Introduction
One reason the phenomenon of akrasia has been of such lasting interest to both action theorists and epistemologists is that philosophers have thought that the possibility of akrasia can tell them something important about the structure of agency more generally.
The existence of practical akrasia—cases in which an agent acts contrary to her considered judgments about what she has most reason to do—has been held to constitute an important form of abductive evidence for the existence of some agential capacity, exercised when acting for reasons, whose failure or weakness in the case of akrasia makes it possible for the agent to act contrary to her evaluative judgments about the force of her reasons.
Just as the existence of practical akrasia has been treated as important evidence for the existence of our practical agency, the alleged absence of epistemic akrasia—cases in which a believer believes some proposition contrary to her considered judgments about what she has most reason to believe—has recently been marshaled as ground for skepticism about the existence of similar forms of epistemic agency (see, e.g., Setiya Reference Setiya2013; Adler Reference Adler2002; Owens Reference Owens2017). If we exert our epistemic agency when we form beliefs on the basis of the reasons we judge to bear on believing, it should be possible for our agency to err, and so we should expect believers to fail occasionally to produce the requisite belief grounded in the recognized reasons. The alleged absence of any such agential breakdown appears to signal that there is no agency involved to break down in the first place.
In this paper, I aim to defend the possibility of epistemic agency against such objections. Rather than argue against the claim that epistemic akrasia is impossible, I will argue that the absence of epistemic akrasia is compatible with the existence of epistemic agency. The crucial mistake, I will argue, is that skeptics about epistemic agency are failing to distinguish carefully between differences in the structure of believing and acting and differences in the structure of normative reasons to believe and normative reasons to act. I will argue that differences of the latter sort can provide an alternative—and superior—explanation for the absence of epistemic akrasia to an explanation involving differences in our practical and epistemic agency.
The argument will proceed in three steps. I will begin by articulating an important distinction between the ways that a certain activity may be restricted: restrictions imposed by the internal structure of an agent's capacities and restrictions imposed by the external environment in which the agent is situated. It is only restrictions on our activity of the former sort, I will argue, that have important implications for our agency.
I will then apply the framework of individual and environmental restrictions to intention and belief. I will argue that for reason-responsive attitudes, such as belief or intention, the relevant environmental restrictions are best understood as restrictions whose source lies in differences between the relevant domain of reasons that the attitude is responding to. For belief and intention, these will be differences between the normative structure of truth-given and goodness-given reasons.
In third part of the paper, I will survey the skeptical arguments against the possibility of epistemic akrasia and show how these arguments all rely on environmental grounds, rather than on internal grounds, for establishing their conclusion. They achieve their conclusion by relying on different features of truth and goodness rather than on different features of the attitudes of belief and intention.
I conclude the paper by explaining how this study of akrasia reveals an important cautionary lesson for cognitive science more generally. When extrapolating from differences in patterns of acting and believing to differences in the structure of our agency, we must be careful to be sure that the differences really stem from differences in the relevant psychological structure of our practical and theoretical attitudes, not from differences in the structure of the reasons that those attitudes are tracking.
1. The Link between Akrasia and Agency
In keeping with contemporary convention in discussion of the phenomenon (following Davidson Reference Davidson and Feinberg1969), I will understand practical akrasia as occurring when an agent intentionally acts contrary to her considered judgments about what she ought or has most reason to do. Such cases can be moral in nature (as with the agent who judges that she ought not cheat on the exam, yet cheats anyway) but need not be (as with the agent who judges that he ought to refrain from having another piece of pie, but takes a second slice anyway).
While not a central or paradigmatic instance of intentional action, the phenomenon of akrasia has important consequences for the explanation of intentional action in general. If agents sometimes act contrary to their judgments of what they have most reason to do, this means that there must be some part of the story of intentional action that can explain how such akratic actions are possible. While an agent may usually act as she judges she has most reason to act, the possibility of practical akrasia thus suggests that even in cases in which an agent does act as she judges she has most reason to act, there must be something further going on explaining why her reasoning makes her act in accordance with her reasons on this occasion but not when she acts akratically.
In this way, cases of rational failure like akrasia constitute an important form of abductive evidence for the existence of some agential or volitional capacity that mediates between the agent's judgments that she has most reason to act and her acting or intending to act for those reasons.Footnote 1 Consider a simple mechanical analogy. The fact that a vehicle's wheels occasionally fail to follow the road suggests that there is some steering mechanism at work that usually keeps the wheels and road aligned. In the same way, the occasional misalignment between our judgments of how we have reason to act and how we act is evidence that we have some agential capacity at work securing their usual alignment when we do act for the reasons we judge ourselves to have.
While the occasional philosopher has expressed skepticism about the possibility of akratic action in the past (see, e.g. Plato's Protagoras 358b-c), the majority now tend to hold that akratic action is a real phenomenon and that explanations of action must be able to account for it. At the very least, those philosophers who are skeptical of the existence of akratic action accept that there certainly seem to be cases of apparent akratic action that lend prima facie support to the view that such cases are possible, absent some persuasive theoretical reason to interpret such cases in some other way.
