1) Introduction

Most of us are trying to do the right thing—we want to act morally. But figuring out what morality calls for is challenging. Our moral judgments are fallible, and we are often uncertain. This can happen because we do not know the relevant descriptive facts—for example, whether an item of clothing was made in a sweatshop or whether a grieving friend would find a phone call comforting as opposed to intrusive. But we can also face difficulties figuring out relevant moral norms, even when the descriptive facts are clear. For example, someone may be unsure whether it’s okay to loyally conceal a friend’s wrongdoing to protect them from serious penalties or, in a case I discuss below, a person may consider whether to break a promise of confidentiality in order get help for a friend struggling with addiction.

This paper argues that being reckless when navigating our moral uncertainty and the fallibility of our moral judgments is, itself, morally wrong. (When I discuss moral uncertainty, here, I’m referring to uncertainty about what is moral to do, which cannot be resolved by settling descriptive questions.) Avoiding this recklessness could include simple steps like steering clear of a morally questionable activity just in case it’s wrong or taking actions that reflect careful thought, not drunken impulses.

This much might seem intuitive. But it has become controversial in recent work. A debate has emerged about whether what we should do is sensitive to what I’ll call our moral-epistemic circumstances: our beliefs, credences, and evidence regarding moral norms. This includes disagreement about whether what we should do depends on factors like our moral uncertainty. One camp thinks that what we should do is sensitive to our moral-epistemic circumstances (Guerrero, Reference Guerrero2007; Lockhart, Reference Lockhart2000; MacAskill, et al. Reference MacAskill, Bykvist and Ord2020; Moller, Reference Moller2011; Sepielli Reference Sepielli and Shafer-Landau2009). The other camp rejects this and says that what we should do is independent of our moral-epistemic circumstances—that what is right for us to do can be settled without peering into our heads (Harman, Reference Harman and Shafer-Landau2015; Weatherson, Reference Weatherson2019). Following Elizabeth Harman, I’ll call the first camp “uncertaintists” and discuss the second camp as critics of uncertaintism.Footnote ¹ But a variety of terms have been used for similar ideas. Prominently, Brian Weatherson refers to the first group as a type of normative internalist and the second group as a type of normative externalist (Reference Weatherson2019). I’ll treat “uncertaintist” as largely synonymous with Weatherson’s “internalist” (I adopt the first term mainly to avoid confusion with debates about motivational internalism in ethics). While uncertaintism gives moral uncertainty the prominence in its name, moral uncertainty is only one of a variety of important moral-epistemic circumstances that could matter to what we morally ought to do—and I’ll take the uncertaintist thesis to be broad enough to encompass others (e.g., being confident in a moral view). What’s key is that the uncertaintist believes that some combination of our beliefs, credences, and evidence regarding morality matters to what we morally ought to do.

The literature on this issue is in a peculiar place. On the one hand, some of the most prominent criticisms of uncertaintism have targeted a version of uncertaintism that says that our moral-epistemic circumstances impact what we morally ought to do (Harman, Reference Harman and Shafer-Landau2015; Weatherson, Reference Weatherson2014, Reference Weatherson2019). On the other hand, there has been surprisingly little direct defense of this type of uncertaintism. Most arguments that our epistemic circumstances matter to what we morally ought to do have focused on credences or evidence about morally relevant descriptive facts, rather than credences or evidence about moral norms.Footnote ² At the same time, there is extensive literature about the normative implications of uncertainty about moral norms, but in most cases, these authors either explicitly defend theses about what we rationally ought to do in the face of moral uncertainty (Lockhart, Reference Lockhart2000; Ross, Reference Ross2006; Sepielli, Reference Sepielli and Shafer-Landau2009; Tarsney, Reference Tarsney2018) or leave open whether the relevant norms are rational or moral (MacAskill, Reference MacAskill2014) (the exceptions tend to focus on narrower questions (Guerrero, Reference Guerrero2007; Moller, Reference Moller2011) or do not take the moral character of these norms as a central claim to be argued for).Footnote ³ This paper fills that gap, defending uncertaintism as an explicitly moral claim: our moral-epistemic circumstances matter to what we morally ought to do—and failing to take them into account appropriately is, itself, morally bad (not only a failure of rationality). I go on to develop a type of uncertaintism that can avoid many of the leading criticisms.

The paper begins, in Section 2, by offering two common sense arguments for the claim that our moral-epistemic circumstances matter to what we morally ought to do. First, acting with respect toward those who could be victims of our wrongdoing requires that we are careful, rather than reckless, as we navigate our moral uncertainty and the possibility that our moral judgments could be mistaken. Second, if we accept commonplace requirements to prepare well for morally important decisions (with, say, careful forethought rather than drunken impulses), then the moral credences we form through those preparations must also matter to what we should eventually do. Without this, it’s unclear why we should bother with good preparations.

