Objectivism and Subjectivism in Epistemology

Objectivism and Subjectivism in Epistemology Clayton Littlejohn King's College London Abstract: There is a kind of objectivism in epistemology that involves the acceptance of objective epistemic norms. It is generally regarded as harmless. There is another kind of objectivism in epistemology that involves the acceptance of an objectivist account of justification, one that takes the justification of a belief to turn on its accuracy. It is generally regarded as hopeless. It is a strange and unfortunate sociological fact that these attitudes are so prevalent. Objectivism about norms and justification stand or fall together. Justification is simply a matter of conforming to norms. In this essay, I shall make the case for objectivism about justification. 1. Introduction I will defend two objectivist views. The first is a view about epistemic norms. Objectivists about norms believe that some norms have objective application conditions (i.e., conditions that don't supervene upon our non-factive mental states or the subjective aspects of our mental lives). The second is a view about epistemic justification. Objectivists about justification believe that justification depends upon whether certain objective conditions obtain. If, say, some objective condition is met and there's a norm that says that you shouldn't believe that it's raining if this condition is met, you couldn't justifiably believe that it's raining because this condition is met. It might not seem that the condition is met. You might have evidence that leads you to think that you violate no norms, but your belief still wouldn't be justified because you violated this norm. Norms identify the conditions under which someone should or should not believe, do, or feel something. Objectivists and subjectivists agree that some norms have subjective application conditions (e.g., if there's a norm that requires probabilistic coherence there is a norm with a subjective application condition). Their disagreement is about whether there are any further norms that require us to believe or refrain from believing when objective conditions obtain. Their disagreement is about norms like these: You shouldn't believe p unless you know p (Only Knowledge).1 You shouldn't believe p unless p is true (Only Truth).2 Objectivists will have their disagreements about these norms (e.g., some would argue that Only Truth needs to be supplemented by further norms and some would argue that Only Knowledge delivers the wrong verdict in Gettier cases), but they wouldn't think that these aren't genuine norms just because we refer to something objective in 1 For defenses of knowledge norms for assertion, belief, or reasoning, see Benton (2011), Bird (2007), Littlejohn (2013), Smithies (2013), Steglich-Petersen (2013), Sutton (2007), Turri (2011), and Williamson (2000). 2 If Only Knowledge governs belief, so does Only Truth. For defenses of Only Truth that aren't intended to support Only Knowledge, see Boghossian (2008), Nolfi (2015), Wedgwood (2002, 2013), and Whiting (2013). For criticism, see Brown (2008), Gerken (2011) and McKinnon (2013). For a helpful overview of the extant debate, see McGlynn (2014). specifying their application conditions. Subjectivists, on the other hand, insist that all genuine norms have subjective application conditions, conditions we specify by reference to subjective aspects of your mental life. Normative evaluation, they'll say, should always be concerned with relations between a subject's attitudes or actions and further features of our subjective mental lives. An action or attitude is justified when it is proper, acceptable, permitted, right, or appropriate. If you have an attitude or perform an action and it's not the case that you shouldn't, the attitude or action counts as justified. Norms identify the conditions under which you should or should not have an attitude or perform an action, so we can characterize justification in terms of norm conformity and violation. A belief is justified iff it violates no epistemic norms, which is to say that it is justified iff the thinker conforms to the epistemic norms. If we think of justification in this way, objectivism about norms and justification should stand or fall together.3 In what follows, I'll offer arguments for Only Truth and Only Knowledge. I'll also argue that our beliefs are justified only when they constitute knowledge. In the paper's final section, I'll consider a recent subjectivist response to these kinds of arguments. I don't have space to address most of the extant objections to the objectivist views defended here and don't have the space to say everything that should be said in support of the arguments for objectivism. My hope is that the reader will agree that these arguments are interesting, potentially important, and worth further discussion. 2. Objective or Subjective? Let's start with a simple argument for objectivism about justification: An Argument for Objectivism O1. Your beliefs are justified iff you violate no epistemic norms in holding these beliefs. O2. You violate no epistemic norms in holding these beliefs iff these beliefs constitute knowledge. 4 3 While Littlejohn (2012), Sutton (2007), and Williamson (forthcoming) all defend views on which beliefs are justified only when they don't violate objective norms, most of the writers mentioned above won't defend objectivism about justification by appeal to some objectivist view about norms. 4 There is an interesting question about (O2) that we don't have space to discuss in detail. Is it possible to know p and violate some epistemic norm that requires you not to believe p? While it's an interesting question whether there could be cases in which someone shouldn't believe what they know, answering this question won't help us settle the debates between the objectivists and subjectivists. There are some epistemologists who think that it's just obvious that anything you know you justifiably believe because they think that justification is a part of knowledge (whatever that means). This is not my picture of justification. On my account of justification, the justification property is a property that a belief has iff it violates no epistemic norm (or, more carefully, does not do so without sufficient reason). Because justification is not part of the nature of knowledge, it isn't trivially true that anything known is thereby justifiably believed and thereby believed without violating any epistemic norms. On my view, knowledge is a relation between a thinker and a fact, a relation that holds between some thinkers and some facts where the thinker's beliefs could not have the OC. Your beliefs are justified iff they constitute knowledge. If the argument is sound, readers should take the 'factive turn'. If (OC) is correct, it is possible for pairs of thinkers to process things in just the same way, draw the very same conclusions using the same reasoning after things appear precisely the same to them, and still form beliefs that differ in justificatory status. This suggests that a kind of rational perfection might fall short of ensuring that our beliefs are justified. If you like to talk about luck, you could say that the objectivist view of justification that identifies justified belief with knowledge allows for a certain kind of luck that many epistemologists won't allow for and doesn't allow for a certain kind of luck that many epistemologists would