In contrast to the case of action, there continues to be widespread skepticism about the existence of akratic belief.Footnote 2 Parallel to akratic action, cases of akratic belief would involve a believer who believes some proposition contrary to her considered judgments about what she has most reason to believe.Footnote 3 Apparent examples of straightforward akratic believing seem much more difficult to come by than the seemingly abundant examples of akratic action. To many, the very possibility of such cases has seemed implausible; philosophers have sensed ‘a whiff of Moore's paradox’ in the claim that, for example, my evidence supports the conclusion that it is raining, but I do not believe that it is raining (see e.g. Setiya Reference Setiya2013:197).
Just as the existence of akratic action has been seen as evidence for important forms of agency with respect to our actions, the alleged absence of akratic belief may appear to give us reason to be skeptical that we enjoy an analogous sort of epistemic agency with respect to our beliefs. Again, it may be helpful to think of a mechanical analogy. Just as the fact that a car's wheels occasionally fail to turn in the direction of the road might suggest that there is some steering mechanism at work keeping them regularly aligned with the road, positing the existence of a steering mechanism may start to sound suspicious if the wheels and road align too perfectly. Capacities are characterized by their occasional misfire or by the possibility of failure. If we exercise a capacity in the performance of some activity, like steering a vehicle, it should be possible for that capacity to err and so effect the possible outcomes. In fact, if we were to insist there was a steering capacity, we would now be pressed to provide some further explanation for why the purported capacity never went awry.
Similarly, if we exert our epistemic agency when we form beliefs on the basis of the reasons we judge to bear on believing, it should be possible for our agency to err, and so we should expect believers to fail occasionally to produce the requisite belief grounded in the relevant reasons. If we never fail to believe what we judge ourselves to have most reason to believe, that is evidence that the transition from judgment to belief is unmediated. If we still insist that our beliefs are the result of our epistemic agency, we are now pressed to provide some further explanation for why this purported agential capacity never goes awry, producing cases in which we believe against our considered judgments.
Doubts about the existence of circumstances in which we believe the evidence supports a proposition but fail to believe that proposition on the corresponding grounds are in this way closely linked to doubts about the existence or extent of our epistemic agency.Footnote 4 The disanalogy between belief and action with respect to akrasia can be extended to a disanalogy between belief and action with respect to our agency more generally. Thus Jonathan Adler (Reference Adler2002: 1–2) argues for the conclusion that akratic believing is impossible and claims that this conclusion ‘undermine[s] … attempts to assimilate theoretical [and] practical reasoning’ and ‘reveals a fundamental divide between them’. David Owens (Reference Owens2017: 37) argues that ‘epistemic akrasia is impossible and that it's impossibility casts a shadow over the whole idea of doxastic control’. Setiya (Reference Setiya2013: 196) expresses ‘skepticism about epistemic agency’, arguing that the impossibility of believing contrary to your judgments of what you have most reason to believe helps ‘reveal a basic contrast’ between what is involved in acting for reasons and believing for reasons. Whereas acting for a reason is to be explained as the result of some psychological capacity or process of rational causation, which moves the agent from her judgments about how she has reason to act to her acting for those reasons, ‘believing for a reason reduces to a conjunction of beliefs [the belief that p and the belief that the fact that q is evidence that p]. There is no further causality that connects them’.
In order to convince those otherwise sympathetic to a more robust sense of epistemic agency of his deflationary account, Setiya marshals precisely the sort of abductive argument from the impossibility of akratic believing described above:
it follows from your beliefs about the evidence that p that you believe it on corresponding grounds. What accounts for these phenomena if believing for a reason is not a mere conjunction of beliefs? (Setiya Reference Setiya2013: 198)
In Setiya's paper, the alleged absence of akratic believing is just one among several important disanalogies between belief and action from which he argues for his deflationary conclusion. This argument from akrasia is thus just one example of a larger methodological trend in the growing body of work at the intersection of ethics and epistemology engaged in comparative analysis of practical and theoretical attitudes and the related apparatus of practical and theoretical reasoning. Theorists will look at the differences between the phenomenon of believing and acting or intending and extrapolate from this to differences between the structure of belief and intention or intentional action. In this instance, theorists are using the alleged absence of akratic belief as the grounds for an abductive argument against epistemic agency. The best explanation of why we never produce beliefs that run counter to our judgments about what we have most reason to believe, in the way we sometimes intend actions that run counter to our judgments about what we have most reason to do, is that we lack the relevant sorts of agential capacities, enjoyed in acting, that would make such beliefs possible.
Given the implausibility that humans exert a universally iron will over their epistemic agency so as to never be swayed or tempted to believe counter to their reasons, the lack of such agency may appear to be the best, and perhaps only, candidate explanation for the complete absence of akrasia in the doxastic realm. As Setiya puts the point in the passage quoted above: ‘what [else] could account for the phenomena?’
