were instead excited by the prospect of starting anew in another state, would not, I think, be open to rational or moral criticism.Footnote 4 The world is full of goods, and it is impossible to value all of the things one judges to be valuable while remaining sane. Nonetheless, we can imagine that Sarah and James do value each of these goods, and so that they will feel the loss of these goods as losses to them. And even though such valuing is not required, it seems clearly to be permitted: these things are valuable, and so they are the sorts of things that are worthy of or eligible for valuing.
Suppose that the prospective losses of goods that Sarah and James value constitute good, pro tanto reasons for them not to pull up stakes. These might, nonetheless, be losses that, all things considered, they have most reason to accept, given the alternatives. But whatever advantages of pulling up stakes make this the case (preserving their savings for their retirement and for their children, for instance), it seems clear these advantages will not compensate Sarah and James in kind (Wiggins Reference D and Wiggins1998a: 259–60; Wiggins Reference Wiggins and Wiggins1998b) for the losses of these valued goods. (I will call goods for the loss of which an agent cannot be compensated in kind non-compensable goods, for short. A good is compensable in kind, or not, for someone, and the same good may be compensable in kind for one person but not for another. For ease of exposition, I omit explicit relativization going forward.) And so, even if it is the right decision, the decision to pull up stakes is nothing like the decision to give up $100 in exchange for $200 (Griffin Reference Griffin and Chang1997: 36). In the latter case (absent some peculiar context), it would be irrational to value the particular $100 bill that one gives up, and so to regret its loss; what is valuable about the $100 bill is simply its monetary value, and by exchanging the bills one gets back the same kind of value that one gives up. But there is nothing irrational about Sarah and James valuing the goods lost in the case of Pulling Up Stakes, and hence regretting their loss, even if accepting their loss is what they have most reason, all things considered, to do. The reason is evident: even if the goods they gain are more valuable than the goods they lose, on some all-things-considered scale, these goods are different and speak to different concerns of Sarah and James.
3.2 Non-compensability as Grounds for Conservatism of Presently Valued Goods over Future, Not-yet-valued Goods
The idea that some of the things we value are non-compensable goods for us is utterly familiar in discussions of love and attachment. It is a platitude, within and outside of philosophy, that one is not compensated for the loss of a child or a spouse by acquiring a new one of equal value.Footnote 5 We love and are emotionally vulnerable to particular persons. It is not irrational of a parent to mourn the loss of a child (whether the ‘loss’ consists in the child's death or merely in the child's being given up for adoption to a loving family), even when she is ‘replaced’ by a newborn sibling (or twins!), because it is not irrational of a parent to love a particular child, and emotional vulnerability to a particular, loved object is constitutively involved in loving it.
A corollary is that it can be rational to hold a conservative preference toward protecting the goods one has and values over securing new goods.Footnote 6 An agent may judge that prospective goods are just as valuable as or more valuable than those she has, but insofar as she values the former and not the latter and insofar as it is not irrational of her to do so, she has good, pro tanto reasons to resist the exchange. Just as it is not irrational to mourn the loss of the child one has and loves, it is not irrational to resist exchanging that child for two, future children each of whom one knows one will love just as much.
The goods that Sarah and James value and that they stand to lose by pulling up stakes are, in important respects, similar to the particular persons that are objects of attachments. They value particular friendships with particular histories. They value respect from others that they have earned in particular ways. And they care about the welfare of a particular town. In each case, their emotional vulnerability to these goods seems to be constitutively involved in their valuing them.Footnote 7 And in each case, the claim that these goods are not compensable to them in kind can be cashed out, at least in part, as the claim that it is not irrational of them to regret the loss of these particular goods, even if something of equal or greater value is in the offing, and so it is not irrational of them to be conservative with respect to these goods. Even if they are the sorts of people who easily develop close friendships, they may resist exchanging the friendships they have for the future ones they would likely develop in North Dakota.