Section 3 of the paper addresses some key criticisms of uncertaintism—remember, critics have especially targeted this variety of uncertaintism that says that our moral-epistemic circumstances impact what we morally ought to do. I argue that a lot of worries about this view are really problems for uncertaintist theories with particular features—features that we can reject without rejecting the core idea of uncertaintism. In particular, influential objections to uncertaintism have tended to suppose that uncertaintists are committed to a series of controversial claims: that moral ignorance is exculpatory, that we should take our moral-epistemic circumstances into account in specific ways involving expectation maximization, and that all moral norms are sensitive to our moral-epistemic circumstances—not just some of them. I outline how uncertaintists can avoid each of these three commitments and why there is often independent reason to think that the best versions of uncertaintism will do so.

In Section 4, I offer a positive account: I defend a framework for our obligations in the face of moral uncertainty and fallible moral judgments that avoids the critics’ challenges, developing in particular this third idea—that moral-epistemic circumstances can matter for some moral norms without mattering for all of them. I argue that we are subject to both procedural moral norms that are sensitive to our moral-epistemic circumstances (analogous to the due process norms of a legal system) and substantive norms that do not need to be sensitive to them (analogous to norms of substantive justice). Procedural norms, on this account, are moral norms about how we should go about trying to satisfy other moral norms given that we face challenges in discovering their content—for example, our uncertainty and fallible judgment. For example, we would be endorsing procedural norms if we suggested that people should “act based on careful moral contemplation” or “avoid actions whose moral rightness is unclear, if they can be avoided at low cost.” In contrast, substantive oughts are ordinary, first-order moral norms that are not focused on the moral strategies to use when trying to comply with other moral norms (e.g., “don’t lie” and “maximize utility” could be substantive oughts).

The arguments I develop have three large implications. First, they suggest that many of the debates about moral uncertaintism have been distorted by a focus on extreme options: either moral-epistemic circumstances always matter to what we morally ought to do or they never matter to what we morally ought to do. But if I’m right, there’s a fruitful middle ground that has been largely ignored, in which moral norms are sometimes sensitive to our moral-epistemic circumstances. And many worries about uncertaintism can be seen, instead, as reasons to reject an extreme uncertaintism in favor of a more moderate form of uncertaintism, rather than counseling for giving up uncertaintism entirely.

A second set of implications follows from my argument that it’s morally bad to be reckless in the face of moral uncertainty—not just a problem of rationality.Footnote ⁴ I’ll leave open here what distinguishes moral norms from other norms (e.g., whether it is something about their gravity or their structure, etc.). But it’s far from a minor, terminological difference. Whether a norm is moral has potential implications for how we think about blame, for instance (we may be to blame for procedural recklessness, itself, even if this does not alter blame for the substantive rightness and wrongness of our actions). And if we morally ought to navigate moral uncertainty in particular ways or using particular strategies, then these responsibilities present a wide swath of underexplored moral norms. Taking a closer look at them may shed light on other moral questions. For example, moral norms calling for interpersonal toleration might be best explained as good strategies for being moral when facing disagreement and uncertainty (Rosenthal, Reference Rosenthal2019). But explanations like this require that using good strategies for being moral is something that we morally ought to do.

Finally, crucially, understanding these strategic responsibilities as moral norms has a large impact on how we determine their content. When navigating moral uncertainty, we can—and should—ask, for example, which risks of wrongdoing are compatible with respecting others, and which personal gains we should be willing to sacrifice in order to manifest that respect. It’s not only that we have strategic responsibilities that happen to be moral—it’s that thinking about these responsibilities’ content and demands will involve engaging directly with moral ideas and values.

2) Common Sense Reasons for Uncertaintism

To see why our moral-epistemic circumstances matter to what we morally ought to do, let’s consider what it takes to treat others with respect when we find ourselves unsure what’s morally called for, or simply aware that our moral judgments might be flawed. This can come up when people are deciding whether to turn in a friend for cheating on an exam, for example, or when they must make end-of-life decisions on behalf of an aging parent. I’ll discuss in more detail a case of someone who considers breaking a promise of confidentiality in order to get help for a friend. Geraldine’s friend, Ellen, wants to confide in her about recent difficult experiences, but first, Ellen asks Geraldine to promise to keep their conversation a secret. Geraldine agrees, and Ellen reveals that she has been using drugs again after a period of sobriety. Ellen has a history of struggles with addiction, and Geraldine wonders whether it would be moral for her to break her promise of confidentiality in order to involve people who might be able to help her friend.Footnote ⁵

I will not try to provide an answer for Geraldine here. Instead, I’ll start by suggesting something, perhaps obvious, about the way she should approach the decision: she should not avoid thinking about the uncomfortable moral questions involved by, instead, deferring to her pet parrot’s advice. We can picture Geraldine asking her parrot, Polly, “should I break my promise to Ellen?” and, having had a long day, figuring that Polly’s answer is good enough. But I’ll assume that ordinary adult humans, with all our flaws, are still better equipped to think through humans’ moral problem than are parrots—who likely do not understand the complexities at stake. So, by trusting Polly rather than engaging in her own careful consideration of the relevant issues, Geraldine seems to be behaving recklessly. She is facing a difficult, weighty moral decision. Making the decision based on a quick chat with a parrot would be morally bad—even if the parrot’s answer accidentally led to a good course of action.Footnote ⁶ Taking action with so little contemplation and attention would be treating her friend, Ellen, in a morally objectionable way.