The rest of this paper is an attempt to provide a satisfactory answer to this question. I will argue that there are actually two possible kinds of explanations for restrictions on attitudes, one stemming from our agency with respect to that attitude, the other stemming from the environment in which the agent is situated. I will further argue that environmental restrictions can better account for the absence of akrasia in the doxastic realm than a lack of epistemic agency.
In making these arguments, I will attempt to maintain an ecumenical approach toward what it is to be an epistemic or practical agent. Notice that in the mechanical case we are justified in making the abductive inference about the presence or absence of a steering capacity from the presence or absence of turning activity even if we lack any mechanical knowledge about the details of how a typical car's steering mechanisms function. In the same way, the absence of epistemic akrasia threatens to cast doubt on the existence of a corresponding mediating agential capacity regardless of how such agential capacities are ultimately understood. As with the skeptical argument it is responding to, the following discussion will not presuppose any view on what the relevant agential capacity involves or even, for that matter, that one accepts the link between akrasia and agency at all. My central goal is rather methodological: I hope to show in this paper that even if one did hold that the capacities that make akrasia possible are essential to agency, the absence of epistemic akrasia would still not provide reason to doubt the existence of our epistemic agency.
2. Agency and the Environment
Consider again the mechanical analogy to akrasia. Suppose we encounter a car that, unlike other cars, never turns or swerves from its route. What are the possible explanations?
There are, broadly speaking, at least two potential sources of explanation for the absence of such activity. The first source is what we might call an internal restriction. The object might fail to exhibit certain activity because it lacks some of the internal structure required for the activity in question. If we see a car that never turns, for example, one explanation may be that it lacks the internal structural steering mechanism required for turning.
The second source is what we might call an external or environmental restriction. An object might fail to exhibit certain activity because something about its external environment prevents the activity from occurring. Though the object has the necessary structural mechanisms or capacities to perform the activity, the exercise of the capacity is curtailed by the environment in which the object is situated. In our mechanical example, it may be that the course on which the car is driving has no side streets on which to turn. If a car were located on such a track, the complete absence of any turning activity would be no evidence for the absence of a steering capacity. Though the driver may retain the capacity to steer as she ordinarily would, the capacities’ expression through turning activity is curtailed, or masked, by the environment in which the car is situated: there are simply no alternative paths for the driver to steer the car on to. (A related point is made by Lewis [Reference Lewis1997]. We cannot characterize capacities, or robust dispositions, in simple counterfactual terms because the environment may ensure that there is no counterfactual scenario in which the disposition is actualized though the individual possesses it.)
More generally, if we know that there exists an environmental restriction, the absence of activity no longer provides any strong abductive evidence for the absence of a mechanism. Were there such a mechanism, we should expect its exercise, in this case, to be masked by the external environment. If we have some antecedent reason to believe such a capacity existed, we should continue to believe the capacity exists.
Let us now return to the case of belief. What are the possible sources of explanation for the absence of cases in which a believer believes contrary to her considered judgments about what she has most reason to believe?
One possible source would be the sort of internal restriction posited by Setiya. There could be some internal agential capacity exercised when a person acts in response to her reasons, a capacity that is absent when a person believes for a reason. Because there is no capacity at work when an agent believes for reasons, the subject is unable to exercise that capacity to believe against what she judges she has most reason to believe.
What, if anything, would be the relevant possible environmental restrictions on believing? The answer is less straightforward. There is no obvious analogue to the car's track for attitudes like belief. With some work, however, an analogy can be formulated. Beliefs and other reasons-responsive attitudes are often understood as evaluative commitments. Intentions (and intentional actions, insofar as they embody similar commitments) are commitments about what to do; beliefs are commitments about what is so. Just as cars are navigating the road, these evaluative commitments are attempting to navigate some normative terrain.
This analogy suggests an answer to our question: while internal restrictions involve how an attitude responds to reasons, the relevant environment for belief and intention consists of the accessible normative reasons that the attitude is responding to. Roughly: in the case of commitments about what to do, we are responding to practical reasons in order to evaluate actions as good or bad. In the case of commitments about what is so, we are responding to epistemic reasons to evaluate propositions as true or false. Explanations that rely on differences between actions and propositions would involve differences not in how we reason, so to speak, but in what we are reasoning about—differences in the external facts about true propositions and good actions that are the objects of our beliefs and intentions, rather than differences in the internal structure of believing and intending.
While discussion of the environment may suggest a strongly externalist picture of reasons, no such assumption is required. What is essential to the analogy is that we can distinguish how behavior may be restricted either by features of our accessible reasons or by features of our agential responses to those reasons. This distinction is compatible with either an internalist or externalist picture of the reasons-in-question. I will refer to such restrictions as external or environmental because just as differences in external environmental factors, such as the lay of the road to which the car is reacting, can mask the car's steering capacity, features of our accessible reasons, to which our attitudes are responding, might mask our agential capacities.
