Knowledge and reasonableness

Abstract

The notion of relevance plays a role in many accounts of knowledge and knowledge ascription. Although use of the notion is well-motivated, theorists struggle to codify relevance. A reasonable person standard of relevance addresses this codification problem, and provides an objective and flexible standard of relevance; however, treating relevance as reasonableness seems to allow practical factors to determine whether one has knowledge or not—so-called “pragmatic encroachment.” I argue that a fuller understanding of reasonableness and of the role of practical factors in the acquisition of knowledge lets us avoid pragmatic encroachment.

1 The problem of relevance

Much of what we take ourselves to know is based on less than conclusive evidence. We know the global climate is growing warmer, based on evidence of rising sea levels and diminishing snowpack; we also know why this is happening, based on evidence such as ice samples showing a historical correlation between temperature and greenhouse gases. Our evidence isn’t infallible—in a different possible world, someone in our total evidential state could be wrong about the cause of global warming; an undetected and surprising source of heat energy is (let’s suppose) possible, however implausible or outlandish such a possibility is. But mere metaphysical possibilities can be ignored—they don’t prevent our knowing that increased greenhouse gases are the cause of global warming. Like a lot of our empirical knowledge this knowledge is based on fallible, but entirely compelling evidence.

How much evidence is enough? On an absolute standard requiring evidence against every counter-possibility no matter how outlandish, we know next to nothing. Alternatively, a less than absolute standard can be set at an arbitrary threshold, as is done in science: for instance in the search for the Higgs boson, researchers were very certain they had found tell-tale marks, but kept going with data analysis until the probability was vanishingly small (“5-sigmas” or a 1 in 3, 500,000 chance) that no boson was responsible for their observations. Neither an absolutely demanding nor an arbitrary standard seems the right one for ordinary claims to know. If you tell me you know the box score from last night’s game, I’m not expecting you to have reached an arbitrary probabilistic standard, and I’m not expecting you to have evidence against all counter-possibilities no matter how fantastic.

One way to articulate a fallibilist epistemology is a Relevant Alternatives approach.¹ On this approach knowing an empirical proposition P requires one’s evidence be sufficient to eliminate only relevant alternatives to P (where alternatives to P are not-P possibilities). Fantastic possibilities are irrelevant, and there is no need to gather evidence against them.

Well and good, but what makes an alternative relevant? Relevant Alternatives theorists struggle with this question.² Fred Dretske, one of the first defenders of the Relevant Alternatives approach, suggests a counterfactual criterion: the alternatives irrelevant for knowing P are those that don’t occur in the nearest possible world(s) where P is false.³ Unfortunately, the suggested counterfactual test makes some unexpected alternatives relevant.⁴ As far as I know, every such suggestion has its counter-examples. The history of attempts and failures to state a criterion of relevance makes it is easy to feel that while relevance theory is attractive, stating a criterion for relevance is a hopeless task.⁵

So much the worse for Relevant Alternatives theory, one might say. But it isn’t only theories that overtly bill themselves “Relevant Alternatives Theories” that face the problem of relevance. Modal theories that would seem to escape the problem by characterizing knowledge in modal metaphysical terms (e.g. reliabilism, safety theory, sensitivity theory) are implicitly offering accounts of which worlds are relevant. These theories also run into problems with counter-examples, where metaphysical facts make the subject’s belief that P unsafe or insensitive or what have you, and yet we would judge the subject to know P. The problem of relevance trips them up, too. Some Invariantist theories would seem to escape the problem by avoiding the idea of relevance in characterizing knowledge or the truth conditions of knowledge claims. Yet any Invariantist theory that invokes conversational pragmatics to account for ordinary knowledge claims will find it natural to appeal to relevance.⁶ Many existing accounts of knowledge face the problem of relevance one way or another.

In this paper, I first suggest why the problem of relevance is so difficult. Once we have a diagnosis, we’ll better see our way to a solution. I will argue that the solution is to think of relevance in terms of reasonableness: a relevant alternative to P is an alternative a reasonable person would want eliminated before taking P to be true.⁷ If we give reasonableness a central place in an account of knowledge we invite an apparently practical notion into the heart of epistemology, and this appearance raises concerns. In response I’ll argue that once we understand reasonableness, we’ll see that practical factors do not control or determine one’s knowledge.