This simple guidance—do not rely on a parrot’s moral advice—helps illustrate two common sense reasons that our moral-epistemic circumstances sometimes matter to what we morally ought to do. First, and most straightforwardly, behaving with respect toward others includes recognizing that they can make moral claims on us, and that mistreating them is something that we should care enough to try to avoid (even knowing there are no options that guarantee this). So, taking morally important actions based on haphazard or reckless processes involves acting disrespectfully toward the people who will be impacted—if Geraldine’s decision was based on a parrot’s advice, her friend Ellen could reasonably complain that she wasn’t given morally important forms of respect (even if Geraldine’s decision was, as a matter of luck, a good one).

But we can only say that it’s morally bad for Geraldine to rely on Polly’s answer, if moral-epistemic circumstances matter to what we morally ought to do. Criticizing Geraldine for following her parrot’s guidance involves criticizing her based on the moral-epistemic circumstances that led to her actions—the credences she formed following her chat with the parrot. If Geraldine’s moral-epistemic circumstances were irrelevant, then it would only matter whether she kept her promise of confidentiality or not—keeping the promise would mean she has done what she morally ought to do (if, let’s suppose, keeping her promise was right), even if she simply followed her parrot’s advice and was lucky it coincided with a morally good choice.Footnote ⁷

It’s worth bearing in mind, as well, that when Geraldine acts disrespectfully by deferring to Polly we cannot simply say that this is just an indication of her character (though it may be that too). Geraldine is wronging a particular person—Ellen could reasonably complain that she, herself, had been mistreated by Geraldine’s recklessness. And this complaint would remain even if making an important decision based on a parrot’s advice was, in fact, out-of-character for Geraldine. She still would have done something wrong when she made an important decision in that way and ran a high risk of mistreating Ellen—even if the parrot’s advice led to a substantively good decision. For this reason, apologizing to Geraldine for handling the decision so recklessly would be appropriate, for example.

Moral obligations like these, that require us to navigate moral uncertainty responsibly, also aren’t a matter of fetishistic focus on our own moral purity (as Weatherson has worried); instead, they involve trying to treat the people in our lives well, even when we are unsure what that involves (Reference Weatherson2014).Footnote ⁸ What this calls for may vary with the context. It might include choosing a “morally safer” action instead of one of uncertain permissibility—respecting a potential romantic partner might involve erring in the direction of being more forthcoming with them, for example. In other cases, it might call for us to seek out good advice or sensible sounding boards—and at a minimum, to avoid making moral decisions based on drunken impulses or advice from pets, when the stakes are weighty, as in Geraldine’s case. Notice that the precautions I’m mentioning are not, for the most part, envelope-pushing suggestions—just reminders of ordinary aspects of common sense morality. In short, we morally ought to treat others with respect, and so (at least sometimes) we morally ought to take steps to make it less likely that we are wronging them. This requires taking some of our moral-epistemic circumstances into account: our uncertainty about the permissibility of some actions, for example, or our improved assessment of a situation following useful advice.

I’ll turn now to a second, but closely related, reason to think it’s morally bad to use reckless strategies for navigating our moral uncertainty and fallible moral judgment—even if we happen to end up acting well. I’ll suggest that if we accept (less controversially) that we have moral responsibilities to prepare thoughtfully for important moral decisions, then we also must accept (more controversially) that our moral-epistemic circumstances matter to which actions we should ultimately take.

To see this, we can notice that strategies to reduce our chances of wrongdoing come in two varieties. On the one hand, there are precautions involving what we do before we act, in preparation for acting. For example, suppose Lisa is considering accepting a political job in which she is likely to face moral dilemmas that make it difficult to avoid harming some people and “getting her hands dirty” (Walzer, Reference Walzer1973), but in which she could also be positioned to bring about important social good. When deciding whether to take the job, she might seek out multiple perspectives from people whose moral insight she respects or who have thought about the relevant values when their communities were impacted by political decisions the past. In another case, Justin might position himself to be less likely to make a morally bad decisions when deciding whether to reveal that a friend cheated on his partner, simply by being sure to consider the issue carefully rather than deciding what to do while rushed or drunk. Of course, there are no guarantees that taking these steps will lead to good actions (perhaps Lisa and Justin’s first thoughtless impulses were best, after all), but because these precautions can make it more likely that we’ll act morally, they can be morally valuable.

On the other hand, we might try to improve our chances of acting morally using precautions that involve the actions we ultimately take, rather than the way we prepare. For example, Amy might avoid a maybe-permissible action in favor of a clearly permissible alternative. Perhaps she is unsure if she has a moral obligation to discuss some aspects of her history with a new romantic partner, but she is confident that it’s morally permissible to be forthcoming about her past. So, she errs in a morally safe direction and has a candid conversation with her partner.