Environmental restrictions, then, would be restrictions about what we can believe or intend, whose source lies in differences between the relevant normative structures of goodness-given and truth-given reasons. These normative restrictions are environmental in that they, like a physical environment, can mask the existence of internal agential capacities. This is perhaps easiest to see by illustration. To see such normative environmental restrictions at work, consider a case from the philosopher Joseph Raz concerning our capacity to act-for-reasons. According to Raz (Reference Raz2001), there are a number of restrictions on when it is possible to perform certain actions for certain reasons. I can decide to go to one play over another for the reason that I love Sophocles, for example, but I cannot decide to have coffee over tea for the reason that I love Sophocles.
Does this reveal some important difference in my capacity to act for reasons with respect to plays over beverages? Raz thinks no. In each case, my capacity to act for reasons is the same—it is a capacity to act for reasons that show the action good or desirable in some way so as to make the action intelligible. In both my beverages and play choices, I am equally capable of being motivated by any reason I take to bear on the desirability of the activity in question. It is just that a love of Sophocles is relevant to the goodness of one play over another, but not to the goodness of one beverage over another, so there will be no such reason available to motivate me.
In this case, a certain absence (cases of deciding to get coffee for the reason that we love Sophocles) does not cast doubt on our agency with respect to beverage choices because we have an explanation of the absence in terms of differences in normative environment. The absence is the result of differences in what is good about the kinds of actions being intended, not a difference in the agency exerted in intending to do the actions. It is, so to speak, a difference in the object of the attitudes—a difference between plays and coffee—rather than a difference in the attitudes themselves, that explains why I cannot perform certain actions out of a love of Sophocles.
Another example can be found in Schroeder's ‘surprise party’ cases (Reference Schroeder2007). My love of being surprised will never motivate me to head to a surprise party being held for me even though my love of other activities will motivate me to go to the activity. This does not reveal any important limitation on the nature of my agency: my intentions are not somehow practically unresponsive to the fun of surprise parties. It is because of the nature of what I find good about surprise parties: to be surprising, I cannot be aware of them, so there will never be anything about the good situation for my regular internal motivational structure to respond to. Such external features of the object of my intention, rather than features of the attitude itself, provide environmental explanations for the absence of such attitudes.
While the above examples involve actions or intentions with different objects, we can apply the same lesson to comparative research concerning different kinds of acts or attitudes, such as comparisons between belief on the one hand and action or intention on the other. Reasons for belief are truth-governed. Reasons for action are good-governed. When extrapolating from differences in when we act and believe to differences in the nature of action and belief, we should take care to make sure that the difference is in the acting and believing, not in the external facts about when reasons to believe and reasons to act are available.
This opens up at least the possibility of an environmental distinction that could explain the absence of epistemic akrasia without calling into question the existence of epistemic agency. In the following section, I will argue that if akratic belief is impossible, it is this environmental diagnosis that provides the most plausible explanation why. Insofar as we have a convincing argument that epistemic akrasia is impossible, the argument will have to rest on differences between the structure of goodness-given and truth-given reasons, not on differences between the structure of intention and of belief. In particular, I will argue, recent objections to the possibility of epistemic akrasia rely heavily on the fact that two contrary propositions cannot both be true, in a way that two contrary actions might both be good.
3. The Arguments Against Akratic Belief
Those skeptical about the possibility of epistemic akrasia claim there is something paradoxical about the possibility of one's simultaneously judging that the evidence supports the conclusion that it is raining, for example, yet believing that it is sunny, in a way in which it is not equally paradoxical to judge that one's reasons count in favor of staying in to grade papers, yet decide to go get ice cream in the park (see especially Pettit and Smith [Reference Pettit and Smith1996: 448]; for similar argument, see Adler [Reference Adler2002: 6–7]; Setiya [Reference Setiya2013]). In this section, I will survey the variety of objections offered by recent skeptics against the possibility of believing akratically by the same methods by which we achieve akratic action. Rather than focusing on evaluating whether such objections are successful, I will instead aim to show that each objection, successful or not, is best understood as grounded in environmental, rather than internal, limitations. Since these objections all rely on environmental limitations to reach the conclusion that epistemic akrasia is impossible, the impossibility of such akrasia will be an inappropriate starting point for arguments that attempt to move from the absence of epistemic akrasia to the absence of epistemic agency.
Following the general Davidsonian line on the origins of practical akrasia (Davidson Reference Davidson and Feinberg1969), recent skeptics begin with the observation that actions and their outcomes can be good in a variety of different ways—they might be moral, or pleasurable, or intellectually stimulating—and these different values can give rise to a variety of conflicting reasons for acting. This variety of ‘goodness-related’ reasons makes akratic action possible. When you act in a way that you judge you should not, you can still be responding to the appeal of some genuine good even when you might have made an all-things-considered judgment that this good is outweighed by some other practical considerations. The more important reasons you have to grade your papers do not make the ice cream taste any less delicious. And so the pleasure-related reasons you have can still make your akratic action intelligible by showing how you see it as good or desirable in some respect even though you think you have stronger reasons to stay in and grade.
