A Normativity Wager for Skeptics

Abstract

Several philosophers have recently advanced wager-based arguments for the existence of irreducibly normative truths or against normative nihilism. Here I consider whether these wager-based arguments would cause a normative Pyrrhonian skeptic to lose her skepticism. I conclude they would not do so directly. However, if prompted to consider a different decision problem, which I call the normativity wager for skeptics, the normative Pyrrhonian skeptic would be motivated to attempt to act in accordance with any normative reasons to which she might be subject. Consideration of the normativity wager will not inevitably cause the normative Pyrrhonian skeptic to lose her skepticism, but there are at least three routes by which it might: first, in considering the wager the agent may spontaneously (non-rationally) acquire a normative belief; second, considering the wager can motivate the agent to cause herself to (non-rationally) acquire a normative belief. Via either of these indirect, non-rational routes, she would cease to be a normative Pyrrhonian skeptic. Thus, consideration of the normativity wager may have value, even if it does not supply a rational argument that will dissuade skeptics. In addition, I consider the possibility of a third (rational) route by which the agent might lose her skepticism.

1 Introduction

Kahane (2017) has recently presented a wager-based argument against normative nihilism. Building on Parfit (2011), Beardsley (2022) proposes a wager-based argument for the existence of irreducibly normative truths. Here I consider whether these wager-based arguments would cause a normative Pyrrhonian skeptic to lose her normative skepticism.¹ I claim they would not do so directly. However, I propose that consideration of a closely related decision problem, which I call the normativity wager for skeptics, would cause the normative Pyrrhonian skeptical agent to be motivated to act as if there are irreducibly normative truths, and to attempt to act in accordance with any normative reasons to which she might be subject. Furthermore, I argue that there are at least two indirect, non-rational routes by which consideration of the normativity wager might lead the agent to cease to be a normative Pyrrhonian skeptic: she might spontaneously (non-rationally) acquire a normative belief, or she might be motivated to cause herself to (non-rationally) acquire a normative belief. In addition, I consider the possibility of a rational route by which the agent might lose her skepticism.

The article proceeds as follows. In Sect. 2, I characterize the normative Pyrrhonian skeptic of interest. In Sect. 3, I summarize Kahane’s and Beardsley’s proposals, and I consider whether the arguments advanced by Kahane and Beardsley would convert the skeptic. I conclude that they would not. In Sect. 4, I describe an alternative decision problem that the skeptic faces. I argue that when confronted with this problem the skeptical agent will be motivated to attempt to act in accordance with any normative reasons to which she might be subject, even if she has not acquired the belief that there are irreducibly normative truths and thus remains a skeptic. In Sect. 5, I argue that after considering the normativity wager the skeptic might acquire a normative belief via a spontaneous process, or she might be motivated to attempt to cause herself to acquire a normative belief; in either case she would cease to be a normative Pyrrhonian skeptic, though her route out of skepticism would not be rational. I also discuss a possible rational route out of skepticism. I offer concluding remarks in Sect. 6.

2 The Normative Pyrrhonian Skeptic

The normative Pyrrhonian skeptic that is the subject of this article is an agent entertaining global normative error theory—she believes it is possible a global normative error theory is true and has suspended judgment on whether it is true or false.² She believes that if there are any normative reasons, there are some categorical, irreducibly normative reasons.³ However, she suspends judgment on whether there are any normative reasons, and she suspends judgment on what the substance of normative reasons would be if there were any.⁴ The agent is “Pyrrhonian” in the sense that she suspends judgment on these questions. In virtue of her psychology, the skeptic tends to reason in accordance with the rules of inference that are familiar to us, but she does so without believing that such rules are rational or having any other normative beliefs about them.⁵ If she appears to the observer to act in pursuit of ends, she does so via processes that do not require her to endorse ends, in the sense of believing that there are normative reasons to pursue ends. In her apparent pursuit of ends, she may employ processes that non-human animals also use, she may act unreflectively on her emotions, she may act on impulse, out of habit, on the basis of ‘drives’, or may employ any similar process.

The skeptical agent may have arrived at her skeptical state from a number of directions. She may have begun with an ordinary, run-of-the-mill set of substantive normative beliefs, or with any other set of substantive normative beliefs—e.g., moral egoism or the normative beliefs of an ideally coherent Caligula who values torturing people for fun. Her normative beliefs may have been shaken loose via a number of routes—consideration of global normative debunking arguments, the proposed ‘queerness’ of normative properties, an existential crisis, etc. Regardless of the route by which the agent arrived at her skepticism, the agent that is the subject of this article has suspended judgment on the existence of irreducibly normative reasons and suspended judgment on what the substance of such reasons would be if there were any, while retaining the metaethical conviction that if there are any normative reasons, there are some categorical, irreducibly normative reasons.⁶

3 Wager-Based Arguments for Normativity

3.1 Kahane’s Anti-Nihilist Wager

Kahane’s (2017) wager-based argument against normative nihilism concerns both evaluative and practical nihilism. Evaluative nihilism is the view that nothing has final value, that “nothing is good or bad” (330). Practical nihilism is the view that there are no practical reasons, that there are “no reasons to do, want, or feel anything” (330). The nihilism at stake is global, having to do with any value and reasons, whether moral, epistemic, pragmatic or of some other kind. The question Kahane (2017) considers is whether to believe that some things matter or believe that nothing matters.