2 Diagnosing the problem of relevance

What is relevant is not codifiable in a detailed rule stateable in advance of encountering the cases covered by it. Attempts to state a standard of relevance have attempted this codification and failed. A diagnosis of the problem should tell us why this difficulty of codification arises.

In a different context, famous legal theorist H.L. Hart observes that the law encounters difficulty in formulating rules to govern our conduct. Hart gives us an example: we aim to keep things peaceful in the park, and so we legislate no vehicles in the park. Now a new circumstance arises we hadn’t anticipated—does a skateboard count as a vehicle?

“… until we have put the general aim of peace in the park into connection with those cases which we did not, or perhaps could not, initially envisage… our aim is, in this direction indeterminate. We have not settled, because we have not anticipated, the question which will be raised by the unenvisaged case when it occurs: whether some degree of peace in the park is to be sacrificed to, or defended against, those children whose pleasure or interest it is to use these things.”⁸

Our general aim (peace in the park) does not settle how we aim to behave in specific unenvisaged situations (will we forbid skateboards?). Our specific aim can only be settled when we have a fresh chance to consider how our interests and aims fare together in the particular situation at hand. Because of this fact,

“…we should not cherish, even as an ideal, the conception of a rule so detailed that the question whether it applied or not to a particular case was always settled in advance, and never involved, at the point of actual application, a fresh choice between open alternatives.”⁹

Hart here makes a deep observation about human aims. Our desires or interests alone cannot completely determine our specific aims in the specific forms they take in specific situations of our lives. Over and above the indeterminacy or “open texture” of language, the fact of our necessarily incomplete knowledge of our aims means that we cannot articulate a rule that ensures our specific aims will be satisfied in particular situations.

Hart’s observation holds true about our epistemic lives as well as our practical lives. In our epistemic lives we also have general aims, and incomplete knowledge of what the satisfaction of these general aims requires in particular situations. As inquirers we have general aims to acquire true beliefs, to avoid false beliefs, to have evidence enough to settle questions for ourselves and for others, to gain knowledge and understanding, and to communicate the knowledge and understanding we possess. But how these epistemic aims balance with each other, let alone how they balance with our other aims, is a complicated matter; we don’t want a rule about how to balance these aims that settles the matter in advance.

Focus for a moment on the general aims of acquiring true beliefs and avoiding false beliefs. We cannot, in advance of finding ourselves in a particular situation, settle the question of how much inquiry to pursue or how much evidence to gather so as to ward off the possibility of a false belief as we try to acquire a true belief. When Justin is invited to take part in the yearly bird count, he does his best over the course of 2 days to correctly identify the birds in his area. Quick counts are part of this, and occasional mis-identifications are bound to happen. An occasional error is tolerated because the aim is to have a fairly capacious count. Imagine for instance the probability of mistaking a rare bird for a native is very small, and what it would take to rule out this alternative is very laborious—it would involve every attempted identification of a native. Justin forgoes the laborious process in order to meet the aim of a generally reliable and capacious count. If the generally reliable count is not what Justin is after, and the undesirability of error is greater, then Justin may look for more evidence, capturing and examining individual birds. In weighing competing epistemic aims, then, situations can differ in the way one’s aims are to be balanced.

The problem of relevance is the problem of saying what it takes for an alternative to be relevant. What is relevant is not codifiable in a detailed rule stateable in advance of encountering the cases covered by it. Attempts to state a standard of relevance have attempted this codification and failed. Hart tells us why codification into a detailed rule is not possible. It is not possible because of our lack of foreknowledge about how our general aims are to be met in specific situations. This points to a diagnosis of the problem of relevance as it bears on the theory of knowledge: at the root of the problem is our lack of foreknowledge of how our general epistemic aims are to be met in specific situations.

3 Hart’s strategy

If this diagnosis is correct, how can we tackle the problem? Hart observes that a rule governing conduct cannot be “so detailed that the question whether it applied or not to a particular case was always settled in advance, and never involved, at the point of actual application, a fresh choice between open alternatives.” A skeptic might jump to the conclusion that therefore no rules or principles can govern our conduct. But this is too fast. Hart counters that the law may yet be encoded in rules, because the law has strategies for avoiding rules that are so “detailed.”