3.3 Conservatism with Regard to Sacrificing Valued Goods for the Sake of Other Valued Goods
In Pulling Up Stakes, Sarah and James exhibit a conservatism different from and on its face more puzzling than the sort of conservatism that I have so far defended. The choice that Sarah and James face is not whether to sacrifice non-compensable goods they already have and value in order to secure possible future (and thus not yet valued) goods. The choice they face is, rather, whether to sacrifice goods that they already have and value, by pulling up stakes, in order to protect other goods that they already have and value (such as the prospects of a secure retirement and an inheritance for the children) against a possible threat. This is what they are unwilling to consider doing, for the time being. How could this be rational?
We can begin by noting two conceivable explanations for Sarah's and James's conservatism, both of which would make it difficult to defend their conservatism as rational. (1) It could be the case that Sarah and James are unwilling to sacrifice the valued goods in question for others under any circumstances—perhaps because they regard the different goods as simply too different to trade off against one another at all. Or, (2) it could be the case that they are unwilling to make a certain sacrifice of valued goods now in order to protect themselves against merely probable future losses of valued goods.
It is, however, easy to imagine their story continuing in a way that looks rational on its face and that is incompatible with each of the explanations above. Suppose that, after a year of seeking energetically and imaginatively for a way to make things work in town, Sarah and James have found nothing and their business is losing money at an accelerating rate so that there is a strong likelihood that they will lose the entirety of their savings soon if they do not act. When push comes to shove in this way, Sarah and James may be willing to exchange the goods involved in staying in their town for the goods they secure by pulling up stakes. Even though they will not be compensated in kind for the loss of the former goods and so will have reasons for regret, they are able to decide that the latter goods are more important to them, all things considered, than the former. On this continuation of the story, Sarah and James are evidently able to commensurate these goods, at least for the purpose of this decision (Millgram Reference Millgram and Chang1997). And they evidently are willing to accept a certain loss of valued goods as the cost of protecting against a loss of other valued goods that is still merely probable.
A more adequate explanation of Sarah's and James's conservatism begins from the fact that, while they are able to compare valued goods and trade them off against one another if they must, there is one possible outcome—finding a way to make things work in town—that would not require them to make this trade-off. This would be a no-regrets outcome, in which no important, valued goods are lost. Plausibly, the same facts that would give them reasons to regret the loss of valued goods also give them (pro tanto) reasons to privilege such a no-regrets outcome in their deliberation and to try to arrive at it if it might be available to them.
The reasons to try for no regrets that non-compensable, valued goods give to Sarah and James interact with the epistemic considerations we noted in discussing Moving On. As with Sandra and Jean, Sarah and James are not in a good position to know whether the outcome that would be from their point of view optimal is actually available to them. As in Moving On, there may be epistemic opportunity costs to considering moving to North Dakota, and epistemic humility may be needed in order to counter bias arising from exposure effects. So, like Sandra and Jean, Sarah and James have reasons to wait and seek to improve their epistemic position regarding options other than moving to North Dakota, as long as the cost of trying this is not inordinately high. Generalizing these epistemic considerations and taking them together with the conclusions we have already reached about the reasons that non-compensable goods give agents to try for a no-regrets solution, we get the following, rough principle:
Try for No Regrets First: In confronting a practical problem, agents have pro tanto reasons to begin by seeking a solution that does not sacrifice important, non-compensable, valued goods (a ‘no-regrets solution’), and to consider solutions that do sacrifice important, non-compensable, valued goods only when the cost of seeking a no-regrets solution is intolerably high.
(A more precise formulation would express the proposition that each valued good that an agent stands to lose gives the agent a distinct reason to seek a solution that does not sacrifice that particular good.)