Moral obligations to take the first type of precaution—involving preparing to act—are less controversial and have not been the main targets of opponents of uncertaintism. Claiming that there’s nothing morally good about making our important moral decisions soberly and carefully would be a puzzling rejection of common sense morality.Footnote ⁹

But, accepting that we have obligations to prepare well for important moral obligations has important wider implications for the cases that do worry uncertaintism’s critics. Preparatory obligations would serve no purpose if we were not also obligated to ultimately act in ways that reflect the morally understanding we have formed during that preparation. So, what actions we should ultimately take will depend upon the newfound beliefs that we developed while seeking advice or engaging in contemplation, for example. But the new beliefs, credences, or evidence we gain by preparing for our moral decisions are part of our moral-epistemic circumstances. So, if what we figure out about morality while preparing to act matters to what we should eventually do, then our moral-epistemic circumstances matter to what we should do.

This dynamic makes it very difficult to reject the core claim of uncertaintism—the claim that our beliefs, credences, or evidence about morality can matter to what we morally ought to do. Rejecting this would require saying either (i) that we never have moral obligations to try to figure out what is moral to do (even when making very weighty decisions) or (ii) that we may have obligations to go through the intellectual exercise of puzzling over morality but no corresponding moral obligations to act in ways that are informed by it. And either of these claims would be very surprising. So far, then, we have two arguments that our moral-epistemic circumstances sometimes matter to what we morally ought to do.

But, in spite of the support for uncertaintism in commonsense morality, objections to uncertaintism sometimes seem to be on to something. There are reasonable concerns that we should try to make sense of. In the next section, I’ll suggest that these challenges are better understood as problems for particular, popular varieties of uncertaintism, rather than problems for uncertaintism as a whole. There are a wide variety of ways to develop uncertaintist positions. So, we can, and should, relocate a lot of these debates: let’s talk about how and in what ways our moral-epistemic circumstances matter to what we morally should do—rather than whether they matter to it at all. In the next section, I’ll identify three features that some criticisms tend to attribute to uncertaintism and spell out why the criticisms depend on these assumptions. But I suggest that these features aren’t essential to uncertaintism, and I close the paper by offering an alternative form of uncertaintism that avoids these problems.

In what follows, I often use language suggestive of moral realism, for ease of discussion. But I take my argument to be compatible with a variety of metaethical views. I only need it to be the case that being moral is something we can be better or worse at (for reasons besides akrasia). If this holds, I’ll suggest we morally ought to try to be “better at” morality, regardless of the underlying metaethical structure. Indeed, insofar as a metaethical view does not leave room for concerns about possible moral mistakes and efforts to avoid those mistakes, that seems, itself, to be a problem for that metaethical theory—an indication that it cannot account for a significant aspect of our moral lives.Footnote ¹⁰

3) Criticisms of Uncertaintism as Criticisms of Only Specific Types of Uncertaintism

Some criticisms of uncertaintism focus on its implications for people who are confidently wrong. Brian Weatherson worries that uncertaintism licenses confident zealots to follow their convictions—after all, if what we should do depends on our moral beliefs, it might seem that the confidently mistaken should act on their wrong moral beliefs (having no uncertainty to hold them back) (Weatherson, Reference Weatherson2019, p. 6). And Elizabeth Harman suggests that uncertaintism involves exculpating people who have done awful things while sincerely believing in their own righteousness (Harman, Reference Harman and Shafer-Landau2015).

But these objections to uncertaintism depend on attributing particular commitments to uncertaintism. Below I highlight three features that one or both of these criticisms attribute to uncertaintism: that it is committed to a decision-theoretic analysis of our moral responsibilities (possibly one involving expectation maximization, in particular); that it must involve moral-epistemic circumstances mattering to all moral norms—not only some of them; and that it must treat false moral beliefs as exculpating (in part because of these first two commitments). I suggest that none of these features is actually required by the core claims of uncertaintism. So, we can take these concerns on board while viewing them as reasons to adopt alternative versions of uncertaintism, rather than reasons to reject uncertaintism entirely.

4) An Alternative, Moderate Uncertaintism: Substantive and Procedural Oughts

I want to turn now to offering a positive account that builds on this recognition that uncertaintism can take many forms and may be able to avoid key challenges as a result. I develop a structure for uncertaintist accounts in which our moral-epistemic circumstances matter to some norms, but not others—and provide independent motivation for this type of uncertaintism. I’ll argue that our moral responsibilities in the face of our moral uncertainty and fallible moral judgments include both substantive requirements and procedural requirements, mirroring the structure of norms of substantive justice and legal due process norms. Rather than competing with each other, these norms address different moral questions—answering procedural moral questions will require information about our moral-epistemic circumstances and answering substantive moral questions may not.

In Section 2, I defended one aspect of this: navigating moral uncertainty recklessly is morally bad and this reveals that our moral-epistemic circumstances sometimes matter to what we morally ought to do. But, why think that there are both moral norms that do not need to be sensitive to our moral-epistemic circumstances and these moral norms that are sensitive to them?