Hart articulates two strategies: the first strategy applies in cases where we can anticipate enough about our interests and future situations to set up an administrative body to handle questions about how to interpret and apply general rules. Hart’s second strategy is, for our purposes here, more interesting: it applies where “the range of circumstances, though very varied, covers familiar features of common experience.” In such cases, Hart notes, “common judgments of what is reasonable can be used…” to interpret and apply general rules to particular cases (Hart, 132).

An absolute and inflexible rule requiring one always to stop, look, and listen for traffic at a light, might needlessly slow one’s journey to the hospital with someone who is bleeding to death. (Hart, 133) We have a general aim to secure people against harm, and that aim is not served by an absolute rule. Contra the skeptic this fact doesn’t force us to abandon efforts to govern our conduct with rules of the road. What’s needed is the flexibility to make further choices about our specific aims in specific circumstances; and we can gain the needed flexibility with a rule that calls for the use of judgment about what is reasonable in the circumstances. Instead of saying that due care in driving absolutely requires us to “stop, look and listen for traffic”, we say that one should “exercise reasonable care.”

Hart’s suggestion is that we can state flexible rules through an appeal to what is reasonable, so that users of the rule are instructed to use their judgment, making “a fresh choice” and deciding how the rule applies to a particular case. In so doing, a judgment of what is reasonable determines how one’s general aims are to be realized in the situation at hand.

Hart’s suggestion raises pressing questions: how are judgments of what is reasonable produced? What makes a person reasonable?

4 The reasonable person

Much ink has been spilled in jurisprudence arguing that the reasonable person standard can be summed up in the rules of a chosen moral theory.¹⁰ Alternatively, John Gardner argues that the reasonable person standard is an “open standard”: its content is only that the reasonable person is justified, or acts on undefeated reasons.¹¹ I will follow the path of so-called ordinary language philosophy as J.L. Austin advocates it. We start with reflection on ordinary usage. We also expect that further philosophical reflection may lead us to revise our understanding of the idea of reasonableness, in order to make best sense of our complex practices with the term.

As we ordinarily use the term “reasonable person”, the following truisms hold:

• (a) A reasonable person sizes up the reasons for and against various courses of action, and in doing so can judge questions of widely disparate kinds. “Am I being reasonable/unreasonable in doing this?” takes an endless variety of objects, each of which require balancing an untold variety of values. (“Am I being reasonable in thinking there are aliens?” requires sensitivity to epistemic values (evidence, likelihoods, testimony); “Am I being reasonable in asking my friend to babysit again?” requires sensitivity to social and moral values, and so on. A key feature of the reasonable person is her general-purpose sensitivity to, and ability to make judgments about, competing values of widely different kinds.

• (b) A reasonable person thinks well, has some degree of intelligence, is sensible; but mere intelligence and logical thinking is not enough to be reasonable. Here we can think of cases real and imagined: Think of a highly crafty individual who sets himself an awful end—Shakespeare’s Richard III who chooses “to prove a villain.” One has to use one’s reasoning to act for the good, not for evil, if one is reasonable. Decency and intelligence are not sufficient ingredients for being reasonable, but they are necessary.

• (c) A person is reasonable when she is sensitive to what’s important in a situation, and shows tact and proportionality in her responses in light of what’s important.

• (d) What is reasonable depends on what one can be expected to already know. If you are answering a question in an exam about probability, it might be reasonable to assign equal probabilities to the outcome that a thumbtack lands on its point, heads up, and to the outcome that it lands on its tail, point up. No one expects you to already know facts of mundane physics in order to demonstrate your knowledge of probability theory. If you are faced with riding a bike behind a truck that drops thumbtacks in your path, it is not reasonable to expect that only half the tacks will present a problem. Everyone expects you to know facts of mundane physics, if you’re riding a bike.

• (e) The reasonable person will see things from another person’s point of view, and shows a sensible concern for others. Being reasonable means being able to judge matters from another’s perspective, in some cases, even forming one’s own reasons with a view to the reasons of other (reasonable) people.

These truisms give us a start on understanding what reasonableness is. A person is reasonable when she is sensitive to what’s important in a situation, shows tact and proportionality in her responses and a sensible concern for others, is disposed to keep in view her own ends and aims as well as those of others, takes into consideration the facts of a situation, sizes up the reasons for and against various courses of action, including the reasons others see, and acts for the good in accord with those reasons.