In contrast to the variety of sources of reasons for action, Susan Hurley (Reference Hurley1993: 133) observes that truth, unlike goodness, is a univocal value. Reasons for belief are united in stemming from this single value. Hurley uses this feature of truth to try to explain why it is impossible to believe akratically by the same method as we achieve akratic action:
In the case of what should be done there may be conflict within an agent. There may be conflicting reasons competing for authority. But in the case of what should be believed, truth alone governs and it can't be divided against itself or harbor conflicts. It makes sense that something is, ultimately, good in some respects but not in others … in a way it does not even make sense to suppose that something is, ultimately, true in some respects but not in others.
Hurley's insight is that the unity of truth-given reasons might deny the believer any reasons from which to motivate her akratic believing. An action or outcome can be good in some respects, bad in others. Even if I hold the professional reasons for grading to outweigh the countervailing reasons of pleasure, the pleasure-given reasons for ice cream still hold some pro tanto force. But since truth, unlike goodness, is a univocal value, there are no multiple sources of value to provide us with countervailing reasons not to believe as we, all-things-considered, judge we ought to believe. And so there are no available epistemic reasons, even outweighed reasons, ever to motivate our akratic believing, as there are often reasons available to motivate our akratic actions.
Some philosophers, while skeptical about the possibility of epistemic akrasia, have claimed that yet further argument is required to establish epistemic akrasia's impossibility. It is not enough to say that the reasons for belief come from a single value, while goodness involves a variety of values. For as Raz (Reference Raz2007: 10–12) points out, there can be conflict among reasons even given a single value. Even on simple hedonic accounts of the good, for example, in which pleasure is the sole value, there can be conflicts among my pleasure-given reasons. I may face interpersonal conflicts between the pleasure of two friends, each of which I might promote. I may face temporal conflicts between my current pleasure of the cake and the future pleasure my good health would help promote.
Though an absence of divided reasons might be sufficient to explain why akratic beliefs cannot get off the ground, by denying a believer the motivating reasons that allow akratic action to emerge, Raz argues that we need some further explanation for why we think the reasons for belief are so united. For the fact that truth is a single value seems by itself insufficient, given the possibility of divided reasons stemming from other unitary values, such as pleasure. However, according to skeptics of akratic believing, such as Adler (Reference Adler2002), Raz (Reference Raz2007), and Owens (Reference Owens2017), a further explanation is forthcoming.
The central thrusts of their arguments are strikingly similar. Here is Owens (Reference Owens2017: 45):
If one thinks the evidence establishes p, one must think that apparently countervailing evidence e′ can be explained on the hypothesis that p and so provides no grounds for thinking not-p to be true. Should one nevertheless be swayed by the appearance of e′, one is being swayed by a consideration whose probative force one can't acknowledge in judgment.
From Raz (Reference Raz2007: 7):
Epistemic reasons can conflict, but all of them are about the truth of the propositions for or against belief in which they are reasons. The weaker reasons are just less reliable guides to one and the same end. … Because there is no possibility that the lesser reason for belief serves a concern which is not served better by the better reason there is no possibility of preferring to follow what one takes to be the lesser reason rather than the better one. The possibility of [practical] akrasia depends on the fact that the belief that a practical reason is defeated by a better conflicting reason is consistent with belief that it serves a concern which the better reason does not, and which can motivate one to follow it.
And again from Adler (Reference Adler2002: 6–7):
When beliefs conflict, they weaken one another, since both cannot be true. When one belief is favored by the evidence, the disfavored belief evaporates, since it has been determined to be false. But when desires conflict, as with desires to pursue careers both in medicine and in ballet, the conflict need not, and typically does not, weaken either. When one desire is acted upon, the other retains a hold, experienced as regret … let us extend [this] observation from beliefs to the evidence or reasons for them. When evidence is adequate … then we accept or fully believe it. Consequently, and this is the crucial claim, previously conflicting evidence (i.e., evidence that supported a contrary of h) is nullified as undermining h.
In each passage, the author trades off what he takes to be an important difference between reasons for belief and reasons for action: that while sufficient reasons for acting may outweigh the reasons for contrary actions, reasons for believing, when sufficient, undercut or nullify any possible reasons for believing contrary propositions.
According to Owens, Alder, and Raz, Hurley is right in locating this feature of reasons for belief as stemming from a distinctive feature of truth, just not the one Hurley appeared initially focused on. What establishes the unity of reasons for belief is not just that the same proposition cannot be true in some respects and false in other respects, whereas, because of the plurality of goodness, the same action can be good in some respects and bad in other respects. It is instead a second distinctive feature of truth: that only one among a set of competing propositions can be true. This distinguishes truth from other unitary values such as pleasure. Though pleasure is a single good, it is a good that can adhere, without contradiction, in multiple objects or multiple outcomes.