allow for. Many epistemologists are opposed to the idea that a thinker might fail to acquire justified beliefs because they suffer the kind of bad luck that would prevent a rational thinker who reasons impeccably from acquiring knowledge. One reason that I'm opposed to subjectivism is precisely because it implies this kind of luck isn't possible. This knowledge account of justification implies that a certain kind of luck that many epistemologists take to have no bearing on justification can rob you of justification. The lucky connection between a thinker and a fact that robs you of knowledge is one that robs you of justification. In the course of our discussion, I shall explain why I think it's important for our view of norms and justification to register that the kind of accidental connections between thinker and present in Gettier cases prevents you from acquiring beliefs that have good standing. Subjectivists will probably say I've just made three significant errors. Justification (and, possibly, norm conformity) doesn't require truth. It also doesn't require a non-accidental connection to truth. Justification also wouldn't require more than rationality because normative assessment is concerned only with the kind of processing that takes us from appearance to belief. To support their case, the subjectivists might offer this argument: An Argument for Subjectivism S1. If p and q seem the same in all epistemically relevant respects to a perfectly rational person, she would accord the same credence to p and q.5 justification property because these thinkers cannot be held accountable for their attitudes or actions. Our beloved dogs and cats presumably know things about their environments, but I don't think we'd want to say that their beliefs could be justified or unjustified. I think that only thinkers that can be held accountable can have justified or unjustified beliefs. On my picture, the relationship between justification and knowledge is analogous to the utilitarian attitude towards the relationship between right action and optimific actions. A dog's actions might be optimific without being right, but the actions of responsible agents could not be anything but right if optimific. 5 In the context of the paper, it is clear that Huemer intends the antecedent to be read as saying that p and q seem the same in all epistemically relevant respects to the particular thinker whose attitudes we're assessing. He is not interested in whether p and q seem the same in such respects to an external observer who knows, say, that p is true and q is false. It should be noted that (S1) is not just a restatement of the idea that justification supervenes upon appearances. It is also not a trivial consequence of that supervenience thesis. Someone could reject (S1) and still defend that supervenience S2. If a perfectly rational person would accord the same credence to p and q, then p and q have the same degree of justification. S3. The propositions p and q seem the same to our thinker in all epistemically relevant respects. SC. Therefore, p and q have the same degree of justification for our thinker.6 We should read (SC) as implying that the thinker is in a position to justifiably believe p iff she is in a position to justifiably believe q. Because it's possible for the thinker's beliefs in p and q to differ in truth-value when p and q seem to them the same in all relevant respects, this argument threatens (O2) and possibly (O1). In choosing my arguments for (O1) and (O2), I chose arguments that show why it's important to resist arguments like this argument for subjectivism. It's clear that Huemer thinks that facts about how things seem or appear have a great deal of normative significance, but this argument doesn't tell us why we should agree with him. Some would ask us to consider error cases to try to elicit intuitions about rationality to shore up support for the key premises. 7 Some would suggest that appearances determine what evidence we have and look to defend this argument by appeal to some norms that tell us that justification is entirely a matter of how our beliefs fit our evidence.8 The arguments I've chosen should help readers see why I don't think that these ancillary arguments will do much to advance the subjectivist cause. Does this subjectivist argument need support from further arguments? I think so. The argument runs from some suggestions about rational credence to a conclusion about the justification of full belief. Is there any good reason to think that the justificatory status of full belief in p and q would be the same if the subject rationally invested the same credence in these propositions? Lottery cases suggest that there isn't. When it comes to believing lottery propositions and believing what you read in the paper, it seems you should have higher credence in lottery propositions but could only be justified in believing fully what you read in the paper. Of course, it could be said that this is just a distraction since the objectivist and subjectivist can agree that you shouldn't believe lottery propositions and reject (S2). I agree. Once we strip away the questionable assumptions about the relationship thesis. Suppose there's some rational requirement that says, in effect, that Agnes should suspend on p and not suspend on q when things appear a certain way to her. Things could appear the relevant way to Agnes, Agnes could be required to suspend on whether p but permitted to believe q, but it also might seem to Agnes that p and q are the same in all epistemically relevant respects. (Someone could say that this means that Agnes is somehow less than fully rational, but I don't see that. Provided that Agnes doesn't violate these rational requirements, I don't see her as rationally simply because she doesn't appreciate the rational difference between the propositions that she entertains but doesn't believe.) For criticism of (S1) and its use in Huemer's argument for phenomental conservatism, see Littlejohn (2011). 6 Huemer (forthcoming). 7 See Cohen (1984) and Wedgwood (2002). 8 See Dougherty (2011) for a discussion of combining Conee and Feldman's (2004) evidentialist view with phenomenal conservatism. between rational degrees of credence and the justification of full belief, it looks like we're left with nothing but the bare claim that if things appear the same with respect to two prospective beliefs, they must agree in justificatory status. Why should we accept this? It's at just this point that the subjectivist should offer some additional arguments, arguments that show that there's some interesting role for appearances in determining what our evidence is or some interesting role for subjective states in the formulation of norms. My arguments for objectivism are designed to head off these responses. Norms aren't just concerned with appearances and the appearances don't determine what evidence we have. 2.1 The Transcendental Argument My first argument for objectivism is a transcendental argument. The argument starts from the assumption that there are practical requirements to conform to objectivist norms. It proceeds to show that these norms are binding only if there are objectivist norms that govern belief. Consider some plausible objectivist norms: You shouldn't sentence the innocent for crimes they didn't commit (Only the Guilty). You should sentence those you know to be guilty (Any Known Offender). You should repair your past (objective) wrongs (Repair).9 I don't have any complicated arguments for Only the Guilty or Repair. They both strike me as compelling.10 Consider what happens if you punish Agnes for something that she didn't do and discover you've done this later. Bad things happen all the time, but this bad thing is something that you have a special obligation to do something about. You should assume responsibility for this harm and have a duty to make things right even though you might not have a similar duty to respond to all the harms that you happen to cause.11 This suggests there's a reparative duty, a duty to right something that you wrongfully brought about. The duty arises because of a morally significant relation between you and Agnes that doesn't hold between Agnes and other parties who weren't involved who could also take steps to make her better off.12 The fact that 9 An objective wrong involves violating an objectivist norm, such as Only the Guilty. I should also add that if there were a group of people who acted as if these norms were binding, it would be hard to see what could be objectionable about it. We can imagine alternative practices where people acted as if only subjective correlates of these norms would be binding. This might not seem to be a terrible practice. It would, however, be a practice that, to my mind, left out some nice features of the practice that involved striving to conform to Repair. In short, it would be nice if morality incorporated norms like Only the Guilty, Any Known Offender, and Repair and it wouldn't be nicer if morality dropped these in favor of subjective correlates. If true, might this be some evidence that morality involves objective elements? Perhaps. See Enoch (2009), Preston-Roedder (2014), and Sayre-McCord (MS) for discussion. 11 For an extended discussion on this point, see Littlejohn (2012). 12 For further defense of this description of the case, see Littlejohn (2012). Zimmerman (2008) defends a kind of subjectivist view, one that insists that what we should do is always determined by a kind of expected value. He thinks (rightly, in my view) that 10 you weren't aware of the morally significant relationship you'd create in punishing her doesn't erase your unique debt to her.13 We should resist the urge to rewrite these norms to fit the subjectivist framework. The subjectivist is right that we need to think about the subjective aspects of an agent's life when assessing the agent and her conduct. The subjective matters when we're interested in things like credit, blame, criticism, and the like, but this is only part of morality and it isn't more important than other aspects of morality. There is another part of morality that isn't directly concerned with such things and it's the part of morality that gives us guidelines that help to determine when someone has the right to proceed without interference, when someone should be protected against those who would harm them or their interests, and when some party need to take steps to apologize and make reparations. It's this part of morality that the objectivist gets right. Getting this right requires identifying standards that should guide action even when not seen perfectly by an agent. I fear that subjectivism, if left unchecked, leads us to draw mistaken conclusions about the kind of responsibility we should assume in the wake of violating some (putative) objectivist norm because it tries to deny the possibility that such guidelines have normative force when some agent has imperfect access to them. The second assumption in the argument is unificationism, the view that says that there is a principled connection between the justificatory status of beliefs about what's required of us in the situations we're in and the justificatory status responses we believe to be required. Unificationism says that if someone justifiably judges that they are required to Φ, she couldn't be required to do other than Φ.14 Segregationism denies this. The segregationist says that we can be required to do other than Φ in situations where we nevertheless justifiably judge that we're required to Φ. The disagreement is about a kind of detachment. Everyone agrees that if someone believes that they should Φ, nothing follows about whether they should. The interesting question is about the case in which they believe they should Φ and this belief is justified. If this belief is sanctioned by the relevant norms, could it be wrong for the subject to Φ in light of it? The unificationist thinks that it couldn't be wrong.15 The segregationist thinks that it can be. subjectivists have to deny Repair and deny Only the Guilty. We shall discuss this further in §3. 13 The intuition that suggests that it's possible for conscientious agents to violate moral norms owing to factual ignorance or mistake is reasonably widespread, even among those who defend subjectivist theories. See Enoch (2012), Herman (1993), and MacFarlane (2016) for interesting attempts to vindicate the intuitions that seem to support objectivist norms. 14 The relevant requirements are those that have a special kind of rational authority such that we count as irrational or unreasonable if we give these requirements no weight. Requirements of the law or etiquette 15 For defenses of unificationism, see Foley (2001), Gibbons (2010), Littlejohn (2012), and Way and Whiting (forthcoming). Greco (2014), Ichikawa and Jenkins (2013), Kiesewetter (forthcoming), Titelbaum (2015), and Smithies (2011) defend limited versions of unificationism that apply just in the epistemic realm. For defenses of segregationism, see Coates (2012), Fantl and McGrath (2009), Feldman (2008), and Lasonen-Aarnio (MS). We can use the Enkratic Requirement to defend unificationism: Enkratic Requirement: You ought to see to it that: if you judge that you yourself ought to Φ, you Φ. This is a wide-scope requirement. It rules out certain combinations. The unificationist accepts a conditional, not a wide-scope requirement. Titelbaum (2015) shows us how to derive the conditional using the requirement. Suppose that you ought not Φ. Can you nevertheless justifiably judge that you yourself ought to Φ? No. We can rewrite the requirement like this: you ought to see to it either (you do not believe that you ought to Φ or you Φ). Assume that you ought not Φ. We can now derive that you ought to see to it that you do not believe that you ought to Φ. Thus, if it's not the case that you ought not believe that you ought to Φ, it's not the case that you ought not Φ. In the case where you justifiably judge that you ought to Φ, you couldn't be in a case where you ought not Φ. We've established that you cannot justifiably believe falsehoods about what you're required to do. We can take two lessons from this. First, it shows that we should reject orthodox accounts of justification: Apparent Truth: If it seems to a subject that p and she has no available defeaters, her belief in p would be justified if based on this seeming.16 Sufficient Strength: If a subject who doesn't believe p has sufficiently strong evidence for believing p, her belief in p would be justified if based on this evidence.17 Mentalism: The justificatory status of a belief supervenes upon the subject's non-factive mental states and the subjective aspects of her mental events.18 Reliabilism: If there is a sufficiently reliable process that produces the subject's belief in p, that belief is justified.19 We can generate counterexamples to these views by focusing on propositions about what you yourself should do.20 Imagine pairs of subjects with the same evidence, with the same mental states, and the same reliable processes coming to believe that they're required to Φ. We should be able to stipulate that one subject is required not to Φ and that the other ought to Φ argue on unificationist grounds that these views deliver the wrong verdicts.21 16 This is an increasingly popular view. See, for example, Brogaard (2016), Huemer (2007), and Tucker (2010). 