5 Relevance as reasonableness

Applying Hart’s strategy to the problem of relevance we say that relevant alternatives are reasonable alternatives. They are the possibilities that a reasonable person would want eliminated before taking P to be true. This rough formulation takes more specific form, depending on how a particular theory uses the idea of relevance. For instance, if we use the idea of relevance in an account of a (partial) analysis of knowledge, we will say something such as this: if S knows P then S has a true belief that P and has evidence sufficient to rule out all the reasonable alternatives to P, i.e. S’s evidence can rule out all the alternatives that a reasonable person would want eliminated before taking P to be true. Alternatively, we might use the idea of relevance in an account of the semantics of knowledge claims. For example, here is how an attributor contextualism might deploy the idea: a knowledge claim “S knows that P” as made by an attributor A is true just in case P is true and believed by S, and S’s epistemic position is such as to eliminate all the alternatives to P that are reasonable as judged in A’s context of utterance. Or, we might use the idea of relevance in an account of pragmatically correct utterances, as some invariantists do. Then we’ll say something such as, “S knows P” is true when S meets invariant standards E, but “S knows P” also pragmatically imparts that all the reasonable alternatives to P in the context of utterance have been eliminated by S’s evidence.

Not only does appeal to the reasonable person standard solve the problem of relevance, it also explains and vindicates important features of our practice in making, criticizing, and retracting knowledge claims. Whichever variety of semantic account we favor—invariantist or contextualist—we can profit from an account of relevance that helps make sense of our practice. A reasonable person standard is a standard to which one has some access in virtue of common world-knowledge and a less than super-human capacity for impartiality. One can use one’s own take on what is reasonable to worry about, and in turn gauge whether one is in any position to take up the commitments of claiming knowledge. Likewise, the standard permits hearers to use their take on what is reasonable to interpret the speaker’s commitments. One is sometimes in the dark about the idiosyncratic needs and practical interests of one’s interlocutors, and coordination around a standard of reasonableness permits us to negotiate our ignorance about each other. Our need to understand each other’s knowledge claims, and to coordinate our epistemic lives more generally, is the impetus for a flexible and accessible standard of relevance, and reasonableness delivers on these fronts.¹²

6 Pragmatic encroachment

How do practical interests or stakes affect knowledge on this view? Does treating relevance as reasonableness open the door to so-called “pragmatic encroachment”?

Pragmatic encroachment is defined differently by different epistemologists. I will understand it to be the view that the nature of knowledge is (at least partly) practical.¹³ To make this claim more concrete, we can take it to mean that knowledge is a relation among several relata, some truth-related (such as the subject’s evidence) and some practical (such as the practical stakes of subject or the attributor).¹⁴ A consequence of encroachment so understood is that a change in practical interests suffices for a change in the knowledge relation.

Now, attributor contextualists can resist pragmatic encroachment by repositioning the effect of practical factors outside the knowledge relation and within the semantics of knowledge claims.¹⁵ Attributor contextualism holds that practical interests in the question of proposition’s truth only affect whether it is correct to ascribe knowledge, while the knowledge relation itself remains a relation among truth-related factors, such as the subject’s evidence, the target proposition, and alternative propositions. (Moreover, if the attributor contextualism takes the form of a “non-indexical” situation semantics, then the knowledge relation is constant across contexts of utterance, while the truth-value of a knowledge claim is determined by the particular situation in which the claim is uttered.¹⁶)

The helpful analogy here is with fitness-ascriptions.¹⁷ Being fit is a matter of all and only physical properties of a person, their body-mass index, resting heart-rate, and so on. We ascribe fitness in a way that varies with features of the context of utterance. In a context where our interests have us assessing middle-aged philosophy faculty, “Jones is fit” may be true, but in the context where our interests have us assessing Navy Seals “Jones is fit” may be false. Our varying practical interests affect the attribution of fitness, while fitness itself owes to purely physical features of the subject. Likewise with knowledge ascriptions, our varying practical interests influence our attribution of knowledge, while knowledge itself owes to purely epistemic features of the subject (i.e. her evidence, how many alternatives it eliminates, etc.).