This principle makes an existential claim about reasons, but it makes no claim about what an agent has most reason, all things considered, to do. That depends, of course, both on the strength of the reasons that Try for No Regrets First identifies and on the strength of competing (and, possibly, undermining) reasons. Plausibly, the strength of the reasons an agent has to try for no regrets varies with the importance for the agent of the valued goods that she stands to lose. Goods might be non-compensable to an agent and yet not very important to her; where that is so, readily parting with some is a mark of maturity. By contrast, the goods that Sarah and James stand to lose are important to them, and it does not seem immature of them to work hard to find a way to keep them all. But even if these goods give Sarah and James strong reasons to try for a no-regrets solution, it is easy to see how these reasons can be outweighed. Perhaps, at the time when they learn about the possibility of moving to North Dakota, their savings are already dwindling so rapidly that if they do not act quickly, they risk losing everything—and so they have very strong reasons to consider this possibility immediately.
A virtue of Try for No Regrets First is that it predicts a distinctive, temporally extended shape that good practical deliberation often takes. As I have described them, Sarah and James are unwilling to consider moving to North Dakota when their friend suggests it to them—but they recognize that if they cannot find a way to make things work, they might have to consider that option eventually. Suppose that, having refused to consider pulling up stakes, they continue to search energetically for a no-regrets solution to their problem and continue to come up empty. This takes time, and during this time, their savings dwindle, and the prospect of losing everything becomes more immanent and more probable. That is to say, during the time that they are trying for a no-regrets solution, the reasons to consider pulling up stakes grow systematically stronger—until at some point they outweigh the reasons Sarah and James have for waiting. At that point, suppose that Sarah and James respond to the reasons they have by considering the possibility of pulling up stakes and then deciding to do it. Given the reasons that Try for No Regrets First identifies, this deliberative pattern appears eminently rational—more rational than simply considering and deciding to pull up stakes immediately without looking hard for a no-regrets solution would have been. And this is true even if, as it turns out, pulling up stakes is in fact what they have most reason to do. It is easy to multiply plausible examples in which recognizably rational deliberation has this temporal shape, because it is easy to find real-life examples in which the reasons that an agent has to sacrifice some valued goods in order to avert a possible disaster grow systematically stronger over time.
3.4 Reasons to Try for No Regrets First Are Not Captured in Expected Value
Try for No Regrets First might seem unobjectionable but unilluminating because it might seem that it does not move beyond the sorts of considerations we identified in Moving On, which can be interpreted naturally within the Expected Value framework. We already know that, when push comes to shove, Sarah and James are willing and able to commensurate non-compensable, valued goods. And we already know that the fact that an outcome involves the loss of a valued good contributes to its disvalue for them, so that they will regard a no-regrets solution as optimal if it is available to them. And so it might seem that in Pulling Up Stakes as in Moving On we can characterize the reasons for and against considering courses of action such as moving to North Dakota simply in terms of their expected benefits and expected costs, measured on whatever scale of value Sarah's and James's push-comes-to-shove decision reflects. Given all this, it might seem that Try for No Regrets First simply articulates a plausible though imprecise theorem, derivable from the more basic principle, Expected Value. (The comparative content of the more basic principle would seem to be preserved, albeit vaguely, in the theorem's language of ‘important’, non-compensable, valued goods and of an ‘intolerably high’ cost of continuing to seek a no-regrets solution.)
The argument that Try for No Regrets First is a theorem of Expected Value rests, however, upon a false premise. The fact that the values of the goods at stake in a decision can be commensurated for the purpose of weighing them against one another and making a decision does not mean that the commensurated values capture or exhaust all of the reasons that these goods give to an agent. We are now in a position to see that, where the goods in question are valued goods, this will often not be the case. When push comes to shove, a parent might be able to make a tragic, Sophie's Choice-style decision—sacrificing one child, for instance, rather than risking a high probability that two will be killed. That is to say, when push comes to shove, she can commensurate the value of her children's lives and conclude that two are more important than (even: twice as valuable as) one. But any such commensuration, of necessity, abstracts from the fact, which makes her children's lives non-compensable goods for her, that she values each of her children as particulars. And the fact that any such commensuration will not capture everything that matters to her gives her strong reason to want not to be in a deliberative situation that forces her to commensurate the value of her children's lives in this way. It is this reason to forestall a circumstance in which commensurating abstraction is necessary that Try for No Regrets First identifies.