I understand Kahane’s argument as follows. Either nothing matters, or some things matter. If nothing matters, then none of one’s actions matter: among other things, it doesn’t matter if one believes that some things matter or not. By contrast, if some things matter, then one’s actions may matter: in particular, it may matter whether one believes that some things matter or one believes that nothing matters. Regardless of whether some things actually matter, if one were to believe that nothing matters, one would lose one’s tendency to act in accordance with one’s ordinary, first-order normative beliefs. Kahane admits that even if some things matter, there is a large range of things that could possibly matter—one’s ordinary first-order normative beliefs could be roughly on track or completely misguided. But if one’s ordinary first-order normative beliefs were roughly on track, he says, it would be better to believe that some things matter than to believe that nothing matters, because in such a condition, retaining one’s tendency to act in accordance with one’s first-order normative beliefs makes one more likely to act in accordance with what (truly) matters. If one supposes that one’s ordinary first-order normative beliefs are roughly on track, one faces the situation depicted in the decision matrix in Table 1. Here outcome (3) is better than (4), and (1) and (2) are not the sort of outcome that can have value, occurring as they do in the condition where nothing matters. With regard to outcomes (1) and (2), as Kahane puts it, “It’s not that these possibilities have zero value—they are, we might say, beyond good and evil” (345). Consequently, both in the condition where nothing matters, and in the condition where some things matter, it would be better or no worse for the agent to have chosen action (A) rather than action (B): (A) dominates (B).

Table 1 The preliminary version of Kahane’s anti-nihilist wager

However, as Peterson (2018) points out, without the assumption that one’s first-order normative beliefs are roughly on track, it is not the case that Table 1 depicts the decision problem one faces, and it is not the case that within one’s set of options for action, one action dominates. Kahane acknowledges this in his own article, and offers another representation of the situation, in which we should not need to assume that one’s first-order normative beliefs are roughly on track. In the revised representation (see Table 2), he splits the “Some things matter” condition into a condition in which our evaluative beliefs are roughly on track and a condition in which our evaluative beliefs are not roughly on track.

Table 2 Kahane’s revised anti-nihilist wager

As Peterson (2018) observes, in this situation, it is not obviously the case that (A) dominates (B). He argues that the condition in which “Things matter” and “Our evaluative beliefs are false” is “radically underspecified” (599). It would be more accurate, he says, to represent the situation using a different decision matrix, where the options for action are the same as in Tables 1 and 2 but where the possible states of the world include—in addition to the state where nothing matters—an indefinite number of other specific, fully characterized states of the world where things matter. Across these world-states, there is great variation in what matters: in some states, the agent’s first-order normative beliefs are roughly on track, and in other states the agent’s first-order normative beliefs are wildly off. Crucially, on Peterson’s picture, in some of the possible world-states, action (A) would make one less likely to act in accordance with what matters than action (B) would. As a result, in the revised decision matrix, (A) does not dominate (B).⁷

In sum, Kahane’s anti-nihilist wager faces a difficulty analogous to the “many-gods” objection raised against Pascal’s wager. A critic of Pascal’s wager may argue that there are possible conditions with non-Christian gods, and as a result there are some conditions where it would be worse to believe in a Christian god than not. In the case of Peterson’s revised anti-nihilist wager, there are possible conditions (however unlikely) where what matters is so different from what the agent believes matters that a belief that some things matter would produce a worse result than would a belief that nothing matters.⁸

Another objection against Kahane’s argument, from Jeffrey (2022), is that “Nihilism should technically undermine our attempts to fill out meaningful values—even 0—on the hypothesis that Nihilism is true” (4–5). The outcomes under the Nihilism hypotheses, she says, contain undefined values. As a result, arguments based on Kahane’s and similar wagers “will fail to generate a truly Pascalian practical reason” (6). She continues, “It is not appropriate to treat null entries as though they are commensurate with the values of the other options. A rational agent should not be able to weigh them against expected values for other options as though they are on the same metric” (6). I understand Jeffrey to be saying that the existing tools of decision theory do not accommodate this type of situation. Clearly, the principle of maximizing expected utility cannot be applied when the values for some outcomes are null. However, the question is whether the principle of dominance can be applied. The dominance principle says choose the action that dominates all other actions, where action A dominates action B if and only if for each possible world-state the outcome associated with A is better or no worse than the outcome associated with B and for at least one possible world-state the outcome associated with A is better than the outcome associated with B. In Kahane’s original wager, it can: that the value of (1) and (2) is null is not an obstacle to stating that in each possible condition, it would be better or no worse for an agent to have chosen (A) rather than (B). In the condition where nothing matters, specifically, it is no worse to choose (A) than (B); whether one chooses (A) or (B) the result is a null value. Thus, I do not believe Jeffrey’s observation poses an obstacle to the construction of wager-based arguments that appeal to dominance or similar properties.

