The Reliability Theory of Knowledge

Chapter 14 The Reliability Theory of Knowledge

Chapter Outline 1. Descartes: Knowledge Is Internally Certifiable 2. What Makes a Thermometer Reliable? 3. Relevance to the Problem of Knowledge 4. Three Concepts of Impossibility 5. To Have Knowledge, You Don’t Have to Be Able to Construct a Philosophical

Argument Refuting the Skeptic 6. A Consequence of the Reliability Theory 7. Thesis of the Relativity of Knowledge 8. What Does the Relativity Thesis Say about Skepticism?

Descartes: Knowledge Is Internally Certifiable According to Descartes, knowledge is internally certifiable. What does this mean? It means that if I know some proposition p, then there exists an argument that shows that p must be true, whose premises are either a priori true or knowable by introspection. Recall how Descartes would explain why I now know that there is a page in front of me:

• I believe that there is a page in front of me.

• My belief that there is a page in front of me is clear and distinct.

• Clear and distinct ideas are true.

•

• There is a page in front of me.

The first two premises I know by introspection—by gazing within my own mind and examining its contents. How do I know that the third premise is true? I know by introspection that I have an idea of a perfect God; I (supposedly) know a priori that an idea of a perfect being must be caused by a perfect being. Putting these two thoughts together, I deduce that God—a perfect being—must exist. The third premise is supposed to follow from this.

Notice that the word “knowledge” does not appear in the displayed argument. So what does this argument have to do with the issue of whether I know that there is a page in front of me? For Descartes, I know that the concluding proposition is true because I know the premises are true. And how do I know the premises? I know these by introspection and a priori reasoning.

I want to name the elements in this argument. There is a subjective premise, an objective conclusion, and a linking premise. The subjective premise describes what is going on in the subject’s mind (“I believe that there is a page in front of me”). The objective conclusion makes a claim about the world outside the subject’s mind (“There is a page in front of me”). The linking premise (or premises) shows how the subjective premise necessitates the objective conclusion (“If I have a particular belief and it is clear and distinct, then it is true”).

This vocabulary can be used to say what is characteristic of Descartes’s approach to the problem of knowledge. His idea—that knowledge is internally certifiable—comes to this: If the subject knows that the objective conclusion is true, then the subject must know that the linking premise is true and must know this independently of sense experience. By introspection and a priori reasoning, I can establish the required connection between what is inside the mind and what is outside it.

The theory of knowledge I’ll discuss in this chapter is very different from Descartes’. It agrees with Descartes that knowledge requires the existence of a connection between what is going on inside the mind and what is going on outside it. Knowledge requires that a particular linking premise be true. But according to the Reliability Theory of Knowledge, the linking premise doesn’t have to be knowable by introspection and a priori reasoning. In fact, this theory maintains that the subject doesn’t have to know that the linking premise is true at all. It just has to be true. This approach to the problem of knowledge was first proposed by Alvin Goldman (in “A Causal theory of Knowing,” Journal of Philosophy, Vol. 64, 1967), Fred Dretske (in “Conclusive Reasons,” Australasian Journal of Philosophy, Vol. 49, 1971), and David Armstrong (in Belief, Truth, and Knowledge, Cambridge University Press, 1973).

What Makes a Thermometer Reliable? The Reliability Theory of Knowledge claims that there is an important analogy between knowledge and a reliable measuring device. If you know that there is a page in front of you now, then your belief is related to the world outside your mind in the same way that the reading of a reliable thermometer is related to the temperature.

Thermometers are devices that form representations of temperature. The height of the mercury column is the representation; the ambient temperature is the thing represented. Thermometer readings represent temperature, just as your beliefs represent the world outside your mind. Thermometer readings can be accurate or inaccurate, just as beliefs can be true or false. What makes a thermometer reliable? Does this just mean that its readings are accurate? To see why this isn’t enough, consider a thermometer that is used just once. Suppose, on that one occasion, the thermometer said “98°F” and the temperature happened to be 98°F. The reading was accurate, but that doesn’t mean that the thermometer was reliable. For all I’ve said, the thermometer may be broken— perhaps it would say “98°F” no matter what the true temperature is. If so, it isn’t reliable, even though its one and only reading was accurate. A reliable thermometer is one for which there is a connection between readings and temperatures. If the thermometer reads n degrees Fahrenheit, then the temperature must be n degrees Fahrenheit. If a thermometer is reliable, then its reading must be correct; it can’t be mistaken in what it says. A stuck thermometer isn’t reliable, even when its readings happen to be correct. It is unreliable because its readings are correct only by accident.