It’s difficult to understand the purpose of uncertaintist norms against moral recklessness without this structure, and an account with both types of norms makes the best sense of ordinary processes of moral deliberation. To see this, let’s look at the role each type of moral norm is playing. Sometimes we are unsure of the content of one set of norms (Geraldine, for example, is unsure whether she should break her promise of confidentiality), and our moral judgments are always fallible. At the same time, we usually behave as though there are ways we can do a better or worse job navigating that kind of uncertainty and fallibility (both through what we do in advance and by ultimately making decisions that are shaped by that preparation). Depending on the case, this could include considering the issue carefully, seeking input from people who may have wise advice or insight into relevant perspectives, avoiding morally questionable actions when more clearly permissible alternatives exist, or avoiding making high-stakes moral decisions while drunk or distracted. My aim here is not to settle the best strategies. Instead, I’m suggesting that if we deny there are better and worse approaches—if we deny there are norms for how we should navigate our uncertainty given our moral-epistemic circumstances—that would be a radical departure from how we ordinarily think about moral decision-making.Footnote ¹⁶ And we treat these norms as making moral claims on us: if Geraldine decided whether to reveal her friend’s drug use by deferring to a parrot, it would be reasonable for her to feel ashamed or for her friend to resent her—even if the parrot led her to a good decision. (Keep in mind that norms about how we make decisions need to come with requirements that we implement those decisions when acting.)

This, together with my arguments in Section 2, already suggests there are some norms that make moral claims on us that are sensitive to our moral-epistemic circumstances—norms concerning what strategies we should use for trying to be moral. But our ordinary practices also recognize the norms that we aim to satisfy with those strategies—which might not be sensitive to our moral-epistemic circumstances. When Geraldine worries about making the right decision about whether to break her promise, she is not thinking that as long as she does what seems best from her limited moral-epistemic circumstances, she will have behaved morally rightly. The reason Geraldine bothers to consider carefully, in the first place, is because she thinks there’s some further set of norms that are important to comply with, but whose content she’s uncertain of. Indeed, it’s difficult to understand the point of deliberating carefully otherwise.

So, it looks like our ordinary ways of thinking about decision-making under moral uncertainty reflect two sets of norms: some first-order moral norms (that might not be sensitive to our moral-epistemic circumstances) and some higher-order norms (sensitive to our moral-epistemic circumstances) that tell us what strategies we should use for trying to comply with those first-order norms. In Section 2, we saw arguments that there are some moral norms concerning the strategies we should use for trying to be moral, that are sensitive to our moral-epistemic circumstances (e.g., norms against recklessness in the face of uncertainty). That these norms make moral claims on us was supported by (i) considerations of respect for those we might be wronging and (ii) the relationship between moral obligations to prepare to act and moral obligations to act in ways that actually reflect that preparation. Now, we have seen that those norms regarding good strategies for trying to be moral cannot exist on their own. Just as obligations to prepare to act only make sense alongside obligations to act in ways that reflect the preparation, these higher-order strategic norms only make sense if they exist alongside first-order norms that they aim to satisfy. Without this, the good strategies would have no purpose and our ordinary ways of thinking about moral decision-making would be difficult to account for. Recognizing both sets of norms does not just avoid particular criticisms of uncertaintism—it’s the only way for uncertaintism to adequately motivate the efforts to do the right thing that are central to the view.

I’ll refer to these two types of norms as procedural oughts and substantive oughts:

Norms of both types make moral claims on us. The contrast between them can be thought of as paralleling the contrast, in legal contexts, between norms of substantive justice (e.g., a trial should convict only the guilty) and norms of due process (e.g., the trial should give the defendant an opportunity to present arguments in their defense). I’ll draw on this parallel to highlight two key features of the account.

First, multiple norms (substantive and procedural or substantive justice and legal due process) can apply to the same act, showing different responsibilities that we have when acting. This is already commonplace in legal contexts. To see this, notice that (like strategies for being moral) legal due process norms can have implications for both the processes that are used and the ultimate decisions that are made. For example, if due process requires that hearsay evidence is not presented to the jury, then due process also requires that we only reach and enforce verdicts that aren’t based on hearsay evidence. Following a process involves adhering to the results of that process.

Complying with these due process norms might coincide with substantive justice. After all, part of the reason we (typically) do not allow juries to convict based on hearsay evidence is that prohibiting (potentially unreliable) hearsay evidence is a good strategy for avoiding wrongful convictions (that is, we hope this strategy makes us more likely to achieve substantive justice). Convicting based on hearsay evidence might run a reckless risk of wrongful conviction in a way that’s unfair to the defendant.

But avoiding recklessness and using good strategies will not always be successful. Sometimes we can follow good strategies and still end up doing something substantively unjust (e.g., convicting the innocent). Likewise, sometimes we can bypass good strategies and be lucky enough to end up achieving substantive justice anyway. But, when this occurs, the procedural violations would still, themselves, be wrong—it would still be unfair to have subjected the defendant to an increased risk of wrongful conviction, even if the conviction was ultimately avoided.