Would Kahane’s anti-nihilist wager convert the normative Pyrrhonian skeptic? I do not expect so, due to the “many-gods”-type objection. Because our skeptic of interest has suspended judgment on the question of which substantive normative claims would be true if any were, she will not face the decision problem that Kahane presents. Rather she will take herself to be in the situation that Peterson describes. There, no dominant action presents itself. We will now turn to Beardsley’s wager-based argument, which also faces a “many-gods”-type problem, and so will not suffice as it stands for the purpose of converting the skeptic, but which is interesting because it features a principle, the principle of normative superdominance, that might be more compelling than the principle of dominance. Then we can ask, if a skeptical agent is confronted with a decision problem in which one action option does dominate or superdominate all others, would it bring her out of her skepticism—would it lead her to believe that in the situation she faces there is a reason for her to take one action over another, and thus that there are some irreducibly normative truths? I introduce such a decision problem in Sect. 4.

3.2 Beardsley’s Wager-Based Argument

Beardsley defends an argument that he calls Parfit’s wager. The argument combines premises from Parfit (2011, p. 619) with a principle Beardsley calls the normative superdominance principle. The normative superdominance principle relies on the concept of superdominance, which Beardsley defines as follows:

“A course of behaviour ‘normatively superdominates’ all possible alternative courses of behaviour if and only if (1) the normatively best possible outcome associated with this behaviour is normatively better than any of the possible outcomes associated with the alternatives, (2) the normatively worst (or least normatively good) possible outcome associated with this behaviour is normatively equal to or better than the normatively best possible outcome associated with any of the alternatives, and (3) the normatively worst possible outcome associated with each one of the alternatives is normatively worse than the normatively worst possible outcome associated with this behaviour.” (2022, 417)

If action option (A) superdominates the other options, then it is also the case that (A) dominates the other options; that (A) (and no other action) is associated with the best possible outcome; and that the worst outcome associated with (A) is better than the worst outcomes associated with each other possible action option. Thus, an action that superdominates would also be recommended by the decision theory principles of dominance, maximax and maximin.⁹

The normative superdominance (NS) principle states:

(NS) “If a course of behaviour (such as a forming a belief) normatively superdominates all possible alternative courses of behaviour (such as forming a contrary belief, or suspending belief), we have at least some normative reason to engage in this behaviour (such as by forming the belief).” (2022, 417)

The Parfit’s wager argument that Beardsley develops draws on the following passage from Parfit (2011, p. 619):

“If we believe that there are some irreducibly normative truths, we might be believing what we ought to believe. If there are such truths, one of these truths would be that we ought to believe that there are such truths. If instead we believe that there are no such truths, we could not be believing what we ought to believe. If there were no such truths, there would be nothing that we ought to believe. Since

(D) it might be true that we ought to believe that there are some irreducibly normative truths,

and

(E) it could not be true that we ought not to have this belief,

we can conclude that

(F) we have unopposed reasons or apparent reasons to believe that there are such truths,

so that

(G) this is what, without claiming certainty, we ought rationally to believe.”

Beardsley’s version of the Parfit’s wager argument is as follows:

“1. Believing that there are some irreducibly normative truths normatively superdominates not believing that there are such truths

2. If a behaviour normatively superdominates all possible alternatives, we have at least some normative reason to engage in this behaviour (NS)

3. Therefore, we have at least some normative reason to believe that there are some irreducibly normative truths

4. Therefore, there are some normative reasons

5. If there are some normative reasons, there are some irreducibly normative reasons

6. Therefore, there are some irreducibly normative truths.” (2022, 419)

The decision problem that accompanies this argument is depicted below (in Table 3). Outcomes (4–6) are normatively neutral. In outcome (1), Beardsley claims, we believe what we ought to believe; hence outcome (1) is better than (2), in which we believe what we ought not to believe, or (3), in which we fail to believe what we ought to believe. As a result, action (A) superdominates (B) and (C).