Do reliable thermometers exist? I think so. The mercury thermometers we use to check whether we have fevers are examples. In saying this, I’m not denying two obvious facts. First, a thermometer can be reliable in one set of circumstances but not in another. A mercury thermometer wouldn’t be very useful for measuring temperature if it were wrapped in insulation before being placed in your mouth. The second point is that in saying that a thermometer is reliable, I’m not denying that it would be unreliable if it were broken. Hitting a reliable thermometer with a hammer will usually be enough to make it unreliable. So there are two things that help make a thermometer reliable. First, it has to be used in the right environment (for example, don’t wrap it in insulation if you want to take somebody’s temperature). Second, the internal makeup of the device has to be right (for example, the glass tube that holds the mercury can’t be broken).

Notice that reliability is an objective feature of the relationship between the thermometer and its environment. The question is whether the thermometer and its environment make the following claim true: If the thermometer says that the temperature is n degrees Fahrenheit, then the temperature must be n degrees Fahrenheit. It is an entirely separate question whether anybody realizes that this thermometer/environment relationship obtains. Whether we notice this fact is a subjective question, but whether the relationship obtains is an objective matter.

Let’s imagine that you take a thermometer out of a child’s mouth, see that it reads 99°F, and then announce: “The thermometer is reliable, and so the baby’s temperature is 99°F.” Suppose a contentious philosopher (like me) comes along and tries to refute your claim. I say:

Your thermometer isn’t reliable. It is unreliable because I can conceive of a circumstance in which its

reading would be false. For example, I can conceive of a situation in which the thermometer is wrapped

in insulation. In that case, the thermometer reading wouldn’t be correct. I also can conceive of the

thermometer’s being broken. In that case as well, the thermometer reading wouldn’t be correct. It

follows from the fact that I can conceive of these things that your thermometer isn’t reliable.

Should you be convinced by my argument? I think not. Whether the thermometer is reliable here and now has nothing to do with what I can imagine. Granted, if the situations I described actually obtained, then the thermometer wouldn’t be reliable. But from this it doesn’t follow that the thermometer is unreliable in the actual circumstances in which it is used.

Relevance to the Problem of Knowledge How does this discussion of reliable thermometers bear on the problem of knowledge? The Reliability Theory of Knowledge says that an individual knows a proposition if the individual is related to the proposition the way a reliable thermometer is related to the temperature it measures. A reliable thermometer wouldn’t say n degrees Fahrenheit unless the temperature were n degrees Fahrenheit. An individual knows that there is a page in front of her precisely when she wouldn’t have believed that there is a page in front of her unless there were one there. Another way to express this idea is by using the concept of causality. A thermometer is reliable in a given circumstance if the only thing that could cause the thermometer to read n degrees Fahrenheit is that the temperature really is ndegrees Fahrenheit. Similarly, S knows that there is a page in front of her in a given circumstance if the only thing that could cause S to believe this is that there really is a page in front of her.

I hope that the connection between thermometers and knowledge is starting to become clear. Suppose S believes that there is a page in front of her. We want to know whether this belief is an instance of knowledge. The answer should depend on the actual relationship that obtains between S and her environment. Suppose S’s sensory system is functioning normally; she isn’t hallucinating, for example. In addition, suppose that there are no evil demons around who might choose to provide S with misleading evidence. If so, it may be true that S’s sensory system is a reliable indicator of the presence of a page. If you want to answer the question of whether S knows that there is a page present, it will be entirely irrelevant to point out that if S had taken a hallucinatory drug or if S were plagued by an evil demon, then S’s belief would be (or might be) false. These hypothetical considerations do nothing to undermine the claim that S’s sensory state is a reliable indicator of what is going on in her environment.