17 See Conee and Feldman (2004). 18 See Cohen (1984) and Conee and Feldman (2004). 19 See Goldman (1986). 20 Proponents of these views might say that their views are not subject to these counterexamples, but see the next section for a response to a line of response I think isn't very promising. 21 It is worth noting that the argument for unificationism could go through even if we dropped our objectivist assumptions about practical norms. Even without these norms, we could cause trouble for these orthodox accounts of justification for it seems that each of these accounts should allow for the possibility of false but justified beliefs about what we ought to believe, feel, or do even if we assume that facts about these oughts supervene upon a subject's subjective mental life. Once we have established Second, we can now offer an argument for Only Truth: A Transcendental Argument T1. There are objective practical norms that sometimes say that we should not Φ even when we're not aware that we should not Φ and not aware of the conditions that constitute the relevant norm's application conditions. T2. If there are such norms, there are situations where we should not judge that we are required to Φ even when we're not aware that we shouldn't Φ and are not aware of the conditions that constitute the relevant norm's application conditions. T3. The best explanation of this is that Only Truth governs belief.22 TC. So, Only Truth governs belief. This is an abductive argument, so much rests on (T3). While (T1) and (T2) only rule out the possibility of a kind of false, justified belief (i.e., one that is about what we're required to do), there doesn't seem to be anything special about such beliefs from the epistemic point of view. Only objectivist norms like Only Truth can explain why the justification of normative belief requires truth. 2.2 Just Follow (Just) the Evidence In recent discussions of the nature of evidence and its theoretical roles, these principles enjoy widespread acceptance: Revise: If you have rational conditional beliefs and you acquire a new piece of evidence p then you ought to adjust your credence in q so that Pnewq = Pold(q|p). that there cannot be false, justified normative beliefs, it will then be difficult to resist an argument for Only Truth because we couldn't appeal to these (apparently) falsified accounts of justification to resist the arguments for Only Truth and (as we'll see below) because once we have the argument for the claim that normative beliefs have to be true to be justified we can then argue that the set of beliefs that would have to be true if justified includes non-normative beliefs. 22 Strictly speaking, unificationism tells us only that a certain kind of belief cannot be both justified and false, not that it cannot be justified because it is false. There are unificationist views that are 'top-down' in that they say that the reason that the relevant belief cannot be both false and justified is that the conditions that ensure that they are justified can 'shift' the status of Φ-ing. We find versions of these views in Foley (2001) and Gibbons (2010). Way and Whiting (forthcoming) also express some sympathy for this view. The proponents of the top-down view could challenge (T3) and say that the reason that normative judgments cannot be both justified and false is best explained by the fact that the justifying conditions can help to shift the status of Φ-ing. I think Only Truth provides a better explanation for two reasons. First, closure principles cause serious difficulties for top-down unificationist views. See Littlejohn (2012). Second, these top-down views give us implausible accounts of responsibility as they either lead to a widespread skepticism about responsibility or sever the connection between culpability and Arpaly's (2002) notion of de re unresponsiveness. See Littlejohn (2014). Remain: If you have rational conditional beliefs you should not update on p unless you acquire p as a new piece of evidence. In combination, Revise and Remain tell us we ought to revise our beliefs on the evidence and nothing but the evidence. These principles can be used in an argument for objectivism about norms and justification. Suppose that you come to justifiably believe p. You see that q would be true if p. What attitude should you take towards q? If we accept closure principles for justification, it would be proper or acceptable for you to believe q. That suggests that if you justifiably believe p and infer q, you wouldn't have violated Remain. That suggests, in turn, that the object of your justified belief belongs to your evidence: Justified Evidence: If you justifiably believe p, your evidence includes p.23 Now let's think about the ontology of evidence. We're interested in propositionally specified evidence or reasons, the things that are (potentially) the subject's reasons for believing, feeling, or doing things. The subject's reasons are the things that figure in her reasoning and stand in logical relations. Ascriptions that specify the agent's propositionally specified reasons entail corresponding propositional knowledge claims.24 If we say, 'Agnes' reason for φ-ing is that p', what we say is true only if, 'Agnes knows that p' is true. There appears to be considerable linguistic evidence for half of Williamson's equation: E=K: Your evidence includes p iff you know p.25 If we combine E=K and Justified Evidence we get our first argument for (O1) and (O2): An Evidentialist Argument for Objectivism E1. If you justifiably believe p, p is part of your evidence. E2. If p is part of your evidence, you know p. EC. If you justifiably believe p, you know p. A belief is justified only if it is knowledge. Thus, there is a norm that requires us to believe only what we know. Subjectivists can either challenge Justified Evidence or argue that it cannot be used to support (OC). Dialectically, it doesn't make much sense for subjectivists to deny Justified Evidence. They wouldn't want to say that satisfying the norms that tell us when it's appropriate to update on some evidence requires more than justified belief. If they did, there would be pairs of subjects who justifiably believe p where only one can properly draw conclusions from this. The requirement for one subject to refrain from updating would have to be sufficiently obscure to the subject so that they didn't threaten the justificatory status of her beliefs but nevertheless sufficiently 'internal' to be binding. It's hard to see how this view could possibly be attractive to the subjectivist. Instead, we should expect the subjectivist to challenge the use of Justified Evidence in an argument for (OC). They'll do so by insisting that justification is a non-factive notion. 23 For defenses of Justified Evidence, see Comesana (2016), Fantl and McGrath (2009), and Littlejohn (2012). 24 See Hyman (1999), Unger (1975), and Williamson (2000). 