Table 3 Beardsley’s normative matrix

Would Beardsley’s version of the Parfit’s wager argument cause a normative Pyrrhonian skeptic to lose her skepticism? There are at least two stumbling blocks. One is the first premise—the claim that (A) superdominates (B) and (C). This premise relies on Parfit’s claim, “If there are [irreducibly normative] truths, one of these truths would be that we ought to believe that there are such truths” (2011, 619). This claim is important for the argument: it is in virtue of this that outcome (1) is normatively better than outcome (2) and outcome (3), and consequently that (A) superdominates (B) and (C). Yet from the perspective of the skeptical agent, who has suspended belief on which substantive normative claims would be true if any were, the truth of Parfit’s claim is not obvious. There is an open possibility, from the perspective of the skeptic, that there exist irreducibly normative truths, yet none of them is that agents ought to believe that there are such truths.¹⁰ Because of this, the familiar “many-gods” type of problem reappears: Beardsley’s normative matrix can be revised to split the “There are some irreducibly normative truths” condition into multiple conditions; in some of the conditions, (A) does not lead to the outcome that we “Believe what we ought to believe”—it may instead lead to the (worse) outcome that we “Believe what we ought not to believe.” As a result, the skeptic would not accept Beardsley’s first premise—she would not believe that (A) superdominates the other options. This alone has the result that the normative Pyrrhonian skeptic will not follow the Parfit’s wager argument, as it stands, to reach a belief in the conclusion that there are some irreducibly normative truths.

Premise 2, the normative superdominance principle (NS), may also be a stumbling block, but this is perhaps less clear. Even if it is not strictly speaking a substantive normative claim¹¹ nor a claim about what substantive normative claims would be true if any were, NS appears to me to be a normative claim on which a normative Pyrrhonian skeptic will have suspended judgment. To see why, it is perhaps helpful to consider whether the skeptic would believe in other possible decision principles, such as a comparable version of the dominance principle, DP: if an action dominates all other possible actions, one has at least some normative reason to choose that action, or the principle of maximizing expected value, MEV: if an action maximizes expected value, one has at least some normative reason to choose that action. I would expect a normative Pyrrhonian skeptic to have suspended judgment on these principles, if only due to the extensive disagreement in decision theory about which decision principle should be applied in cases where DP recommends something different than the principle of maximizing expected value (e.g., in Newcomb’s problem). Is there something special about NS that sets it apart from other decision principles, making it less contentious? An anonymous reviewer suggests that NS might in fact be a conceptual truth, such that it would be accepted by the normative Pyrrhonian skeptic if only she has an adequate grasp on the relevant normative concepts. If the reviewer is right, this may point to a rational route out of skepticism for the normative Pyrrhonian skeptic. I remain unsure, though, whether NS is a conceptual truth. Even if it is not, I nonetheless think that when paired with a different decision problem, it may bring the normative Pyrrhonian out of their skepticism in a non-rational manner. I will return to these potential routes out of skepticism in Sect. 5.

In the next section, I present a decision problem closely linked to those that Kahane, Parfit, and Beardsley discuss. I label this problem the normativity wager for skeptics. However, I do not use it as the basis for an argument leading to a conclusion about the existence of irreducibly normative reasons or truths. I put the decision problem to a different purpose. If prompted to consider the normativity wager decision problem, I claim that a normative Pyrrhonian skeptic will acquire a motivation to attempt to act in accordance with any normative reasons to which she might be subject. The skeptical agent will wager on normativity, even if she maintains her skepticism. In addition, as I will discuss in Sect. 5, consideration of the wager may lead her to lose her skepticism.

4 The Normativity Wager for Skeptics

A normative skeptic who finds herself entertaining the question of normative nihilism or the question of whether there are some irreducibly normative truths faces a practical question about what, if anything, to do next.¹² In particular, for as long as the agent remains in a state of suspended belief with regard to these questions, she faces a practical question about whether to attempt to comply with any decisive, categorical, irreducibly normative reasons to which she might be subject, or to not attempt this.

The decision problem may be depicted as in Table 4. From here on out, I will use the term ‘reason’ to stand in for ‘decisive, categorical, irreducibly normative reason.’

Table 4 The normativity wager for normative Pyrrhonian skeptics

When a normative Pyrrhonian skeptic considers this problem, I claim that they will become motivated to take action (A). Why? Action (A) exhibits two interesting properties, either of which would suffice to produce the motivation to take (A). One is the property of superdominance.¹³ The second is what I will call the property of sole possible value. I do not claim that the agent will decide to select (A) on the basis of a reason that she believes herself to have; I claim only that upon considering the decision problem, the agent—without necessarily ceasing to be a skeptic—will acquire a motivation to (A).

Let us examine the action options and states of the world in this problem. The agent has suspended belief on the substance of what she would have reason to do if there were anything that she had reason to do. What, then, is involved in action (A)—attempting to comply with any reasons to which she may be subject—given that she does not have a belief about what those reasons would be and remains uncertain about whether any such reasons exist? Human agents can form a motivation to take an action, without a fully elaborated picture of what is involved in the execution of the action, and without being certain that they can achieve the action.¹⁴ In this case, the agent will acquire the motivation to attempt to act in accordance with any reasons to which she might be subject, even though she lacks a full picture of what that entails. The skeptic remains open to a broad range of possibilities: it may be that what there is reason for the agent to do is to maximize her own happiness during the year that she is 23; to act in accordance with Bible teachings; to never look at hedgehogs in the light of the moon; or to make the surface of the earth as purple as possible, etc. One may think of a motivation to take action (A) as a motivation to act rightly de dicto without a belief that there is something that would be right for one to do. In Sect. 5, I will comment on what further motivations the agent would be likely to acquire, and further actions she would be likely to take, after acquiring a motivation to (A).