Here’s how the Reliability Theory of Knowledge characterizes what knowledge is:

• S knows that p if and only if

1. S believes that p.

2. p is true.

3. in the circumstances that S occupies, if S believes that p, then p must be true.

The third condition can also be formulated in either of the following two ways:

• In the circumstances that S occupies, S wouldn’t believe that p unless p were true. • In the circumstances that S occupies, it is impossible that S believes p and p is false. Notice that the Reliability Theory of Knowledge makes use of the concepts of necessity and impossibility. To understand this theory of knowledge, we need to look more carefully at what these concepts mean.

Three Concepts of Impossibility I want to discuss three kinds of impossibility. Consider the following three statements:

1. Joe can’t be a married bachelor. 2. Joe can’t go faster than the speed of light. 3. Joe can’t tie his shoes now.

The word “can’t” in each of these indicates that something is impossible, but different kinds of impossibility are involved.

The first statement is logically necessary. It has to be true, just by virtue of logic and the definitions of the terms involved. If definitions and their deductive consequences are a priori true (Chapter 8), then (1) expresses a necessary truth that is a priori. In contrast, statement (2) isn’t a priori. We know that it is true because Einstein’s theory of relativity—an a posteriori theory—says that it is so. Statement (2) is said to be nomologically necessary (from nomos, meaning “law”). It is necessary because of a law of nature, in this case, an a posteriori law of physics.

Statement (3) differs from the first two examples. Its truth doesn’t follow from logic or definitions. Nor does its truth follow from any law of nature—physical or biological or whatever. Rather, to see that (3) is true, you have to take into account particular facts about Joe. For example, suppose that Joe can’t tie his shoes now because he is carrying several bags of groceries. I’ll call this third sort of necessity circumstantial necessity. By this I mean that the statement is necessarily true because of facts about the circumstances in which Joe finds himself. There is an additional fact about type (3) necessity that I should note. Whether we judge (3) true or false depends on how we interpret it. That is, (3) is ambiguous. If we interpret it to mean that Joe can’t tie his shoes because he is otherwise occupied, (3) may be true. Suppose, however, we take (3) to mean that even if Joe put down the bags of groceries, he still couldn’t tie his shoes. If that is how we interpret (3), we may

judge it to be false. This is an important feature of statements that are circumstantially necessary—they are ambiguous.

When I say that a thermometer is reliable, I mean that it and its environment are related in a special way. I’m saying that its circumstances are such that its readings must be correct. In saying this, I’m using the concept of circumstantial necessity. Similarly, when I say that S knows that there is a page in front of her, I’m saying that she is related to her environment in a special way. I’m saying that she can’t be mistaken in believing what she does. Here again, I’m using the concept of circumstantial necessity.

Suppose that a real printed page is the only thing that could get S to believe that a printed page is in front of her. Her senses are functioning normally. There are no evil demons lurking about who provide misleading evidence. If this is so, and if S subsequently believes the proposition in question, then S will know that there is a printed page in front of her. This is what the reliability theory says. In this circumstance, her belief will be related to the world the way the reading of a reliable thermometer is related to the temperature.

To Have Knowledge, You Don’t Have to Be Able to Construct a Philosophical Argument Refuting the Skeptic It is important to notice that the claim that S knows some proposition p isn’t refuted by the fact that S may not be able to defend her knowledge claim against clever philosophical interrogation. If I asked S how she knows that her impressions aren’t caused by an evil demon, she may draw a blank. If I ask her how she knows that she isn’t dreaming, she may admit that she can’t construct an argument proving that she isn’t dreaming. Such philosophical puzzles may even lead Sto say, “I guess I don’t know that there is a printed page in front of me.” But her comment doesn’t show that she lacks knowledge. Rather, it shows only that Sdoesn’t believe that she knows there is a printed page in front of her.