25 For defenses of E=K, see Williamson (2000). Hyman (1999) defends a similar view about potential motivating reasons. For criticism that doesn't focus on the truthrequirement on reasons or evidence, see Hughes (2014) and Locke (2015). Justified Evidence can be understood in different ways. We can read into it some orthodox account of justification and then try to show that any belief that has the relevant justification-making properties can properly figure in reasoning and give us premises to draw conclusions from. 26 Alternatively, we could resist the idea a proposition gets turned into evidence by virtue of being the content of a belief that has the justification-making properties that figure in non-factive accounts of justified belief. We could say that justification turns on whether a belief's contribution to rational processes is a good one. We ask whether the belief's object is the kind of thing that could be a good tool or instrument. We thus use some independent standard that distinguishes genuine reasons from spurious reasons to ground a standard that we use to distinguish the good beliefs from the bad, the beliefs that can properly function in reasoning from the ones that cannot. I prefer the second reading of Justified Evidence. Since I think we have good independent reason to think that a propositionally specified reason couldn't be a genuine reason unless it was a fact, I'd argue from the fact that genuine reasons are facts 'back' to the conclusion that justified beliefs are the ones that provide us with facts to figure in our reasoning. How should we decide between these options? Should we combine Justified Evidence with some orthodox non-factive account of justification and reject E=K or should we combine Justified Evidence with E=K and accept (EC)? One reason to prefer the second option to the first is that the first faces serious difficulties that don't arise for factive accounts of justification. Let's say that the 'target view' is a view that incorporates Justified Evidence and allows for false evidence because it sees justification as a property that a belief has if it's produced by a reliable process or because it's the belief that results from the excellent use of a subject's rational capacities even when the resulting belief is mistaken. Difficulties arise for the target view when we think about the properties of processes like conditionalization and the properties of 'ought'. Imagine that Agnes completes Here I Am, a carefully researched if somewhat self-indulgent autobiography. Because it is so carefully researched, Agnes has a wellfounded belief that corresponds to each claim in the body of the book. Thus, it would seem that she should have a justified belief that corresponds to each claim in the body of the book. After completing the body, her fact-checker tells her that she found precisely one error in the manuscript. Alas, her fact-checker dies suddenly and unexpectedly before she can reveal the error's location. After searching in vain for the error, Agnes fails to find the error. She notes in the book's preface that the book contains an error and it is sent to press. Agnes satisfies the conditions that orthodox accounts of justification impose so we should say that she has a justified belief that corresponds to each claim in the body of her book (i.e., p1-p1,000,000) and in the preface. Thus, the target view supports (P1)-(P3) in this argument: P1. Agnes justifiably believes p1, Agnes justifiably believes p2, ..., and Agnes rationally believes p1,000,000. P2. Agnes justifiably believes ~(p1 & p2 & ... & p1,000,000). 26 See Fantl and McGrath (2009) and Comesana (2016) for examples of this. They think that arguments from error also give us good reason to allow that a subject's reason or evidence could be a false proposition. For responses to these arguments, see Alvarez (2010), Dutant (forthcoming), Hornsby (2007), Littlejohn (2012), and Lord (forthcoming). P3. Whatever Agnes justifiably believes belongs to her evidence. P4. Agnes ought to update on p1, Agnes ought to update on p2, ..., and Agnes ought to update on p1,000,000. [(P1), (P3), and Revise] P5. Agnes ought to update on ~(p1 & p2 & ... & p1,000,000). [(P2), (P3), and Revise] P6. Agnes ought to (update on p1 & update on p2 & ... & update on p1,000,000). [(P4), Agglomeration] P7. If Agnes ought to update on ~(p1 & p2 & ... & p1,000,000), Agnes ought not update on (p1 & p2 & ... & p1,000,000). P8. Agnes ought not update on (p1 & p2 & ... & p1,000,000). [(P5), (P7)] P9. If Agnes ought to: (update on p1, update on p2, ..., update on p1,000,000), it is not the case that Agnes ought not update on (p1 & p2 & ... & p1,000,000). P10. It is not the case that Agnes ought not update on (p1 & p2 & ... & p1,000,000). [(P6), (P9)] Because (P10) contradicts (P8), we have to reject some premise in the reasoning that supports (P8) and (P10). If we accept the objectivist view of justification on which justified beliefs have to be pieces of knowledge, we block the reasoning at the outset. We know that (P1) and (P2) couldn't both be true. I don't see how a proponent of the target view could plausibly reject both claims.27 Justified Evidence is part of the target view, so its defenders cannot reject (P3). Revise generates the 'ought' claims that support (P6) because of agglomeration. As for (P7), this is the plausible claim that if you ought to update on p, you should not also update on its negation. As for (P9), this is plausible given that a process like conditionalization is cumulative. The result of conditionalizing on the conjuncts is the same as the result of conditionalizing on the conjunctions. It is hard to see how there could be a requirement to update on p, say, a requirement to update on q, and also a requirement to refrain from updating on the conjunction when we know in advance that the result of the one-step update and twostep update has to be the same. The trouble with the target view is that it identifies a non-factive epistemic relation as sufficient for the possession of evidence. This means the target view allows for inconsistent evidence. It cannot be that such things are things we ought to update on. Theoretically, the best way to save agglomeration and accommodate norms like Revise is adopt a factive account of evidence. When we combine that with Justified Evidence, we get our argument for (EC). The reason that justified beliefs have to be true is not that they need some maximal independent support, but because only beliefs 27 Some writers would argue that we shouldn't accept (P1) and (P2). Some would argue this because they think that there's some kind of defeater operative that prevents them from both being true (e.g., Ryan (1991)). For arguments that we shouldn't deny that it's possible for (P1) and (P2) to be true, see Easwaran and Fitelson (2015) and Worsnip (2015). Some would try to show that we don't have the right evidence or grounds for (P2) to be true (e.g., Smith (2016)). Using testimony, it's hard to see how denying (P2) could be a viable strategy. that contribute truths to rational processes meet the standard that distinguishes the beliefs that can do what they're supposed to do from those that cannot. 