I claim that (A) superdominates (and thus also dominates) (B). Furthermore, (A) is the only action associated with an outcome of any value. I call this the property of sole possible value (SPV). One can imagine an agent that does not comply with the decision principles of dominance, maximax, maximin, or superdominance, but only selects actions if they have the property of SPV¹⁵:

The SPV decision procedure: If, within one’s set of action options, (A) is associated with at least one outcome of value, and no other action is associated with any outcome of value, choose (A).¹⁶

This could alternately be presented as a decision principle with a structure like NS or DP:

The SPV principle: If, within one’s set of action options, (A) is associated with at least one outcome of value, and no other action is associated with any outcome of value, one has at least some normative reason to (A).

I expect it will be rare for an agent to face a problem in which she could apply the principle of SPV. Nonetheless, the principle is interesting because if superdominance is a more minimal normative assumption than dominance, SPV is a more minimal assumption still. Furthermore, detecting the property of SPV requires fewer cognitive resources: ascertaining that an action in a decision problem has the property of SPV requires only that one registers the presence or absence of value in each possible outcome; by contrast, ascertaining that an action in a decision problem superdominates requires not only that one registers the presence or absence of value in each outcome but also that one compares whether the values of some outcomes are higher or lower than the values of other outcomes.

Why think that (A) superdominates and has the property of SPV? Outcomes (3) and (4), as in Beardsley’s and Kahane’s decision matrices, are not the sort of thing that have value. As I will argue below, in a condition where there are no reasons to which the agent is subject, none of the outcomes in that condition have an associated value. Outcome (1) is better than outcome (2). I will argue that in the condition where there is at least one reason to which the agent is subject, it is better if the agent responds to the wager by attempting to comply with any reasons to which she is subject, than if the agent does not attempt this. I claim, further, that there is no value associated with outcome (2).

Why does the “many-gods”-type objection not arise for this problem? For (1) to be better than (2) it must be the case that for any set of reasons to which the agent might be subject, taking action (A) rather than (B) (at the decision moment in question) will be more conducive to the agent’s compliance with those reasons. Were this assumption to not hold, Table 4’s characterization of the states of the world would be inadequate: there would be some possible conditions where (B) would produce better outcomes than (A), and (A) would not dominate (nor superdominate) (B); nor would (A) possess the property of SPV. I claim that this assumption holds. For any set of reasons to which the agent could possibly be subject, responding to the wager by doing (A) rather than (B) will be more conducive to acting in accordance with the reasons one is subject. I suppose that a behavior may be conducive to the agent acting in accordance with the reasons to which they are subject by increasing the extent to which the agent acts in accordance with the reasons or increasing the probability that the agent acts in accordance with one or more of the reasons.

One might doubt that the assumption holds because one is worried about peculiar situations such as the following: there is a demon that determines what (if anything) is of value and what (if anything) the agent has reason to do; the demon fixes things so that the world will go better if the agent does (B). Would this be a situation in which outcome (2) is better than (1)? Let us first consider whether this case is an instance of condition I—the condition in which the agent is subject to at least one reason. It might appear that there is reason for the agent to do something in this situation: namely, to choose (B). However, in such a situation, although the world will go better if the agent does (B), the demon has not thereby made it the case that there is reason for the agent to do (B). It is true that the world would go better if someone knocked the agent out and prevented her from forming the motivation to do (A)—thereby causing her to do (B). But there is no decisional route by which an agent could select (B). Imagine that having been informed about what world she is in, the agent comes to believe that it would be better if she does (B) than if she does (A), she spontaneously comes to believe that there is a reason for her to (B), and, on the basis of this belief she decides to do (B).¹⁷ Via this cognitive process, the agent will have thereby already performed action (A)—she will have made an attempt to act in accordance with the reason to which she now believes she is subject. (Given the demon’s decree, her response will have already made the world worse.) In this peculiar situation, the agent is subject to no reasons, because there is nothing the agent could decide to do that would make the world go better. Contrary to what one might have thought, the agent in this situation is not in condition I but rather condition II.

To reinforce this point, consider the following hypothetical world, involving a capricious god that determines what is valuable. The capricious god has fixed value in the world so that the world goes best if someone (anyone) constructs a building resembling the Eiffel Tower in Australia. If no one does this, the world has gone in the worst possible way. Nothing else matters. Our agent is located on a remote island in the Atlantic, with no means of travel and no way of communicating or assisting the people in Australia. By stipulation, nothing the agent could possibly do will have an influence on whether the Australian Eiffel Tower is built. Although this world can go better or worse, there is nothing the agent can do to influence how well it goes. Consequently, in the peculiar world of the capricious god, our agent is in condition II: subject to no reasons.¹⁸ Because of the way in which the capricious god has fixed what is valuable in the world, there is nothing that could be gained or lost via any of the agent’s decisions (or behaviors, more generally). If the agent happens to behave in accordance with (B) because it never occurs to her to attempt to act in accordance with any reasons to which she might be subject, nothing is gained or lost; likewise, if she is motivated to do (A) or if she choses to do (A) (e.g., on the basis of a false belief that there is a reason for her to do (A)), nothing is gained or lost.