The foregoing semantic contextualist response to pragmatic encroachment is discussed at length in the recent literature. (It is the view I personally favor.) I won’t discuss it further here. But there is a further concern about encroachment we need to address. What about views that are not contextualist, that make use of the idea of relevance? If they adopt my view of relevance as reasonableness, how can they address the worry about encroachment? Consider for example a simple invariantist relevant alternatives account of knowledge, such as Fred Dretske’s.¹⁸ On the view I’ve offered of relevance as reasonableness, this sort of account of knowledge might seem to suffer pragmatic encroachment: after all, isn’t it the case that which alternatives it is reasonable for a person to want eliminated are simply fixed by her practical interests in the question at hand? If so, then a change in practical interests suffices for a change in the knowledge relation (an increase in practical interests will suffice to make more alternatives relevant, while a decrease will suffice to make fewer relevant). And this just is pragmatic encroachment.

In what follows, I will argue that this line of thought is mistaken. The mistake lies in the idea that appeal to a reasonable person standard of relevance means that practical factors determine what is relevant. To see why, we have to understand more about the way practical factors affect the reasonable person’s judgments about what is relevant.

7 Reasonableness and practical rationality

Our first step is to see more clearly the relation between reasonableness and practical rationality. Recall we have said that a person is reasonable when she is sensitive to what is important in a situation, shows tact and proportionality in her responses and a sensible concern for others, is disposed to keep in view her own ends and aims as well as those of others, takes into consideration the facts of a situation, sizes up the reasons for and against various courses of action, including the reasons others see, and acts for the good in accord with those reasons. Does this mean that being reasonable is a matter of being practically rational? And if so, doesn’t the reasonable person standard of relevance allow practical interests to determine relevance? How does reasonableness relate to practically rationality?

There are different views of practical rationality. And different views entail different relations between reasonableness and practically rationality. For instance, on Hume’s view of the nature of practical reason, reasonableness and practical rationality come clean apart. For Hume, practical rationality is a matter of determining the means to ends that are settled independently by one’s desires or preferences. On this understanding of practical rationality there is no particular force a practically rational person will feel in the face of considerations that are provided by another person’s desires or preferences, no particular need to attend to another’s reasons. But the reasonable person is reasonable insofar as she does take into account the reasons that others see, and balances these all in deciding on a course of action. So practical rationality and reasonableness come apart, on the Humean view.

For a different view, consider Philippa Foot’s account of the nature of practical rationality. In brief summary, Foot sees practical rationality as deriving from goodness of the will.¹⁹ Here’s an analogy. An owl hunts by night, and having night vision is crucial to its success. Facts about the life of owls means night vision has a special role. If an owl does not have night vision, that is a defect. The owl’s vision is not good, we say. If an owl does have night vision, the owl’s vision is (to that extent) good. It serves an owl in the pursuit of its owly life. Analogously, a human being has a will—roughly, a disposition to consider and then act on factual considerations in matters that involve choice and following through on one’s choices. Facts about the life of humans give this feature, the will, a special role. If a person fails to consider and act on reasons, that is a defect. And if a person does consider and act on reasons, her will is to that extent good (Foot, 66). The sort of goodness here is “natural goodness”: a feature, or trait, or its operation, is good, or as it should be, if it improves the creature’s chances of carrying out its natural life functions and succeeding in its natural life cycle, in its natural habitat. A blind owl can live successfully in a zoo, but that doesn’t alter the fact that its vision is bad. It’s crucial to Foot’s view that natural goodness can be self-regarding or other-regarding, since some improvements of a creature’s chances depend on the chances of others of its kind—its mate, its offspring, or its hunting partners. Since some species are cooperative, a feature can be good because it helps a conspecific (mate, offspring) succeed in its natural life cycle. Given our highly social form of life, taking others’ interests into account is indispensable. Just as having night vision is a good thing in an owl, having a will that is responsive to factual considerations in a way that promotes successful shared human life is a good thing in a human.

On Foot’s view of practical rationality, it involves being sensitive to the various reasons for action a given situation presents and acting for the good on an appropriate balance of these reasons. And importantly, on her view, practical rationality requires the ability to weigh other-regarding reasons alongside self-regarding reasons. In consequence, practical rationality as Foot conceives it is closer to what we have described as reasonableness.