3. Subjectivism about Justification In the previous sections, I've offered a transcendental argument for Only Truth and an evidentialist argument for Only Knowledge. I've also argued that these norms support an account of justification on which you can only justifiably believe what you know. Most epistemologists reject (OC), but it isn't always clear where they think this kind of argument goes wrong. Huemer (forthcoming) suggests that the problem with this style of argument for (OC) is not that they assume that there are objectivist norms, but with the further assumption that these norms tell us something about justification. He agrees that there's a sense in which we 'ought' to conform to epistemic norms like Only Truth. He agrees we cannot justifiably believe what we 'ought' not believe. Still, he thinks, there's no good reason to think that justification is an objectivist notion. To square this circle, Huemer claims that much of our normative language is ambiguous. Terms like 'ought' and 'should' admit of objective and subjective readings (or 'external' and 'internal' readings, as he puts it), but some terms like 'justification' admit of only one subjective reading. The difficulty with this argument is that (O1) is true on the subjective reading only and (O2) is true on the objective reading only: An Argument for Objectivism O1. Your beliefs are justified iff you violate no epistemic norms in holding these beliefs. O2. You violate no epistemic norms in holding these beliefs iff these beliefs constitute knowledge. OC. Your beliefs are justified iff they constitute knowledge. As he sees things, justification is a matter of conforming only to subjective norms. When it comes to debates between the objectivist and subjectivist about norms, he thinks we should see that these debates aren't substantive: Now suppose someone asks, “Does what a person should do depend upon external facts of which that person may be unaware, or does it only depend upon information available from the subject’s own point of view?” This would be a misguided question. There is no deep issue there; there is only the perfectly shallow, semantic question of whether you want to use the internal sense or the external sense of the word “should”. Both senses, as far as I can tell, are established in ordinary English; neither is more correct than the other (Huemer: forthcoming: 7). Because these debates rest on the mistaken presupposition that there is a substantive debate to be had, they should end. The objectivist and subjectivist about norms should see that their opponent's views get something important right. Live and let live. When it comes to debates about justification, however, he thinks these debates shouldn't continue because (O1) is true only on its subjective reading: The truth of this thesis [i.e., that justification is entirely a matter of conforming to subjective norms], by the way, strikes me as just an obvious semantic point. Unlike the word “should”, I think the word “justified” is not ambiguous; rather, it has only the internal meaning in standard English (Huemer forthcoming: 8). Even if the objectivist is right about epistemic norms, there is nothing that the objectivist about justification gets right. When it comes to justification, it only matters whether we conform to subjectivist norms. Only the subjectivist about justification has a sensible view and that's why this debate should end. Live and let die. In discussions of the subjective 'ought' and the (alleged) ambiguity of normative language, philosophers tend to be dividers or debaters.28 Dividers think that pointing to the ambiguity will show that there's no need to debate an issue because both parties to the debate are right about something. The parties to the debate are only mistaken in that they fail to see their opponent's insights. Maybe Huemer has his finger on something important here in helping us see why so many epistemologists would be open to Only Truth but convinced that justification couldn't require truth. Is Huemer right that the debates between objectivists and subjectivists about norms aren't substantive? I think not. Objective and subjective norms can come into conflict. In some cases of conflict, it looks as if there's a difficult and substantive question about how we should deal with these conflicts. In turn, it's not at all obvious that the justified response would be to conform to subjective norms and violate objective ones. The suggestion that the debates between the objectivist and subjectivist aren't substantive might be plausible if the subjectivist could vindicate the intuitions that drive objectivism, but we'll see that they cannot do this. The upshot is that subjectivists should be debaters, not dividers. Consider one disagreement that dividers take to be non-substantive, a debate between objectivists about punishment and subjectivists who accept norms such as this one: You should sentence the accused iff their guilt is beyond all reasonable doubt (Beyond Reasonable Doubt). According to Only the Guilty and All Known Offenders, Agnes would have done nothing wrong if, say, she sentenced 1,000,000 persons standing trial provided that she knew of the accused to be guilty. Assuming, as we will, that Agnes believed the accused to be guilty beyond any reasonable doubt if she knew them to be guilty, the subjectivists who accept Beyond Reasonable Doubt would agree that Agnes did no wrong by sentencing these 1,000,000 persons for the crimes they committed. The interesting issues don't arise when everyone agrees that Agnes got it all right, but when it seems to someone that Agnes got something wrong. Let's imagine again that Agnes serves on 1,000,000 juries and that the evidence in each case leaves no room for reasonable doubt. Thus, we get: 1. Agnes subjectively ought to convict in case 1. 2. Agnes subjectively ought to convict in case 2. 3. Agnes subjectively ought to convict in case 3. ... 1,000,000. Agnes subjectively ought to convict in case 1,000,000. Suppose that after her 1,000,000th case is brought to a close Agnes learns that one of the people convicted was innocent. She thus learns: 28 See Sepielli (forthcoming). 1,000,001. Agnes objectively ought not: (convict in case 1 & convict in case 2 & ... & convict in case 1,000,000). If Agnes knows that she objectively ought not do something, it seems plausible that she subjectively ought not do that thing: 1,000,002. Agnes subjectively ought not: (convict in case 1 & convict in case 2 & ... & convict in case 1,000,000). If the subjective 'ought' behaves like the 'ought' of standard deontic logic, agglomeration and (1)-(1,000,000) give us: 1,000,003. Agnes subjectively ought to: (convict in case 1 & convict in case 2 & ... & convict in case 1,000,000). And if the subjective 'ought' acts anything like the 'ought' of standard deontic logic (1,000,002) gives us: 1,000,004. It is not the case that Agnes subjectively ought to: (convict in case 1 & convict in case 2 & ... & convict in case 1,000,000). Since (1,000,003) contradicts (1,000,004), something has to give. It seems we have four options: Option 1: Deny at least one of (1)-(1,000,000) and deny Beyond Reasonable Doubt. Option 2: Deny (1,000,001) and deny Only the Guilty. Option 3: Deny that the subjective 'ought' behaves like the 'ought' of standard deontic logic. Option 4: Deny the bridge principle that states that you subjectively should not do what you know you objectively should not do. Dividers want to show that the disagreement between the subjectivists and objectivists isn't substantive, so they cannot avoid the contradiction by means of Option 1 or 2. If they deny Only the Guilty or Reasonable Doubt, they would enter into just the debate that they wish we'd all stay out of. The