Examples like this help to show that due process norms and norms of substantive justice can make claims about the same decisions while bringing out different responsibilities that we have when acting: we have a responsibility to adhere to due process norms and we also have a responsibility to do what’s substantively just. This is not an exotic metaethical claim—it’s just the ordinary way that we already think about our responsibilities in legal settings.

Procedural and substantive oughts about morality are similar to legal norms in this respect: one decision can be subject to multiple norms including obligations to do the substantively right thing and also obligations to use (and act on) responsible strategies for trying to. So, in the example above, Geraldine might worry about whether revealing her friend’s drug use is substantively right. And she has to decide what strategies to use (or not use) to navigate that uncertainty—those strategies are what she procedurally ought to do. Procedural oughts can apply both to how we should position ourselves to act well (e.g., how much trouble we should go through to seek more moral knowledge) and to questions about what course of action to pick in light of the (perhaps still incomplete) knowledge we have acquired (e.g., should we reduce our chances of wrongdoing by hedging our bets and avoiding a potentially wrong act, just in case). On this view, violating procedural oughts is, itself, morally bad, even if a substantively good decision gets made, and following procedural oughts will not always ensure substantively good actions—just as legal due process violations are wrong, even if a substantively just verdict is reached, and complying with due process norms does not guarantee substantive justice. (See footnote 18 for discussion of the relationship between this framework and work on descriptive uncertainty that distinguishes actions satisfying the criteria of right action from actions resulting from following the best decision procedures.)Footnote ¹⁸

This interplay between different types of norms also helps show a second, key feature of the account that I’ll turn to now. Substantive and procedural oughts are distinct norms that cannot, by themselves, settle each other’s content—they are addressing different moral questions, and moral-epistemic circumstances will matter to the answers in the case of procedural norms, even though they do not have to impact substantive norms. To bring this out, let’s continue the due process analogy. The exact extent to which due process norms are shaped by the goal of achieving substantive justice is a subject of disagreement—and the right answer may depend upon the particular due process norm. But, for many legal due process norms, it’s both the case that the procedural requirements are designed to generate substantively better outcomes, and that this goal will not, on its own, settle what due process requires.

For example, substantive justice requires convicting only the guilty, but processes that reduce the risk of mistaken convictions can be costly in other ways (e.g., because of financial expense or increased jury service requirements). Our actions toward a defendant may ultimately be determined by whether we accept only unanimous jury verdicts in criminal trials (our current practice) or convict defendants based on, for example, a two-third supermajority of jurors (which might avoid some retrials, and so modestly reduce the court resources and number of jurors needed). Knowing that substantive justice requires convicting only the guilty will not settle how to trade-off increased risks of wrongly convicting a defendant against costs in time or money that are more certain. There are distinct normative questions about the trade-offs we should make and the strategies we should use for pursuing substantive justice—questions like “How much time and resources must we be willing to give up in order to achieve a small reduction in the risk of wrongful conviction? What about a large reduction in that risk?” These are procedural questions that require answers, and that aren’t addressed by the substantive requirement to convict only the guilty.

The moral context has a similar structure. Procedural oughts can address questions about what strategies we morally ought to use when trying to satisfy substantive oughts—questions like, “if I’m morally uncertain whether taking an exciting career opportunity would make me unacceptably complicit in unethical behavior, should I decline the job just in case?”Footnote ¹⁹ or “if my moral uncertainty leaves me unsure whether to take a particular action, should I spend time seeking advice and additional perspectives (and then act in ways that take that input into account)—or can I shrug and ‘wing it’?”Footnote ²⁰ Substantive oughts are silent on these questions. They address what we substantively ought to do, not what hard trade-offs to make when we’re not sure what we substantively ought to do—moral questions like these are the domain of procedural oughts.

I will not aim to exhaustively settle the factors that determine procedural norms here, but some examples will be illustrative. The likelihood that a particular action is substantively wrong will matter to procedural norms, yes—but it’s not the only consideration that’s relevant. What I procedurally ought to do—that is, how I should go about trying to do the substantively right thing (given my uncertainty and fallibility)—might depend, for instance, on how I should weigh possible substantive norms when they have different levels of plausibility and severity, or how much I should be willing to sacrifice in order to modestly reduce my risk of substantive wrongdoing. And they may require knowing which options were met with skepticism when I sought advice from wise friends, for example, or which actions seem, to me, like they might stray into a moral grey area—information about my moral-epistemic circumstances (that is not relevant to what I substantively ought to do).

Our moral-epistemic circumstances matter to the answer to procedural questions, like those about whether to seek additional perspectives and advice. The questions, themselves, may not arise under idealized conditions of infallible moral judgment and complete moral certainty. And determining, say, whether we should spend time getting additional input may depend on what we believe so far, for example. But this only means we are taking into account the information that’s most relevant for answering our initial question—and that happens to include some information about our moral-epistemic circumstances. Because procedural and substantive oughts depend on different information, it’s possible for one, but not the other, to be sensitive to our moral-epistemic circumstances.