However even on Foot’s view, there are still key differences between practical rationality and reasonableness: As we noted above, it is a truism that a reasonable person will try to see things from another’s point of view. This is not part of Foot’s account of practical rationality, which only requires a disposition to weigh other-regarding reasons, and does not incorporate the demand that the person’s own reasons are formed with a view to reasons as other (reasonable) people see them. We also noted that a reasonable person is sensitive to, and able to make judgments about a variety of values; the reasonable person is thus able to weigh both theoretical and practical reasons in the same course of deliberation. There is no such general-purpose sensitivity or ability on Foot’s view of practical rationality, so reasonableness is not equivalent to practical rationality.

To recap our first step: being reasonable at first glance seems to be a matter of being practically rational, but this first glance is misleading; reasonableness is a trait that enables one to balance complex relations among practical and theoretical considerations, and among one’s practical and theoretical aims.²⁰

8 The role of practical factors in judgments about reasonable alternatives

Now let’s return to our question, does thinking of relevance as reasonableness bring pragmatic encroachment in its wake? More specifically, recall, we were considering whether an invariantist relevant alternatives theorist would be committed to accepting that practical factors determine what count as relevant alternatives, because they determine what a reasonable person wants eliminated. And this in turn would mean that whether one knows is itself determined by practical factors.

In this section, I argue that practical factors do not determine what a reasonable person wants eliminated. An increase (/decrease) in practical interests does not suffice to make more (/fewer) alternatives reasonable, and so a change in practical interests does not suffice for a change in the knowledge relation.

Start with a simple motivating case. Harley has a momentary concern about whether he has his house keys. The practical stakes are modest for him—it would be embarrassing to call his neighbor who has a spare key. He considers the possibility that he didn’t pick up his keys when he left the house. (There’s no other way he would fail to have his keys. He hasn’t used them since morning, etc.) He feels the familiar clump in his pocket, and rules out this reasonable alternative. Compare Gregor, also momentarily concerned about whether he has his house keys. Alike in every respect to Harley, except that the practical stakes for him are high—he needs to get inside without delay. Gregor considers the alternative that he didn’t pick up his keys when he left the house. Are more alternatives reasonable for him to consider? We are supposing Gregor’s case is in every respect the same as Harley’s, except in practical stakes. Does the mere increase in practical stakes make it reasonable to consider more alternatives relevant? What would these alternatives be?

That there is a question here is important. The fact is that it is not at all clear that further alternatives will reasonable for Gregor, despite his greater practical interests.²¹ This fact is prima facie evidence that greater interests don’t in themselves suffice to make more alternatives reasonable. And we have a simple explanation for this fact: if some new alternatives are relevant for Gregor, it will only be because a reasonable person would judge them in need of eliminating, and a reasonable person will not take greater practical interests to simply dictate that more possibilities need eliminating.

Here is why. I said before that the problem of relevance arises because of our lack of foreknowledge about how to satisfy our epistemic aims. I also said that the problem of relevance is resolved by appeal to a reasonable person standard because of the way the reasonable person balances epistemic and other aims. Now here’s a key point about this balancing: in judging whether an alternative is relevant, the reasonable person’s epistemic aims are primary. Such judging takes place in the course of thinking about whether to take P to be true, and this is an epistemic endeavor. In thinking about whether to take P to be true, a reasonable person will consider features of her practical situation, but because the question before her is whether to take P to be true, her aim is epistemic, and she does not lose sight of this. Consequently, the reasonable person will not allow her practical interests to decide whether to take P to be true. Specifically, she will not allow her practical interests to determine what is a relevant alternative to P. Doing so would involve allowing her practical interests to decide whether to take P to be true, and thus make her practical aim primary instead of her epistemic aim.²²

But if greater practical interests do not dictate that more possibilities need eliminating, can practical factors make any difference to what is relevant? They can. Practical factors can make a difference to what is relevant—after all, the reasonable person considers a wide array of reasons and aims, as she pursues her primary epistemic aim. But non-epistemic, practical, factors do not determine relevance.