If S is like a reliable thermometer, she has knowledge. Being like a reliable thermometer, however, doesn’t require that S has the ability to construct fancy philosophical arguments that show she is like a thermometer. Thermometers can’t construct philosophical arguments, yet they are sometimes reliable. S may be similar. She may be unable to refute the skeptic, but that doesn’t show she lacks knowledge. S may even be a skeptic herself. She may believe that she lacks knowledge, but that doesn’t entail that she really lacks it. Just as it is possible to mistakenly believe that a thermometer is unreliable, so it is possible to mistakenly believe that an individual lacks knowledge. In fact, it is possible to have false

beliefs about one’s own situation; S can believe that she lacks knowledge and yet be mistaken.

The KK-Principle According to the Reliability Theory of Knowledge, knowing a proposition doesn’t imply that you know that you know it. That is, the reliability theory rejects what is called the KK- principle. The following several paragraphs about this principle will be somewhat tough to grasp—so take a deep breath, read them slowly, and then read them again. The KK-principle says that if S knows that p, then S knows that she knows that p. But I’ve just argued that the following is possible: S knows that p, but S doesn’t believe that she knows that p. It follows that S might know that p even though she doesn’t know that she knows that p.So if I’m right, the KK-principle is false. If the KK-principle were true, S could prove that she lacked knowledge just by becoming a philosophical skeptic. This would be very convenient—wouldn’t it be nice to be able to prove some thesis just by believing it? Suppose S believed, maybe for no reason at all, that she didn’t know that p is true. The KK- principle would conclude from this that, indeed, Sfailed to know that p is true. To see why, consider the following argument:

1. If you don’t believe that you know that p, then you don’t know that you know that p. 2. If you don’t know that you know that p, then you don’t know that p.

If you do not believe that you know that p, then you don’t know that p.

The argument can be symbolized as follows:

1. If –B[K(p)], then –K[K(p)]. 2. If –K[K(p)], then –K(p).

If –B[K(p)], then –K(p).

The argument is deductively valid (what is its logical form?). Notice that premise (1) is correct; it follows from the fact that knowing a proposition requires that you believe the proposition. Premise (2) would be true if the KK-principle were correct. So if the KK- principle were correct, believing that skepticism is true would be enough to ensure that skepticism is true. However, if knowing is like being a reliable thermometer, this has got to be wrong.

The Reliability Theory of Knowledge holds that whether an agent knows something is settled by the objective relationship that obtains between the agent’s belief and the environment. Whether the agent believes that this relationship obtains is irrelevant. The reliability theory rejects the KK-principle. In summary, the Reliability Theory of Knowledge explains why S knows that there is a page in front of her by describing the following argument:

• S believes that there is a page in front of her.

• In the circumstances S occupies, she wouldn’t believe that there is a page in front of her unless there were a page in front of her.

•

• There is a page in front of S.

This argument doesn’t use the word “know.” So what does it have to do with knowledge? According to the reliability theory, S knows that the concluding proposition is true because the premises are true. But S doesn’t have to know that the premises are true, nor does S have to produce an argument independent of sense experience for the premises. This shows why the reliability theory rejects Descartes’ claim that knowledge is internally certifiable.

One virtue of the reliability theory is that it explains what is wrong with a standard skeptical argument. The skeptic claims that S doesn’t know that there is a printed page in front of her on the grounds that it is possible to imagine that S is deluded in some way. You can imagine a situation in which S believes what she does but she is mistaken. The reliability theory shows why this act of imagination is irrelevant to the question of whether S has knowledge in the real world situation that she occupies.

A Consequence of the Reliability Theory To conclude this discussion, I want to consider an implication that the reliability theory has. It involves a fact about circumstantial necessity that I mentioned before. This is the fact that claims of circumstantial necessity are often ambiguous. I’ll restate the point and then show how it is relevant to the reliability theory.

Consider the statement “Joe can’t tie his shoes now.” Suppose I make this comment to you while we are looking at Joe, who is carrying two heavy bags of groceries. Is the quoted remark true? Here are two ways of interpreting it:

1. If Joe tries to tie his shoes while holding the bags of groceries, he will fail. 2. If Joe tries to tie his shoes after first putting down the groceries, he will fail.

If we interpreted “Joe can’t tie his shoes now” to mean (1), it will be true; if we interpreted it to mean (2), it will be false. Which reading you choose affects whether you will say that the quoted claim is true.