My condition II (There are no reasons to which I am subject) diverges somewhat from the corresponding conditions presented by Beardsley (There are no irreducibly normative truths) and Kahane (Nothing matters). This is because the practical question that the normative Pyrrhonian skeptic faces is not a question of curiosity about what exists, but the very pressing question of what to do next. As in Kahane’s and Beardsley’s decision matrices, though, the outcomes of condition II are beyond value. If the building of a mock Eiffel Tower would be valuable, but (by stipulation) there is nothing the agent can possibly do to aid that effort, this fact has no implications for what there is reason for the agent to do. For the purpose of making decisions about what to do, an agent in the world of the evil demon or the capricious god might as well be in a world of no irreducibly normative truths or a world where nothing matters—it is still the case that the outcomes associated with condition II are beyond value, because of the fact that none of the actions the agent can take will affect how well the world goes.

So, there are some possible worlds, which one might think supply an instance where outcome (2) is better than (1), but which in fact turn out to be worlds where the agent is subject to no reasons. These cases thus do not produce the “many-gods”-type objection—they introduce no condition where choosing action (B) is associated with a better outcome than choosing action (A).

Now let us consider a possible counterexample that is slightly more complex: the extended-duration evil demon world. In this world, value is fixed by an evil demon in the following way. The best world trajectory is one in which at T₀, when the agent considers the wager, the agent does (B), and then at each subsequent moment, the agent also does not attempt to comply with any reasons to which she might be subject. The worst trajectory is one in which the agent does (A) at T₀, and then at each subsequent moment the agent continues to attempt to comply with any reasons to which she might be subject. There are many trajectories of intermediate value: the greater the number of moments in which agent is not attempting to comply with any reasons to which she might be subject, the better the trajectory of the world. Now, suppose the agent is informed that this is the condition she is in. As in the first evil demon case discussed, at T₀, the agent forms a motivation to (B) and thereby does (A). This is not a good start, but as we have discussed, at T₀ the agent cannot choose to do anything better. However, after T₀, the world may still go better or worse, and the agent’s choices can influence how well it goes. If the agent recognizes this, and does some further instrumental reasoning, she may decide at, say, T₅, to take an action that will prevent her at each subsequent moment from attempting to comply with any reasons to which she is subject. For instance, she may decide at T₅ to take a drug that causes her henceforth to cease to engage in practical reasoning. This is a better world trajectory than the one in which at every moment she is attempting to comply with any reasons to which she might be subject. In the extended-duration evil demon world, then, the agent is subject to some reasons: over time, there are some actions that the agent could choose to take that would affect how well the world goes. The agent in this world is in condition I.

Does the extended-duration evil demon world supply a counterexample to the claim that with regard to any set of reasons to which the agent might be subject, taking action (A) rather than (B) (at T₀) will be more conducive to the agent’s compliance with those reasons? No. The world trajectory that the agent initiates at T₀ is not the best possible trajectory—the world would have gone better if someone had knocked out the agent at T_−1. Nonetheless, it is still the case that if this is the world where the normative Pyrrhonian skeptic unknowingly happens to be, she is more likely to act in accordance with more of the reasons to which she is subject if she does (A) at T₀. If she were to respond to the wager problem by doing (B)—not for a reason (since I have argued that it is not possible to choose to do (B) at T₀ for a reason), but, say, on a whim—she makes herself no more likely to act in accordance with the reasons to which she is subject than if she had never considered the wager at all. It is only by attempting to act in accordance with any reasons to which she might be subject that the agent can increase the probability that she will act in accordance with those reasons.

In sum, in worlds where the agent is subject to at least some reasons, at T₀, regardless of the set of reasons to which an agent is subject, doing (A) is more conducive to acting in accordance with the reasons to which the agent is subject than doing (B). As a result, outcome 1 is better than outcome 2. In fact, outcome 2 involves no value at all—it is only by doing (A) that the agent can increase the extent to which she acts in accordance with the reasons to which she is subject or the probability that she acts in accordance with the reasons. Consequently, (A) is the only action option associated with an outcome of any value (outcome 1). Thus, (A) has the property of SPV, and (A) superdominates (B).