‘Maximize expected utility’ is a criterion for a successful decision made at a moment—even where waiting is one of the possible courses of action to be decided among. Try for No Regrets First, by contrast, only makes sense as a prescription for a temporally extended pattern of deliberative activity. Once agents have heeded the counsel of Try for No Regrets First and have found no no-regrets solution to their problem, and once the costs of waiting have become intolerably high, the principle that a successful decision maximizes expected utility may then become the appropriate criterion by which to judge their decision. In invoking this latter principle in the context of non-compensable goods, however, something is lost.
Try for No Regrets First depends for its force upon the sorts of epistemic limitations discussed with respect to Moving On—limitations that can often be mitigated by waiting. If there were no chance that waiting would improve agents’ epistemic position and so open up the possibility of new solutions, there would be no use forestalling a decision that involved commensurating valued goods. But realistic cases like Pulling Up Stakes and Moving On remind us that epistemic limitations are ubiquitous, and that time (and focused deliberative effort during that time) is sometimes a remedy. Pretheoretical norms of rationality recognize these facts—as when they license the common-sense judgment that it is not irrational of Sarah and James to refuse, for a time, to consider pulling up stakes.Footnote 8
4. An Application: Moral Reasons Not to Consider a Course of Action
Beyond its inherent interest, the idea that I have been developing and defending—that agents can have reasons to limit the scope of their deliberation, for a time, by ignoring some of the options available to them—has a payoff in moral psychology. It can help to illuminate some intuitions that seem to play an important role in motivating deontological moral thought, without recoiling into an implausible deontological absolutism. And it can help make sense of the thought, memorably voiced by Elizabeth Anscombe (Reference Anscombe1958), that there is something wrong with many of the thought experiments that drive philosophical theorizing on these matters. My discussion is, inevitably, sketchy and cursory, especially in its characterization of the views that it brushes aside as inadequate. My goal in this brief, concluding section is not to discuss these views fairly, but simply to indicate that the account I have offered gives us resources for a new approach to an old set of problems.
Anscombe remarked that ‘if someone really thinks, in advance, that it is open to question whether such an action as procuring the judicial execution of the innocent should be quite excluded from consideration—I do not want to argue with him; he shows a corrupt mind’ (Reference Anscombe1958: 17). For Anscombe, this provocative remark is underwritten by absolute prohibitions: asking whether procuring the judicial prosecution of the innocent should be excluded from consideration is evidence of a corrupt mind, she thinks, because procuring the judicial prosecution of the innocent is something one ought never to do. (In Anscombe's preferred idiom, we are meant to forswear the use of ‘ought’ and instead say simply that procuring the judicial prosecution of the innocent is always unjust.) In Anscombe's own thought, these absolute prohibitions seem, in turn, to be underwritten by divine promises (cf. Lovibond Reference Lovibond2004 for a subtle discussion). For many of us, however, a moral view underwritten by divine promises is simply not an option. And without those promises, it is difficult to maintain confidence in absolute prohibitions as the stakes become very high. Is it really the case that it is always wrong to kill one in order to save 20 or 100 or 1000 or to prevent ‘hydrogen bomb war’ (Anscombe Reference Anscombe1958:17)? Compromise positions, such as ‘threshold deontology’, according to which consequentialist thought becomes appropriate only once the consequences of conforming to prohibitions become sufficiently disastrous, seem unstable and hard to motivate (Alexander Reference Alexander2000). And yet, the lingering feeling remains: a consequentialist response to cases like this leaves much that is morally important out of account,Footnote 9 and the fault for this may lie, somehow, in the examples themselves.