Understanding that substantive and procedural oughts are answering different moral questions (and relying on different information to do so) can also help show why procedural oughts aren’t the same as “subjective oughts,” at least not in the usual sense. I’m not primarily concerned with policing how far “subjective oughts” terminology from subjective/objective oughts discussions can be made to stretch, and I will leave open whether to characterize procedural oughts as special cases of subjective oughts or a fully distinct phenomenon. But there are important reasons not to run procedural oughts together with typical subjective oughts discussions because the features that distinguish procedural oughts from typical subjective oughts are also features that make the content of procedural oughts “interesting” in Harman’s sense and important to attend to.

First, though less relevantly here, procedural oughts can be specified in more and less subjective ways, depending upon whether the procedural oughts take into account our actual beliefs and credences or only our evidence. And Amelia Hicks has even argued that belief-sensitive norms about how we should navigate moral uncertainty are all, in fact, objective (Hicks, Reference Hicks2022).

Second, more crucially, objective/subjective oughts debates are typically concerned with different interpretations (objective and subjective interpretations) of the same norm, addressing the same question—say, should a doctor give a patient the medicine in bottle A—but answering that question with different limitations on the information used (roughly, subjective oughts draw only on the credences of the person acting, and objective oughts draw on all relevant information). In contrast, substantive and procedural oughts aren’t different interpretations of the same obligations or answers to the same moral question with different restrictions on our information. Instead, substantive and procedural oughts identify different moral obligations and address different questions, using the information that is relevant to those questions. We ought to do the substantively right thing, and also we ought to be moral in the way we go about trying to do the substantively right thing. So, showing why procedural oughts are interesting or important (in spite of Harman’s concerns) is straightforward: we should care about procedural oughts simply because they are additional obligations that we have, and we genuinely wrong people if we violate them.

The parallel to legal processes is again illuminating. We’ve seen that the epistemic circumstances of a trial impact what procedures we should use—for example, we need to consider how likely different procedures are to yield correct verdicts, given that we do not know who has actually committed a crime. But saying that a requirement of juror unanimity gives a subjective interpretation of our duty to convict only the guilty, for example, would be misleading. It obscures the way in which these due process norms are duties of their own (“interesting” in Harman’s sense, as a result) and address distinct questions, using the information relevant to those questions.

Similarly, when what we procedurally ought to do depends on our moral-epistemic circumstances, this only means we are taking into account the information that’s most relevant for answering our procedural questions—and that happens to include some information about our moral-epistemic circumstances. This is not a case where the most relevant information is excluded in order to focus on our own perspective, as it is with typical subjective oughts. Given this, procedural oughts are worth distinguishing from the norms that subjective oughts terminology usually identifies. Procedural oughts are their own moral norms, addressing different moral questions than substantive oughts, and using the most relevant information to do so.

Before moving forward, it’s worth briefly addressing a couple of possible concerns about the structure that I’ve been proposing. Although substantive and procedural oughts depend on different factors, it might still seem surprising that substantive and procedural oughts can directly contradict. What one forbids, the other may require, or vice versa (e.g., if our best strategy for reducing moral risks turns out to lead us to act in ways that are substantively morally wrong, despite our best efforts). But it’s already recognized that conflicting moral norms can sometimes bear on the same physical act, as with an act that is both required by legitimacy and forbidden by norms of substantive justice, for example. And, as Andrew Sepielli has noted, in practice in morality, we typically only receive guidance from oughts at one level at a time (Reference Sepielli and Timmons2017, p. 110). Put in the language of my account, we do not normally encounter the demands of substantive morality and then decide how to handle conflicts with procedural norms—if we genuinely knew exactly what substantive morality required, then there would be no conflict with procedural norms because the best procedure would involve using that knowledge of substantive morality. But in real life, we cannot be sure if we have that substantive knowledge, so we are often dealing with procedural oughts.

Along similar lines, these conflicts mean that we may sometimes be unable to satisfy all of morality’s demands because we face conflicting requirements. But this is a variation on something that is already commonplace. The conflicting moral requirements only arise because we lack full moral knowledge, and if gaps in our moral knowledge make it difficult for us to comply with all of morality’s demands, that’s an ordinary and unsurprising predicament.

Recognizing that guidance from substantive and procedural oughts can point in different directions also helps address a problem facing obligations to use good procedures: even the very best procedures sometimes lead us to act badly. If these procedures were our only obligations, that might seem to make a lot of bad actions permissible. Indeed, concern over this possibility motivates many critics of uncertaintism, as we saw in Section 3. But once we recognize that there are also substantive moral norms that may give different guidance, this concern loses much of its bite. If we have done something morally right by following good procedures and nevertheless been led astray, there is still something important (and, arguably, more important) that we have done morally wrong—we violated substantive moral norms.

So far then, we have a structure in which our moral epistemic circumstances impact what we morally ought to do, without necessarily shaping every moral norm (e.g., they might matter for what we procedurally ought to do without mattering for what we substantively ought to do). This avoids two of the “problem features” we examined above. It directly removes the assumption that moral-epistemic circumstances must matter to all moral norms, if they matter to any. And, indirectly, it addresses the related worries about exculpating people based on their moral ignorance that depended on that assumption. If we adopt my earlier recommendation (in Rosenthal (Reference Rosenthal2021)) to reject decision theory as a guide for navigating moral uncertainty, we can also avoid the remaining “problem feature” (decision-theoretic approaches). With this alternative structure for thinking about moral uncertainty on the table, I want to turn now to briefly addressing a final worry that might arise about this type of uncertaintism.