third option is problematic for dividers. Suppose we've established (1)(1,000,000) and (1,000,002) and look to avoid the contradiction by denying agglomeration. It seems that, barring cases of duress, a subject who fails to do what she subjectively ought to could be held responsible for that failure or subject to criticism. We shouldn't blame or criticize someone for carrying out all of the acts described in (1)(1,000,000) if we'd blame or criticize them for failing to carry them all out. (Does it make sense to say that someone would avoid criticism iff they sentenced in each case but would not avoid criticism if they sentenced everyone?) The fourth option is the one that remains. It looks as if dividers have good grounds for challenging the bridge principle.29 Consider the mineshaft case.30 Ten miners are working together underground, but Agnes doesn't know whether they are in Shaft A or Shaft B. She knows that it's raining and that if she does nothing all will die at the bottom of the mine. She has three options: block shaft A, block shaft B, 29 See Wedgwood (2013) for discussion. Wedgwood offers a revision of the Enkratic Requirement that is supposed to avoid the difficulties that arise for this bridge principle. The principle holds only when Φ is a fine-grained option. In our case, it is not. 30 Discussed in Parfit (2011). The case is a variant on Regan (1980) and Jackson (1991). partially block both shafts. If she completely blocks the opening to A, there will be enough air inside for 10 to live but B will flood and kill anyone in that shaft. If she blocks B completely, there will be enough air inside for 10 to live, but A will flood and kill anyone in that shaft. If she partially blocks both openings, she will be guaranteed to save 9 miners wherever they happen to be but 1 will be killed. In this sort of case, Agnes knows that she objectively ought to either block Shaft A or B, but doesn't know which shaft it is. Thus, she knows that she objectively ought not partially close the shafts. Still, it seems to many commentators that Agnes subjectively ought to partially close the shafts. This suggests that the bridge principle is mistaken. Wedgwood notes (rightly) that the Enkratic Requirement but only when it's read as pertaining to fine-grained options (e.g., blocking shaft A completely). Thus, it's plausible that the bridge principle is only superficially similar to the Enkratic Requirement and shouldn't be accepted. If dividers reject the bridge principle, would this solve the problem? No, not quite. If they deny the bridge principle, they can reject (1,000,002). They could say that while there is a sense in which Agnes succeeded in doing everything she should do, there is another sense in which she failed to do what she should do. Suppose we say that. If Agnes then asks if she's required to make reparations, what would the divider say? The divider has to ask the subjectivists what to say since they want to avoid disagreements with them. Subjectivists might adopt one of two views. They might offer a version of Repair, suitably modified to fit with their theoretical orientation. They might say that upon discovering that she sentenced an innocent person to prison, Agnes has a reparative duty to right some past wrong. On this view, it turns out to be true, as the objectivist says, that Agnes failed to do what she ought to do, but crucially the truth of this depends upon what was revealed after conviction. Alternatively, the subjectivist could insist that since Agnes did no wrong, violated no norm, she never failed to do what she ought to do and reparation isn't called for. The subjectivist would say that the objective facts that figure in objectivist accounts of norm have to be excluded from the framework so that they don't figure in our account of rights, obligations, or duties to repair or make reparation.31 The accused would have the right to a fair trial, but not to escape sentencing if innocent or to receive reparations if wrongly convicted. The problem with the first response is that it leads us right back to the problem we were trying to avoid by denying the bridge principle. If the subjectivist says that Agnes has a reparative duty because she failed to do something she ought to do by virtue of sentencing the accused in each case, the subjectivist is offering us an account on which (1,000,004) comes out as true. As we're assuming that the subjectivist wants to offer an account of 'ought' on which 'ought' satisfies agglomeration, this proposal won't do. The truth of (1,000,004) would force us to give up at least one of (1)(1,000,000) but the subjectivist has no good grounds for giving up any of these claims. The new information doesn't change the fact that Agnes had adequate evidence at the 31 The most completely worked out version of this view is found in Zimmerman (2008). He is a subjectivist in the sense that he thinks that what a subject ought to do depends upon her evidence and not upon any further facts that don't supervene upon that evidence. He is not a divider. For further defenses of the subjectivist debater, see Kiesewetter (forthcoming) and Lord (forthcoming b). time of sentencing and doesn't change the fact that she still has adequate evidence for believing of each conviction that it was appropriate. The second subjectivist response causes trouble for dividers. If the subjectivist offers this response, she's now engaged in a genuine, substantive disagreement with the objectivist. The divider cannot say that there's a kernel of truth in both views because our subjectivist is rejecting the idea that there's some sense in which we ought to make reparations. Even if Agnes learned the identity of the individual who was framed, our subjectivist would insist that the objectivists would be wrong to say that Agnes did something that violated this person's rights, that she failed in her obligations to the accused, and wrong to say that Agnes has a reparative duty to right some past wrong. At just this point there is a substantive question that we need to settle. Upon discovering the identity of the innocent person she convicted, would Agnes' decision not to make reparations be justified? If you say that it would, you are taking up a substantive view and siding with a subjectivist debater. If you say that it would not, you are taking up a substantive view and siding with an objectivist debater who takes justification to depend upon conforming to objectivist norms. It seems clear to me that Agnes would have no justification for failing to try to make reparations, but this interesting fact about justification would appear to be connected to the past violation of an objectivist norm (Only the Guilty), not a subjectivist norm. Thus, the debate turns out to be substantive and the objectivist view about justification turns out to have an important virtue that the subjectivist view lacks. The objectivist (and the objectivist alone?) has a coherent position that vindicates important intuitions about reparation. 