It is thus the principle of noncontradiction, not the principle of bivalence, that plausibly generates the appropriate reasons for belief to prevent epistemic akrasia. Because truth cannot contradict truth, if some proposition p is true, then any true considerations that seemed to indicate that p was false must in fact be compatible with p after all. While ice cream may still taste sweet and so be good in some respect even though I know I should be studying, it cannot still be true in some respect that it will rain, for example, if the truth is that it will remain sunny outside. Thus, my alleged evidence of rain (e.g., the dark clouds in the sky) cannot really be foreshadowing future rain, if there is no rain to foreshadow. The clouds must actually be compatible with its being sunny later. Even if I do not yet see how my apparently contrary evidence is compatible with what I believe to be true, I see that the evidence must not, in fact, have the normative force I had taken it to have.
Having lost my reasons for believing not-p, the argument goes, I have lost any possible motivating grounds for adopting the belief. My believing not-p is thus incompatible with my simultaneously judging that the evidence supports belief in some contrary proposition, and so Moore-paradoxical in precisely the way skeptics of epistemic akrasia had claimed.
With the arguments for the impossibility of epistemic akrasia on the table, we can now assess whether, if they were sound, their conclusion would be an appropriate abductive starting point for skepticism about epistemic agency. I think, in this case, the answer is clearly ‘no’. As in the case of the surprise party and the Sophocles enthusiast, it is clear in the arguments presented above that it is the particular nature of truth and truth-given reasons, not the particular nature of belief, that explains why epistemic akrasia is impossible when practical akrasia is not. According to the skeptical arguments surveyed, beliefs and intentions are similarly structured in that the object of their respective evaluative commitments constrains the reasons for the attitude. They differ only in whether truth-given or goodness-given reasons for akratic actions or attitudes are available for the agent to hold. Akratic believing is impossible, according to the argument, not because we as believers lack any capacity to be tempted by reasons for believing propositions we judge we ought not believe, but because there will never be any such reasons to tempt us.
4. Objections
In the previous section, I have shown that insofar as we have good arguments for the conclusion that epistemic akrasia is impossible, these arguments rely on an external, environmental source. Epistemic akrasia is impossible because of distinctive features of truth and truth-given reasons, not because of distinctive features of our internal capacities as believers to respond to such reasons.
Given the environmental source of the absence of epistemic akrasia, the absence cannot be grounds for doubting that we have epistemic agency. The absence would not reveal, as proponents have sometimes thought, a fundamental divide between the agential structure of our practical and theoretical attitudes. Moreover, given general grounds for holding that believing-for-reasons and acting-for-reasons are otherwise analogous, we might plausibly hold that we do have such epistemic agential capacities (see Flowerree [Reference Flowerree2017] for an example of a parity argument for epistemic agency along these lines). Those previously inclined to posit agential epistemic capacities could acknowledge the skeptical arguments for the absence of epistemic akrasia and hold that the absence of epistemic akrasia is the result of these capacities being constrained, in the case of belief, by their normative environment. In this section I will consider and respond to two further objections one might raise even if one accepts my diagnosis of the environmental source for the absence of akratic believing.
4.1. Environmental Constraints on Capacities
I will begin by considering an objection to my claim that discovering an environmental source for the absence of an activity removes any grounds for skepticism about the presence of a capacity to produce that activity. One might worry that environmental restrictions on the exercise of a capacity, if sufficiently global, would make it puzzling why such a capacity would have come to exist. We would need an explanation for why we have evolved an internal capacity to perform an activity the external environment prevents us from performing.
One may object, then, that we have overlooked an important disanalogy when moving from the case of the car on the track to the case of akratic belief and the normative environment. Unlike the contingent features of the track that prevent turning activity, the features of truth-given reasons that prevent akratic believing appear conceptually necessary. Perhaps if there were no possible environment in which the car could turn left or right, the environmental explanation for the absence of turning behavior would still be sufficient to rule out the existence of a steering capacity. It may seem odd to posit a capacity whose every exercise is necessarily masked in any possible environment in which it might be situated.
While I agree there is something strange about a capacity that it is, perhaps conceptually, impossible to exercise, I think it is a mistake to hold that if akratic believing is impossible, epistemic agential capacities fall into such a category. This is because akratic beliefs would not be the only exercise of our epistemic agential capacities if such capacities were to exist, any more than akratic action is supposed to be the only time our practical agency is exercised. While akratic action might be especially important for revealing the existence of a mediating agential capacity, it is, in fact, a rather peripheral case of intentional behavior. The typical exercise of our practical agency will be cases in which our mediating agential capacities are exercised so that we act in accordance with the reasons we judge ourselves to have. Likewise, if we do have analogous epistemic capacities, these capacities will be exercised in the actual environment every time we believe for the reasons we judge ourselves to have.
Thus, the skeptics’ objection is not (or ought not to be) that in the absence of epistemic akrasia there is no behavioral evidence for an agential capacity. Every instance of believing for the reasons we judge ourselves to have is a possible exercise of an agential capacity, just as (presumably) our acting for the reasons we take ourselves to have is an exercise of a practical agential capacity. The problem is that for belief (unlike action) all the available behavioral evidence is equally good evidence for an agentially unmediated capacity that directly moves from judgments to beliefs. What the normative environment ensures (and so explains) is the absence of any distinguishing behavior: doxastic behavior that could only be produced by an agentially mediated, rather than an agentially unmediated capacity. The differences between truth and goodness provide a normative environment for belief that ensures that even if we had a mediating agential capacity, such differentiating behavior would never be exhibited.