5) Infinite Regresses

This framework can invite concerns about infinite regresses. I’ve been focusing on uncertainty and fallible judgment about substantive morality’s content. But we could also ask how to handle uncertainty and fallibility regarding what second-order procedural norms require.Footnote ²¹ Questions about procedural norms are hard, and it’s by no means obvious what the correct second-order, procedural account calls for. So perhaps we’d need a third-order theory of how to approach these second-order questions, and a fourth-order theory to address uncertainty about the right third-order theory—with the theories multiplying ad infinitum.

Procedural norms, so the worry goes, are supposed to be action guiding—they are supposed to sort out what we should do when we are mired in uncertainty.Footnote ²² But this infinite pile of third-, fourth-, or seventeenth-order theories may seem to fail at that. So, you might wonder why we are bothering with anything beyond the first-order, substantive questions at all.

But this is less of a problem than it seems because these layers of norms do not have to prevent reasonable and tractable moral guidance in most cases but also because, if my argument so far is right, procedural norms are important regardless of whether they are action-guiding. We need a theory of procedural oughts simply because we need theories about the ways in which our actions might fail to be moral—and failing to follow procedural norms are some of those ways. But I also do not need to give up action guidance so quickly for two reasons.

First, action guidance does not require settling all questions about what to do, with no room for further uncertainty about whether the guidance, itself, is correct. No theory of procedural norms can do this because no theory can do this at all. But we need a working theory of procedural norms to move forward when we face substantive uncertainty, even if that working theory is not the last word. And this type of deliberative role for procedural norms is suggested by the way that actual deliberative processes reflect distinct substantive and procedural norms.

Second, while there may be infinite higher-order, procedural questions, we usually do not need to get mired in questions about seventeenth-order norms just to respect other people and take care not to wrong them. We can often do that without any attention to these very high order norms at all. There are third-order, procedural norms that (at least sometimes) require us to carefully try to comply with second-order norms and explore their content, precisely because this does not seem like wasted effort—we are sometimes able to identify actions that would be bad strategies for being moral and adjust our behaviors. (Indeed, sometimes we are so unsure about first-order, substantive norms that we must attend to second-order norms to figure out what to do at all.) But it’s not clear that investing additional time in eighth- or ninth-order norms would be morally called for in a similar way. It’s very unlikely that the effort would yield important new guidance that gave us reasons to materially change our behavior. So, without evaluating the content of eighth-order norms, for example, we might say that ninth-order norms (about how we approach eighth-order norms) do not require us to cautiously try to determine the content of eighth-order norms and increase our chances of complying with them—the best, ninth-order procedures might be ones that involve ignoring eighth-order norms! And we could generalize this to say that, past a certain level, it is not the case that we morally ought to attend to these higher-order norms (at least barring special circumstances that provide evidence to the contrary).

Even if we do not need to attend to these norms, are there still infinitely many levels of higher-order moral norms that we might be violating? In principle, there can be. But this need not mean that the higher-order moral claims at stake are especially weighty or especially easy to transgress. Indeed, being completely inattentive to these higher-order norms will often be all we need to do.

6) Conclusion

In sum, then, there are common sense reasons to think that what we morally should do sometimes depends on our moral-epistemic circumstances—that is, there are reasons to be uncertaintists. Respecting others includes reducing our risk of wronging them. And, if it’s morally bad to make morally weighty decisions without taking steps that position us to act well (e.g., considering the relevant values carefully and not deciding drunk)—then this requirement only makes sense if we also ought to act in ways that reflect this preparation and decision-making. While a number of significant concerns have been raised about uncertaintism, I’ve argued that these concerns are better seen as problems for specific varieties of uncertaintism—concerns that give us a reason to adopt alternative varieties of uncertaintism, rather than abandon it. The structure of substantive and procedural oughts that I’ve offered provides the framework for one type of alternative that is also an independently more attractive approach to uncertaintism.

Examining procedural oughts also suggests that we have a large body of moral responsibilities that are under-explored. Looking more closely at procedural norms opens the door for making better sense of a variety of moral and political phenomena, perhaps allowing us to explain the value of practices like interpersonal toleration, or exceptions to deontological rules in high-stakes cases, in part through their capacity to reduce moral risk, for example (Rosenthal, Reference Rosenthal and Timmons2018; Rosenthal, Reference Rosenthal2019). And, when we see reckless navigation of moral uncertainty as, itself, morally bad, this begins to show how figuring out what we should do under moral uncertainty may require very different conversations than the ones that have dominated this literature. To identify the best procedures for navigating moral uncertainty, we will need to engage directly with questions about the nature of respect and what sacrifices we should make to reduce our risk of wronging others (although resolving these questions is far outside the scope of this paper).Footnote ²³