4. Conclusion I have defended two objectivist views, one about norms and another about justification. There is no simple case to be made for either objectivist view, but it seems that the objectivist framework best fits with some plausible claims about the kinds of obligations we're under and with some popular claims about the normative significance of our evidence. What emerges from this is a simple idea that will undoubtedly require further defense. For belief to play its role well, it has to provide us with reasons and put us in touch with the part of reality that consists of facts. A belief only does this when it constitutes knowledge. This is why Only Knowledge governs belief. There is little theoretical gain to be had by insisting that this point, if correct, tells us nothing about justification. Justification loses all of its theoretical significance if we characterize it as a status that a belief can have or fail to have quite apart from whether that belief conforms to all the norms that govern belief. This is why objectivism about norms and justification stands or falls together. References Alvarez, Maria. 2010. Kinds of Reasons. Oxford University Press. Arpaly, Nomy. 2002. Unprincipled Virtue. Oxford University Press. Benton, Matthew. 2011. Two More for the Knowledge Account of Assertion. Analysis 71: 684-687. Bird, Alexander. 2007. Justified Judging. Philosophy and Phenomenological Research 74: 81-110. Boghossian, Paul. 2008. Content and Justification. Oxford University Press. Brogaard, Berit. 2016. Staying Indoors: How Phenomenal Dogmatism Solves the Skeptical Problem without Going Externalist. In B. Coppenger and M. Bergmann (ed.), Intellectual Assurance. Oxford University Press. Brown, Jessica. 2008. Subject-Sensitive Invariantism and the Knowledge Norm for Practical Reasoning. Nous 42: 167-189. Coates, Allen. 2012. Rational Epistemic Akrasia. American Philosophical Quarterly 49: 11324. Cohen, Stewart. 1984. Justification and Truth. Philosophical Studies 46: 279-95. Comesana, Juan. 2016. Falsehood and Entailment. Philosophical Perspectives 29: 82-94. Conee, Earl and Richard Feldman. 2004. Evidentialism. Oxford University Press. Dougherty, Trent. 2011. Introduction. In T. Dougherty (ed.), Evidentialism and its Discontents. Oxford University Press. Dutant, Julien. Forthcoming. Knowledge-Based Decision Theory and the New Evil Demon Problem. In F. Dorsch and J. Dutant (ed.), The New Evil Demon. Oxford University Press. Easwaran, Kenny and Branden Fitelson. 2015. Accuracy, Coherence, and Evidence. In T. Gendler and J. Hawthorne (ed.) Oxford Studies in Epistemology. Oxford University Press. Enoch, David. 2009. Wouldn't It Be Nice if p, Therefore, p (for a moral p). Utilitas 21: 222-4. Fantl, Jeremy and Matt McGrath. 2009. Knowledge in an Uncertain World. Oxford University Press. Feldman, Richard. 1988. Subjective and Objective Justification in Ethics and Epistemology. the Monist 71 405-19. Foley, Richard. 2001. The Foundational Role of Epistemology in a General Theory of Rationality. In A. Fairweather and L. Zagzebski (ed.), Virtue Epistemology: Essays on Epistemic Virtue and Responsibility. Oxford: Oxford University Press. Gibbons, John. 2010. Things That Make Things Reasonable. Philosophy and Phenomenological Research 81: 335–61 Goldman, Alvin. 1986. Epistemology and Cognition. Harvard University Press. Greco, Daniel. 2014. A Puzzle about Epistemic Akrasia. Philosophical Studies 167: 201-19. Hawthorne, John and Ofra Magidor. Forthcoming. Reflections on Reasons. . In D. Star (ed.), Oxford Handbook of Reasons and Normativity. Oxford University Press. Hornsby, Jennifer. 2007. Knowledge in Action. In A. Leist (ed.), Action in Context. De Gruyter, pp. 285–302. Huemer, Michael. 2007. Compassionate Phenomenal Conservatism. Philosophy and Phenomenological Research 74: 30-55. Huemer, Michael. Forthcoming. Defending a Brain in a Vat. In F. Dorsch and J. Dutant (ed.), The New Evil Demon. Oxford University Press. Hughes, Nick. 2014. Is Knowledge the Ability to ϕ for the Reason that P? Episteme 11: 457-62. Hyman, John. 1999. How Knowledge Works. Philosophical Quarterly 49: 433-51. Ichikawa, Jonathan and Ben Jarvis. 2013. The Rules of Thought. Oxford University Press. Jackson, Frank. 1991. Decision-Theoretic Consequentialism and the Nearest and Dearest Objection. Ethics 101: 461-82. Kiesewetter, Benjamin. Forthcoming. You ought to φ only if you may believe that you ought to φ. Philosophical Quarterly. Lasonen-Aarnio, Maria. MS. Enkrasia or Evidentialism: Learning to Love Mismatch. Littlejohn, Clayton. 2011. Defeating Phenomenal Conservatism. Analytic Philosophy 52: 35-48. Littlejohn, Clayton. 2013. The Russellian Retreat. Proceedings of the Aristotelian Society 113: 293-320. Littlejohn, Clayton. 2014. The Unity of Reason. In C. Littlejohn and J. Turri (ed.), Epistemic Norms. Oxford University Press. Locke, D. 2015. Knowledge, Explanation, and Motivation. American Philosophical Quarterly 52: 215-33. Lord, Errol. Forthcoming. Defeating the Externalist's Demons. In F. Dorsch and J. Dutant (ed.), The New Evil Demon. Oxford University Press. Lord, Errol. Forthcoming b. Acting for the Right Reasons, Abilities, and Obligation. In R. Shafer-Landau (ed.), Oxford Studies in Metaethics. Oxford University Press. McKinnon, Rachel. 2013. The Supportive Reasons Norm of Assertion. American Philosophical Quarterly 50: 121-35. Mantel, Susanne. 2013. Acting for Reasons, Apt Action, and Knowledge. Synthese 190: 3865-3888. McGlynn, Aidan. 2014. Knowledge First? Palgrave Macmillan. Mitova, Veli. Forthcoming. Truthy Psychologism about Evidence. Philosophical Studies. Nolfi, Kate. 2015. How to be a Normativist about the Nature of Belief. Pacific Philosophical Quarterly 96: 181-204. Parfit, Derek. 2011. On What Matters. Volume I. Oxford University Press. Preston-Roedder, Ryan. 2014. A Better World. Philosophical Studies 168: 629-44. Regan, Don. 1980. Utilitarianism and Cooperation. Oxford University Press. Ryan, Sharon. 1991. The Preface Paradox. Philosophical Studies 64: 293-307. Sayre-McCord, Geoff. MS. On a Theory of a Better Morality. Sepielli, Andrew. Forthcoming. Subjective and Objective Reasons. In D. Star (ed.), Oxford Handbook of Reasons and Normativity. Oxford University Press. Smith, Martin. 2016. Between Probability and Certainty. Oxford University Press. Smithies, Declan. 2011. The Normative Role of Knowledge. Nous 46: 265-88. Steglich-Petersen, Asbjorn. 2013. Truth as the Aim of Justification. In T. Chan (ed.), The Aim of Belief. Oxford University Press. Sutton, Jonathan. 2007. Without Justification. MIT University Press. Tucker, Chris. 2010. Why Open-Minded People Should Endorse Dogmatism. Philosophical Perspectives 24: 529-45. Turri, John. 2011. The Express Knowledge Account of Assertion. Australasian Journal of Philosophy 89: 37-45. Unger, Peter. 1975. Ignorance. Oxford University Press. Way, Jonathan and Daniel Whiting. Forthcoming. If You Justifiably Believe that You Ought to Φ, You Ought to Φ. Philosophical Studies. Wedgwood, Ralph. 2002. Internalism Explained. Philosophy and Phenomenological Research 65: 349-69. Wedgwood, Ralph. 2013. The Right Thing to Believe. In T. Chan (ed.), The Aim of Belief. Oxford University Press. Wedgwood, Ralph. 2013b. Akrasia and Uncertainty. Organon F 20: 484-506. Whiting, Daniel. 2013. Nothing but the Truth: On the Norms and Aims of Belief. In T. Chan (ed.), The Aim of Belief. Oxford University Press. Williamson, Timothy. 2000. Knowledge and its Limits. Oxford University Press. Williamson, Timothy. Forthcoming. Justifications, Excuses, and Skeptical Scenarios. In F. Dorsch and J. Dutant (ed.), The New Evil Demon. Oxford University Press. Worsnip, Alex. 2015. Belief, Credence, and the Preface Paradox. Australasian Journal of Philosophy 94: 549-562. Zimmerman, Michael. 2008. Living with Uncertainty: The Moral Significance of Ignorance. Cambridge University Press.