Once we see that it is only the distinguishing behavior that is masked by the environment, we can accept that this environment may be quite global without this causing any questions about why an internal capacity with the power to cause such activity would have evolved. The answer is that akratic believing is not an activity we evolved to perform. We would have developed the capacity in order to believe for reasons generally. Since this behavior is produced in our actual environment, it can explain why we may have evolved such agential capacities even if the capacity will never produce the distinguishing behavior that we would have expected it to produce if the environment were different.
We can see actual examples of this distinction between distinguishing behavior of a capacity and the evolved-for behavior in various cases of parallel and convergent evolution. Consider the case of the convergent evolution of various species of mangrove trees. Mangroves often grow in brackish water of various degrees of salinity. Since salt is toxic to trees, different species of mangroves, in the same environment, have evolved different mechanisms for salt tolerance. Black mangroves use ion transporters to excrete excess salt sucked up in the root system, while red mangrove trees use a root filtration system to prevent salt from entering the root system in the first place. While each of these mechanisms is equally effective at producing salt tolerance, one of the capacities has a characteristic side effect: In water with sufficient salinity, a side effect of the ion transporter mechanism (but not of the filtration system) is the accumulation of salt crystals on the leaves of the plant.
Suppose we are wondering if a particular tree is a black mangrove. In an environment with no salinity, we might be rightly curious why we should posit the existence of ion transporters in the absence of any possible behavior that the transporters could cause. Moreover, given their characteristic side effect, if we find trees in an environment where the water has a sufficiently high degree of salinity, and we never find accumulated salt crystals, this is strong abductive evidence that this mangrove is using a filtration mechanism. If it were using an ion transporter mechanism, we would expect to see at least the occasional salt crystal accumulated. The absence of salt crystal accumulation in high salinity environments might serve as a defeater even if we had some other prior reason to suspect the tree was a black mangrove rather than a red mangrove.
However, if the mangroves are growing in a body of water with a lower degree of salinity, the absence of salt crystals will no longer provide any evidence to distinguish between the two mechanisms. There is some evidence for each mechanism—the trees’ ability to grow in salt water is evidence that they have one of the two capacities. But the fact that the distinguishing behavior is absent provides no reason to posit one capacity over the other once we can explain the absence in terms of environmental features. And insofar as you had some antecedent reason to favor the hypothesis of black mangrove over red mangrove, you should continue to believe these are black mangroves even though, if the water were of a higher salinity and the salt crystals were absent, that absence would have given you reason to abandon your belief.
The case of convergent evolution is not perfectly analogous to the case of akrasia because while we might imagine a global low-salinity environment, such environments are not necessary. However, I think we can come up with further cases in which it seems plausible to posit a capacity even if the features of the environment masking the distinguishing activity are necessary ones. Consider a modified version of our previous mechanical example. Suppose I buy a blimp from my mechanic. My blimp might have motor A or motor B, each of which power blimps, and does so equally effectively. Suppose motor type A makes a whirring sound when working underwater, but motor type B does not. It is, plausibly, impossible that my blimp would ever whir underwater, since it is physically—and let us suppose conceptually (it would cease to be a blimp if it ran underwater) —impossible for the blimp to inhabit such an environment. This does not seem to show that the absence of whirring behavior of blimps in the sky makes an attribution of motor type A any less plausible than an attribution of motor type B. If so, the similarly necessary absence of akratic belief should not make attribution of underlying mediating agential capacities any less plausible.
4.2 Environmental Constraints on Agency
Suppose that, given the environmental source of the absence of akratic belief, one accepts that we can reasonably posit agential epistemic capacities constrained by their normative environment. One might still worry that such a constrained capacity to believe for reasons will constitute something of a Pyrrhic victory. For one might object that as freedom requires the possibility to do otherwise, an agential capacity that can never be exercised to override our judgments about what reason requires is no better than the absence of a capacity as far as our agency is concerned. Perhaps our lack of epistemic agency is not a lack of some internal capacity, but rather just consists in our (in this case environmental) inability to do otherwise than what epistemic reason demands.
In this section, I will argue against this response by considering a variety of Putnam's (Reference Putnam1973) Twin Earth thought experiment. Suppose we create a normatively sparse world in which the structure of an agent's reasons to act mimics her reasons to believe so that she would never have available reasons to act as she judges she has most reason not to act. Suppose we move an individual from our world to this new world. I contend that it is not clear that in such a world practical akrasia would still be possible. However, we would be wrong in thinking that this agent's actions are no longer the result of her agency. By parity, I will argue that in the actual world we should not doubt that a subject's beliefs are the result of her agency.