I have argued that in the normativity wager decision problem, (A) superdominates (B) and has the property of SPV. What are the implications of this for how the skeptic will respond to the decision problem, given that she has suspended belief on any associated normative decision principles? When confronted by a decision problem in which one action superdominates or where one action has the property of SPV, I propose that any human agent, including the skeptical agent, will acquire the motivation to choose that action. This is so, even for the normative Pyrrhonian skeptic, because of the action-selection and decision-making capacities that humans possess independent of their capacity for practical rationality. Many of these capacities are shared with non-human animals. I expect that while she remains in a state of skepticism, the agent’s unendorsed, default reasoning processes and inclinations will determine how she reasons and acts. As a result, I suspect that (A)’s possession of either of the properties of superdominance or SPV would suffice to motivate the agent to (A). Furthermore, given that the SPV principle is more cognitively basic than the superdominance principle (or even the dominance principle), I am inclined to think that even if the agent were not disposed to act in accordance with the principles of superdominance or dominance, she would still tend to comply with the principle of SPV. I would go so far as to propose that there is a universal psychological mechanism—across biological agents capable of representing value—such that if the agent perceives one action option as associated with possible value and another option as offering no possible value, and she perceives these as being the full set of options, then the agent will be motivated or otherwise disposed to take the first action.

What I have in mind here is a very minimal notion of value and possible value, and capacity for representing such, which can also be possessed by non-human animals, such as birds engaging in optimal foraging behavior. As the bird considers where to go, something in their cognition is responsive to information bearing on which possible destinations possibly offer something the bird represents as of value. In these species, I expect that in the same way we sometimes find behavior plausibly produced by a decision procedure implementing something resembling the principle of maximizing expected value or the principle of dominance, we will also find that organisms are similarly disposed to act in accordance with the principle of SPV.

Let me emphasize that I do not assume that the principles of superdominance, dominance, or SPV are true, nor that our agent believes them to be true. An agent can believe that one option dominates another, and can be inclined to choose the dominant option, without assenting to a principle of dominance—without believing that she has a reason to follow a principle of dominance. Furthermore, my own conclusion is not that there is a reason for the agent to do (A). My claim is that if the skeptical agent considers the wager in a sustained way, without being distracted away from it, she will acquire a motivation to (A).

5 Implications: Further Motivations and Beliefs

We can now turn to the question of whether consideration of the wager would prompt a skeptic not only to acquire a motivation to (A) but also to cease to be a skeptic, because she has acquired a substantive normative belief.

First let us consider the possibility suggested by a reviewer, as mentioned in Sect. 3, that NS (or, indeed, the principle of SPV) might in fact be a conceptual truth, such that it would be accepted by the normative Pyrrhonian skeptic if only she had an adequate grasp on the relevant normative concepts.¹⁹ Suppose this is so, that one or both of these are conceptual truths. Then, the normativity wager may supply a rational route out of skepticism. If the skeptic would sufficiently reflect on the NS principle and the concepts that constitute it—normative superdominance, possible courses of behavior, normative reason, normatively better, and so on—and obtain an adequate understanding of the principle, she would come to form a belief in NS. Likewise for the SPV principle, and adequate reflection on and understanding of the concepts of value, action options, normative reason, and so on. With a belief in the principle of NS or SPV, and confronted with the normativity wager—a situation in which (A) possesses both the property of superdominance and sole possible value—the agent would likely be inclined to (rationally) form a number of substantive normative beliefs.²⁰ One normative belief she might acquire is belief in the proposition that q: There is reason for me to (A). Another normative belief she might acquire is belief in the proposition that p: There is at least one reason to which I am subject. Since I am currently unsure about whether NS and SPV are conceptual truths, though, I remain unsure about this argument.

There are also at least two ways that contemplation of the normativity wager can (non-rationally) bring an agent out of normative Pyrrhonian skepticism. First, contemplation of the decision problem might produce a condition where it appears to the agent that there is a reason for her to (A).²¹ This perception can then cause the agent to believe that q, rendering the agent non-skeptical. Second, if the agent believes that believing p would make her more likely to do whatever she might have reason to do, this would produce a motivation to seek to acquire the belief that p, regardless of any evidence she possesses bearing on the truth-value of p. Such a motivation would be rooted in the agent’s motivation to do whatever she has reason to do, rather than in a motivation to pursue the truth; in this sense, the agent’s acquisition of the belief that p would be pragmatic. Like Pascal’s subject, who wagers on the existence of god and attempts to cause herself to acquire a belief in god by engaging in religious practices and associating with religious people (Hacking 1972, p. 188), the agent might take steps such as living as if p is true, surrounding herself with people who believe in irreducibly normative reasons, etc. Regardless, if via this pragmatic process the agent successfully acquires the belief that p, the agent will have ceased to be a normative skeptic.

However, it is not inevitable that contemplation of the normativity wager will bring an agent out of skepticism. An agent may contemplate the wager and acquire a motivation to do (A), yet not have the perception that there is a reason to do (A). In such a case, she will not be led by perception to acquire the belief that q, nor any other normative belief. Likewise, if an agent does not believe that believing p makes her more likely to comply with any reasons to which she is subject, the pragmatic route to belief in p that was described above will also not proceed.