Some cases illustrate these two points. Smith is traveling to New York and his itinerary says something about “CHI”, so he wonders if his plane stops in Chicago. He’d like to see that airport. “CHI” might mean something else—let’s suppose for the sake of argument that this alternative is the only reasonable alternative given Smith’s circumstances—so, he checks the display at the gate, which gives full city names. Now imagine Williams, whose personal interest in the question of where his plane stops is marginal, less even than Smith’s. Williams is a reasonable person, and he is aware of how little interest the question holds for him, but insofar as the question before him is whether to take the plane stops in Chicago to be true, his aim is epistemic. He is aware that “CHI” might mean something else, so this is also a reasonable alternative in his situation. The fact of his marginal practical interest does not suffice to make fewer alternatives reasonable to consider. Now imagine Jones, whose personal interest in the question of where his plane is heading is significant—he has a very important meeting in Chicago. Considering his interests, it might be reasonable for Jones to check more than Smith or Williams. He might reasonably wonder whether the display is accurate (and so he may double-check with the gate agent). Greater practical interests can make more alternatives reasonable. Acknowledging this fact does not mean holding that increased practical interests suffice to make more alternatives relevant. Imagine Vargas, who has still greater practical interest in the question than Jones—he will suffer serious harm if the plane doesn’t stop in Chicago. Do Vargas’s interests make it reasonable for him to consider more possibilities relevant than Jones does? Perhaps. But the point to note is that reasonableness is still in control here. Vargas’s greater interests do not suffice to make more possibilities relevant—what makes more possibilities relevant (if they are relevant) is the fact that a reasonable person with those interests in view would want more them eliminated before taking P to be true. Greater practical interests alone do not suffice to make still more alternatives relevant. If ever greater interests were sufficient to make more alternatives relevant, then we should expect to see more alternatives whenever interests are further elevated. A series of people with ever greater practical interests, Xavier, Zena …, whose practical interests are each weightier than the last do not have to see further possibilities as relevant.

A reasonable person is sensitive to a wide range of interests—political, ethical, social and practical—as she goes about her business. In taking P to be true, a reasonable person pursues an epistemic aim, and what she deems a reasonable alternative to P will not be determined by her non-epistemic interests, on pain of demoting her epistemic aim. Practical factors may play a role in the reasonable person’s judgments of relevance, but they do not determine what a reasonable person wants eliminated. With a reasonable person standard of relevance, we can accommodate the sensitivity of relevance to practical interests, without giving practical interests a deciding role.

9 Objections: objectivity, disagreement

I acknowledge that there are a host of questions one might still have about the reasonable person standard and the notion of reasonableness. My aim here has been primarily to motivate its use in solving the problem of relevance. However, before closing I’d like to briefly address one important line of questions centering on the objectivity of the notion of reasonableness.

Begin by noting that we talk about being “more reasonable, or “not at all reasonable” and so it seems that “reasonable” is a graded notion. One line of questioning starts from this fact. Does the reasonable person standard demand the highest level of reasonableness, or absolutely ideal reasonableness? If not, then how can the reasonable person standard deliver objectively correct claims—won’t its use mean a collapse into some sort of subjectivism?

To discuss my answer, let’s help ourselves to a particular deployment of the idea of relevance. For illustrative purposes, let’s suppose we’re working with a relevant alternatives account of knowledge paired with an attributor contextualism about the semantics of knowledge claims. That is, suppose we hold that for “S knows P” to be true in attributor A’s context of utterance, then S’s epistemic position enables him to rule out all the relevant alternatives in A’s context of utterance. Given the account of relevance as reasonableness, these alternatives are the possibilities that a reasonable person in A’s context sees as defeaters.²³

Now, we can formulate concrete instances of some of our questions about the objectivity of the reasonable person standard.

First, is there an objectively correct answer to the question, “which possibilities would a reasonable person in A’s context see as defeaters?” In answer, yes there is. But we should be clear that there are different species of objective correctness one might mean here: (i) one might say the answer to the question “what would a reasonable person see as a defeater?” depends on what people in A’s epistemic community take the answer to be. This would make for Minimal Realism about the question of defeaters; alternatively (ii) one might say the answer to the question “what would a reasonable person see as a defeater?” is determined under ideal conditions, by an ideal judge.²⁴ This would make for Modest Realism about the question of defeaters. I favor Modest Realism but haven’t the space to argue for it here.