To create a world with such restricted reasons to act, we can imagine a world in which some powerful being has arranged the environment so that the goodness of states of affairs is structured in the same way that the truth of propositions is structured in the actual world.
How would a powerful being construct such a world? To mimic the structure of truth in the actual world, this would have to be a world arranged so that the variety of goods aligned and so ‘spoke with one voice’ as truth does. Imagine, for example, the powerful being has arranged the world so that every moral action would always be in your own prudential self-interest; the powerful being has arranged it so that donating to famine relief really will always come back to help make you richer later on. Our imagined world would also have to be a world in which an action that is good in some respect could not be bad in that same respect. Imagine, for example, a world designed so that every pleasurable activity in the short term would also be in the interests of your long-term happiness; if a cupcake tastes good, it has been arranged so that it will also be healthy for you. Finally, to mimic the structure of evidence in the actual world, our imagined world would have to be a world in which having the most practical reason to act in one way means that any reasons an agent thought they might have had to act another way will be undercut rather than outweighed. This would be a world arranged so that, if you have most reason to choose to go to the cinema rather than the park, it will turn out that it will be raining, so there would be nothing enjoyable about the park to regret missing out on when choosing to act as you had most reason to act.
With the environment structured so that good actions are fully good and alternative actions fully bad, we will have created a world in which there will always be one available course of action that will best attain all the variety of goods we might be interested in. Imagine we pluck a full practical agent from our world and place her in this new Panglossian world. What would happen to someone lucky enough to find herself knowingly in such an environment? There would never be any case in which the reasons for two actions were counterbalanced nor would there be any case in which an action one ought not to do had anything to be said for it that an agent could take as her motivating reason to perform that action akratically. On the Davidsonian picture of akrasia assumed by the skeptics of epistemic akrasia, an agent in such circumstances might be no more able to act contrary to what she judges she has most reason to do than she would be able to believe contrary to what she judges she has most reason to believe. Though her will is just as susceptible to temptation as ever, her new environment is structured so as to lack any reasons to be tempted by and thus motivate her to act akratically.
Though her ability to exercise her agential capacities may have been impacted, her agential capacities themselves have not. And surely it would be strange to think that a previously free agent with a capacity to act at will, placed in such alternate circumstances but intrinsically unchanged, has somehow lost her agency. We should not say she has suffered a change to the structure of her will. Rather, what we should say is that her environment is structured in a way that ensures that she will always exercise her will so that she acts in accordance with what she judges herself to have reason to do.
If this is right, we ought to say the same thing about ourselves as epistemic agents in the actual world. It is not that we lack epistemic agency, we just find ourselves in a normative environment that is structured in such a way that we will always exercise our epistemic agency to believe in accordance with what we hold ourselves to have most reason to believe.
While the twin earth thought experiment above imagines a normatively restrictive practical environment, a similar argument might be made by imagining a world with a more normatively expansive epistemic environment, in which the normative reasons for belief mirror our practical reasons for acting. Though such cases are more difficult to construct, they are not impossible. While it is a conceptual truth that no proposition can be both true and false, we could imagine a world in which a powerful being makes it the case that whatever we believe comes true, so that a believer could have reasons showing that either a belief that p or a belief that not-p would both be true, if she were to believe it. Suppose we pluck a believer from our world and knowingly place her in this new normative environment. It seems plausible (at least given the skeptical grounds provided in section 3), that this agent could have truth-given reasons to believe either p or not-p, regardless of what (if anything) she judged herself to have most reason to believe. There is, of course, much more that would have to be said about such cases of self-fulfilling belief to make this revised version of the thought-experiment complete, and I do not hope to address everything here (I discuss such cases in more detail in Antill [Reference Antill2018 and Reference Antill2019]). However, thinking about such cases of self-fulfilling belief in the present context can help explain why these sorts of seemingly obscure cases are of particular importance to epistemology more generally. These cases are ways of imagining alternative normative environments by changing our available reasons for belief, and so they provide us with a methodological avenue for distinguishing whether a certain feature of how we believe is due to the nature of belief or due to the nature of the theoretical reasons typically available to believe.
5. Conclusion
In this paper I have defended the existence of epistemic agency against a threat to that agency coming from the alleged absence of epistemic akrasia. Rather than showing that epistemic akrasia is possible, I have argued that the link between agency and akrasia has been overstated. This is because the possibility of akrasia rests not only on an individual's agency but also on the environment in which she is situated. She must have some motivating reason for choosing to act contrary to her considered judgments. If her environment is such that she lacks any motivating reasons, akrasia might be beyond her, even though she has the agential capacities to act on such motivating reasons if they were available.
I have then surveyed the arguments for the impossibility of epistemic akrasia and have shown that they reach their conclusion by relying on just such environmental factors. Epistemic akrasia is impossible, opponents allege, not because believers are incapable of being motivated to believe akratically, but because they lack any available reasons to be motivated by. Given that epistemic akrasia, if impossible, is impossible because of environmental factors, the absence of epistemic akrasia is not actually grounds for skepticism about the extent or force of our epistemic agency.