I have argued that after consideration of the normativity wager, the skeptical agent will be motivated to do (A). This is so regardless of what she believes about the probability of p.²² Furthermore, consideration of the wager, and subsequent acquisition of a motivation to (A), could stop an agent from inquiring further into the question of whether p is true. If the agent believes that possession of a true belief about p will not affect her probability of complying with whatever reasons to which she is subject, if any, I expect she will not pursue the question of whether p any further. If the agent is unsure about how possession of a true belief about p would affect her probability of complying with whatever reasons to which she is subject, whether she pursues further inquiry into p will depend on her other beliefs, including beliefs about what will make her most likely to comply with any reasons to which she is subject.

This brings me to the topic of what other motivations we can expect the agent to acquire, after having acquired a motivation to (A). The agent’s motivation to do (A), in combination with her non-normative beliefs about the world, may prompt her to acquire a number of further motivations, including the following:

• (a) A motivation to attempt to ascertain the substance of what she would have reason to do if she were subject to reasons. I expect the agent to acquire this motivation if the agent believes that, in general, an agent is more likely to accomplish some particular task if the agent attempts to do that particular task—if the agent’s representation of the task she is attempting to achieve is more fully specified.

• (b) A motivation to attempt to ascertain the means to achieve what she would have reason to do if she were subject to reasons. I expect the agent to acquire this motivation if the agent believes that having true beliefs about what means will achieve one’s ends makes one more likely to achieve that end.

• (c) Various further instrumental motivations, tied to the outcomes of (a) and (b)—such as a motivation to stay in existence long enough to do whatever she has reason to do; a motivation to acquire the capacities to enable her to do what she has reason to do in the future; and motivations with epistemic implications: for instance, if one believes that a general disposition to seek justified beliefs would increase one’s probability of doing whatever it is one ought to do, one may acquire a motivation to develop such a disposition.

None of these additional motivations necessarily draw the skeptic out of her skepticism. Nonetheless, they are interesting because they may produce behavior resembling that which would be performed by the non-skeptical agent who believes that there are at least some reasons to which she is subject.

6 Conclusion

I have argued that the decision problems presented by Kahane and Beardsley would not suffice to move a normative Pyrrhonian skeptic from her skepticism, because each faces an objection equivalent to the “many-gods” objection to Pascal’s wager. As a result of the “many-gods” issue, in neither decision problem is there an action option that dominates (or superdominates). However, Kahane’s and Beardsley’s arguments raise the question of whether a skeptic confronted with a decision problem that did feature a dominant (or superdominant) action would be caused to lose her skepticism. I have introduced the normativity wager, as a candidate decision problem that does feature a superdominant (and dominant) action option. In addition, the superdominant action is the only action associated with an outcome of any value—it has the property of sole possible value. I have argued that if a skeptic is confronted with this decision problem, she will acquire a motivation to act in accordance with any reasons to which she is subject. She will not necessarily thereby cease to be a skeptic, but there are at least two indirect routes by which an encounter with the normativity wager could prompt her to lose her skepticism—namely, via a perception that she is in a situation in which she is subject to a reason, or via pragmatic reasoning, in which she becomes motivated to cause herself to believe that she is subject to at least some reasons, because she believes that acquisition of such a belief would be conducive to compliance with any reasons to which she might be subject. Furthermore, depending on whether NS or SPV are conceptual truths, the combination of the normativity wager with belief in at least one of these principles may produce a rational route out of skepticism.

There is an additional way to test my predictions about the motivations and beliefs of the normative Pyrrhonian skeptic who encounters the normativity wager. Those readers that are capable of putting themselves in the position of the normative Pyrrhonian skeptic may test my claim empirically. To do so, they must come to believe that if there are any normative reasons, there are some categorical, irreducibly normative reasons; they must suspend judgment on whether there are any normative reasons; and they must suspend judgment on what the substance of normative reasons would be if there were any. Having put themselves in this position, they may then consider the normativity wager. My prediction is that they will immediately acquire a motivation. I can attest that this is my experience when I put myself in the position of the Pyrrhonian normative skeptic. If an agent attempts this process and does not acquire such a motivation, it may be that they have failed to successfully put themselves in the position of the Pyrrhonian normative skeptic; they have misunderstood or not fully reflected on the wager; or they have been distracted from the question. If the agent is quite confident that none of these apply, and they still have not acquired a motivation, they may have uncovered evidence against my prediction. In any case, though I do not claim they have a reason to do so, nor that it is rational to do so, I invite all normative Pyrrhonian skeptics to consider the wager, because I both believe and hope that it will prompt them to acquire the motivation to do (A). I promote the normativity wager, banking on the assumption that if we are in a world with irreducibly normative truths, the world is more likely to go better if normative Pyrrhonian skeptics acquire a motivation to attempt to act in accordance with any normative reasons to which they might be subject.