Next, note that the attributor A does not herself have to be an ideal judge of what counts as a defeater. The reasonable person standard invokes a character, the reasonable person, who is not the subject S or the attributor A. (Of course, a competent attributor seeks to approximate the ideal judge; if she fails in this, her knowledge claims will not be reliably true.) No collapse into subjectivism about knowledge claims is in the offing.

This fact leads to another point worth noting, and that is that although the reasonable person standard is set by an ideal, it is not a superhuman ideal. Our idealization has limits—we don’t say the ideal judge of defeaters has infinite time to make decisions about which defeaters matter, or infinite world-knowledge, or infinite computing capacity. Why limit the idealization of the reasonable person standard? I fear that the argument for such limits is long, and I haven’t space to fully articulate it. Roughly, it has to do with the fact that concepts and linguistic meanings are tools for coordination, and so we need the application conditions of the concept of knowledge (or the meaning of “knows”) to be reliably deployed, on pain of making our coordination impossible. Ordinary speakers have to be able to reliably approximate our idealized judge, if their knowledge claims are reliably interpretable. And ordinary speakers do not have infinite time to make decisions about which defeaters matter, or infinite world-knowledge, or infinite computing capacity.

Perhaps these brief remarks begin to address concerns about the objectivity of the reasonable person standard of relevance. But now a different worry arises: what of the possibility that two reasonable people may differ in their judgments about which defeaters need ruling out? In answer, I believe we must acknowledge this possibility, in light of the undeniable facts about limited reasoners—as noted above, our limited nature as reasoners give rise to what Rawls calls “burdens of judgment.” In light of such facts, our not-too-idealized ideal judges may disagree over what heuristics to apply in answering a question, or how to weigh the evidence, or which prior probabilities to adopt. Consequently, two such reasonable people may differ in their judgments about what defeaters need ruling out. I want to point out, however, that this fact of disagreement is consistent with Modest Realism. A reasonable person who is a not-too-idealized judge is still an ideal judge–a character who is not simply the subject or the attributor of the knowledge claim.

But the worry continues, if we tolerate reasonable disagreement about alternatives won’t this undermine the (extent of, or possibility of) determinate truth of knowledge claims? In cases where there is no single answer to the question of what the reasonable person will see as relevant, there is no answer to the question whether one has knowledge or not. A knowledge claim in such a case is without a truth-value. If this happens routinely, many knowledge claims will be without truth-values.

This worry deserves a more discussion than I can give it here. Briefly, several responses: The very idea of knowledge claims with indeterminate truth-value does not shock. The worry must rather be that relevance as reasonableness opens the door to more indeterminacy than we expect, too much for the account to be plausible. How frequent is unresolvable disagreement among reasonable people about what is relevant to a knowledge claim? This is an empirical question, and I won’t venture a guess. Two things can be said in response to the worry about indeterminacy: (i) In many cases there will still be a determinate answer to the question of whether a subject has knowledge: a subject who can rule out all the possibilities that any reasonable person sees as relevant, determinately possesses knowledge; and a subject who cannot rule out those possibilities that all reasonable people agree are relevant, then she determinately does not possess knowledge.²⁵ Further, (ii) facts about reasonableness work to limit the frequency of destructive disagreement. Reasonable people who disagree initially, due to differences in risk aversion or other burdens of judgment, may yet reach reasonable agreement about what to treat as relevant. Expecting people to reach agreement about what is relevant is not the same as expecting them to reach agreement about moral claims. A reasonable person may be willing to do more epistemic work than they were first inclined to do, in order to satisfy other reasonable people who are more risk averse. Or more risk averse people may be willing to tolerate checking a bit less. Reasonable people may after all reach mutual agreement on what alternatives to treat as relevant, as long as they have the chance to air their differences and reflect on them.

These facts about reasonable disagreement encourage me to think disagreement is not a formidable threat to the possibility of human knowledge. What is a threat to knowledge is the possibility of our failing to be reasonable. Not just one person at a time, a small failure here or there, but widespread community-wide failure. Such a loss of reasonableness poses a much greater threat to knowledge. Loss of reasonableness brings with it the loss of the ability to rise to the standards of knowledge, and to make and assess knowledge claims. This is a much more disturbing possibility than the possibility of reasonable disagreements about evidential standards. On my view, the need to cultivate reasonableness for the sake of knowledge is an important fact revealed by thinking of relevance as reasonableness.