To see how this idea applies to the Reliability Theory of Knowledge, let’s switch examples—from knowing that there is a printed page in front of you to knowing that there is a barn in the field next to the road on which you are driving. Suppose one day you go for a drive in Dane County, which is where Madison, Wisconsin is located. On

your drive, you look at a field and say, “There is a barn in that field.” Suppose that the proposition you’ve asserted is true. You believe the proposition. Do you know that it is true? The reliability theory says that we must ask whether, in the circumstances you occupy, there is anything besides a real barn that could have caused you to believe a barn is present. The problem I want to focus on is that it isn’t clear how to decide what is included and what is excluded by the expression “in the circumstances you occupy.”

First, I need to define the idea of a fool’s barn. Fool’s gold looks like gold, but isn’t gold at all. A fool’s barn is something that looks like a barn but isn’t one. Suppose there are no fool’s barns in Dane County. The only things that look like barns are real barns. This means that if I describe the circumstances you occupy by saying that you are in Dane County, then it will be true that the only thing that could have made you think a barn is present is a real barn. Hence, you know that there is a barn in the field.

Now let’s broaden our vision. In Hollywood, there are fool’s barns. These are the facades used in movie sets. When viewed from one angle, they look just like barns, but they aren’t buildings at all. If you were driving around Hollywood and came to believe that a barn is present, your belief would not count as knowledge (even if your belief happened to be true). The reason is that, around Hollywood, there are things besides real barns that can make you believe that a barn is present.

So if your circumstances are restricted to Dane County, you might know that a barn is present; if your circumstances are restricted to Hollywood, you wouldn’t. But now consider a puzzle. I initially suggested that the circumstances you occupy are limited to the objects in Dane County. Given this array of objects, I argued that a real barn is the only thing (in the circumstances you occupy) that could cause you to think that a barn is present. But why describe your circumstances so narrowly? Why not describe your circumstances as including all the objects in the United States? If I describe your situation in this way, it will be false that the only thing that could cause you to believe a barn is present is a real barn.

Here is the point: When we assert or deny that the only thing in the circumstances that could cause you to believe that a barn is present is a real barn, we are making reference to the environment you occupy. The environment may be thought of as composed of a set of objects. If we describe that set narrowly, it may be true that the only thing in the set that could cause you to believe that a barn is present is a real barn. If we describe the set more broadly, this may no longer be true. So whether you know, on your drive through Dane County, that there is a barn on the hill next to the road depends on how we choose to describe the “circumstances you occupy.” There are many true descriptions we might select. Narrow ones will entail that you have knowledge; broader ones will entail that you do not.

Thesis of the Relativity of Knowledge The claim I just made about knowledge may be put this way: Knowledge is relative. Whether S knows that p depends on (is relative to) a choice. “S is in Dane County” and “S is in the United States” are equally correct descriptions of the circumstances that S occupies. Which one we consider depends on our interests.

Let’s be clear what this relativity thesis asserts. When you say that something is relative, you should always be prepared to say what it is relative to. The thesis that knowledge is relative is the thesis that whether an agent knows a proposition depends on (is relative to) a specification of his or her circumstances. These circumstances can be specified in different equally correct ways, and so there is no unique answer to the question of whether the agent has knowledge.

Let’s consider a down-to-earth example of relativity that does not concern knowledge. Mary and Alice are walking down the street, side by side. Is Mary walking to the left of Alice? Well, that depends on the point of view. If you look at them from the front, you get one answer. If you look at them from the back, you get the other. The key thing to notice is that x’s being to the left of y isn’t a relation between x and y alone. It involves some third item, z.

Does S know that p? The suggestion now being considered is that the question is incomplete, that is, whether S knows that p depends on a third item besides the agent S and the proposition p. It depends on a specification of S’s environment. There are different true ways to describe S’s environment. S is in Dane County. But it is also true that S is in the United States. Relative to one specification of the environment—call it E 1—S knows that p. But relative to another equally true specification of the environment—call it E 2— S doesn’t know that p.