We can make a start at bringing the thoughts of the last section to bear by noting that for a morally decent agent facing the sort of moral dilemma that is a staple of twentieth- and twenty-first century philosophy, important, valued goods are at stake. Perhaps, in a horrendous enough case, lying, killing, cheating, stealing, or torturing might be the right thing to do. But even if that is so, an agent who does these things must live with the knowledge that she is now a killer or a liar or torturer; and as Hume pointed out, ‘the invaluable enjoyment of [her own] character’ is a good that a decent person may value, for whose loss she cannot be compensated in kind (Hume Reference Hume, Selby-Bigge and Nidditch1975: 283).Footnote 10 In an awful enough situation, perhaps this is a price worth paying; indeed, in circumstances where others’ lives, for instance, are at stake, perhaps this price to the agent ought not to weigh in her deliberation at all. But it is, nonetheless, a source of not-irrational regret.
In many real-world versions of these moral dilemmas, the considerations discussed in section 2 regarding appropriate epistemic humility and epistemic opportunity costs will also apply. Where this is the case, so will Try for No Regrets First: decent agents will have good reasons to want to forestall considering morally horrible courses of action. (In what may be the original, real-world lifeboat case, R v. Dudley and Stephens [Reference Dudley and Stephens1884], the protagonists could have avoided a morally horrible act had they only waited longer.) Whether it is reasonable to forestall consideration, and for how long, will depend, again, on the costs of waiting—including distinctively moral costs. When the moral costs of trying for a no-regrets solution become intolerable, an agent then faces a moral dilemma with the structure of those that populate textbooks of moral philosophy. At that point, considering all of her options, she might decide that what she has most reason to do, all things considered, is to take a morally horrible course of action, such as lying or killing.
The picture has an obvious similarity to standard threshold deontological accounts. As in a standard account, an agent's choices are constrained by deontic requirements so long as the badness of the consequences of obeying those requirements is below some threshold. But whereas in standard threshold deontology the threshold concerns the badness of, for instance, refusing to lie or cheat or steal, on the view I have been arguing for, the threshold concerns the badness of continuing to look for a no-regrets solution. And so whereas standard, threshold deontology describes a static, yes-or-no decision at a moment (obey the deontic requirements or violate them), the view for which I am arguing describes how morally decent deliberation and decision may unfold over time with a distinctive temporal shape. Unless the costs of looking for a no-regrets solution are intolerably high at the start, an agent will begin by allowing her thought to be bounded by deontic constraints and will eventually step over these constraints only if she fails to find a no-regrets solution before the costs of continuing to look grow intolerably high.
While the account that I have just given applies to many real-world moral dilemmas, it seems to shed no light on the dilemmas of philosophy textbooks precisely because those cases are stipulated so that considerations of appropriate epistemic humility and epistemic opportunity costs are not relevant. Such stipulation is, of course, a necessary feature of these examples if they are to uncover the structures of moral thought they are meant to uncover. I do not want to suggest with Anscombe that we could or should do without such examples. But we are now in a position to see that her complaint against these stipulated examples, like the complaints of countless undergraduates in introductory philosophy courses, has merit. In abstracting from the features of real-life examples that make epistemic humility appropriate and from features that that introduce epistemic opportunity costs, these examples illuminate certain aspects of the structure of moral thought at the expense of others. I do not think that this shows a corrupt mind, and I do not think that we should stop considering such examples. (We do need to understand the aspects of structure they illuminate). But we should do so with a keen awareness of what they leave out and of the structural features of rational practical deliberation that these omissions obscure. To fail to do so is to risk corrupting the mind, in just the way that Anscombe identified:
The point of considering hypothetical situations, perhaps very improbable ones, seems to be to elicit from yourself or someone else a hypothetical decision to do something of a bad kind. I don't doubt this has the effect of predisposing people—who will never get into the situations for which they have made hypothetical choices—to consent to similar bad actions, or to praise and flatter those imagined as doing so . . . when the desperate circumstances imagined don't hold at all. (Anscombe Reference Anscombe1958: 13)