What Does the Relativity Thesis Say about Skepticism? The thesis that knowledge is relative has an interesting implication about skepticism. Skepticism, recall, is the position that people don’t have knowledge. Its opposite is the more common-sense idea that people often (if not always) know the propositions they believe. If the relativity thesis is true, then each of these theses is true in one sense but false in another. If the agent’s environment is given a very broad specification, then skepticism is true. If the environment is given a narrower specification, then the common-sense position will be correct. Notice, however, that there is no conflict between the following two claims:

1. S knows that p, relative to E1. 2. S doesn’t know that p, relative to E2.

These statements don’t conflict with each other, any more than the next pair conflicts with each other.

3. Mary is walking to the left of Sue, when they are viewed from the front. 4. Mary is not walking to the left of Sue, when they are viewed from the back.

So the dispute between skepticism and common sense seems to end in a stalemate, not in a victory for either side, if the relativity thesis is correct. Each position is correct in one sense but incorrect in another. This consequence follows from the Reliability Theory of Knowledge once we acknowledge that claims about circumstantial necessity are ambiguous.

What’s Relative about Einstein’s Theory of Relativity? Sometimes people say that Einstein proved that “Everything is relative.” This isn’t true. Neither is the more modest claim that Einstein proved that everything in physics is relative.

To say that a statement is relative isn’t to say that it is subjective or arbitrary. Rather, it is to say that the statement is incomplete in a particular way. Whether Sue is walking to the left of Mary is a relative matter. That means that there exists some third item, beyond the two people mentioned, that must be referred to if the statement is to be true or false. To say just that Sue is to the left of Mary is to fail to express a complete thought.

It takes no enormous insight or creativity to see that whether one person is walking to the left of another is a relative, not an absolute, matter. In contrast, Einstein’s idea that simultaneity is relative is anything but obvious. It was a brilliant theoretical conjecture, something that isn’t at all suggested by our common-sense talk about space and time. Consider two events that occur at different places. Common sense suggests that either the events are simultaneous with each other or they aren’t. Einstein’s theory says that this isn’t true. Whether two events are simultaneous depends on the choice of some third item— something called a rest frame. Relative to one rest frame, the events are simultaneous, but relative to another equally correct rest frame, they aren’t. Simultaneity is relative, not absolute. Einstein didn’t say that everything in physics is relative. For example, in the special theory of relativity he defines a quantity called the space-time interval. This quantity measures the amount of separation that there is between events; it takes account of both spatial and temporal distances. The space-time interval between two events isn’t relative to the choice of a rest frame. The space-time interval is absolute, not relative.

Review Questions

1. At the beginning of this chapter, I said that Descartes held that knowledge is “internally certifiable.” What does this mean? Does the Reliability Theory of Knowledge agree?

2. What is the difference between logical necessity, nomological necessity, and circumstantial necessity?

3. What does it mean to say that claims about circumstantial necessity are “ambiguous”?

4. What does it mean to say that a thermometer is “reliable”? What analogy does the Reliability Theory of Knowledge see between reliable thermometers and knowledge?

5. How does the Reliability Theory of Knowledge assess the following skeptical argument: “I can imagine that my senses are now malfunctioning. Hence, I don’t know now that there is a printed page in front of me.”

6. What does it mean to say that knowledge is “relative”? Relative to what? 7. Does the relativity thesis entail that skepticism is correct?

Problems for Further Thought 1. For any proposition p, we can construct the sentence “S knows that p.” Since

“S knows that lemons are yellow” is a sentence, we can construct the sentence “S knows that S knows that lemons are yellow.” We can repeat this operation as many times as we please. The KK-principle says that if S knows that p, then S knows that S knows that p. Formulate similar principles for truth (the TT-principle), belief (the BB-principle), and surprise (the SS-principle). Are any of these principles plausible? Defend your answers.

2. Whether a ring is made of gold depends on the materials that the jeweler chooses when the ring is made. Does this mean that whether a ring is made of gold is a relative matter? How is this example different from the two claims about relativity discussed in this chapter?

3. Do you think that nonhuman organisms have knowledge? For example, do dogs? What is required for dogs to have knowledge, according to the reliability theory? How do these requirements differ from what would be demanded by the view that knowledge is internally certifiable? If dogs have knowledge, does this count for or against the Reliability Theory of Knowledge?

MySearchLab Connections