The theories discussed in Chapter 4 gave accounts of justification elaborating on the evidentialist idea that justification is a matter of fitting one's beliefs to one's evidence. Although no decisive objection to evidentialism emerged from our discussion, many contemporary epistemologists reject that view. No one denies that evidence often matters to justification and no sensible epistemologists urge people to disregard their evidence when forming beliefs. Rather, they think that making one's belief fit one's evidence is only part of the story. The larger story, they think, brings in considerations about the processes that initiate and sustain beliefs. Four theories of this sort will be examined in this chapter.1
I. THE CAUSAL THEORY A. The Main Idea The first proposed replacement for the TAK [abbreviates: Traditional Account of Knowledge –D.] is the causal theory of knowledge. Alvin Goldman was an early proponent of this view.2 To appreciate the idea, first think about any device that receives and processes information about its environment. Familiar devices such as thermometers and barometers are examples. The temperature surrounding the thermometer and the pressure in the air around the barometer cause the measuring devices to be in particular states or conditions. We sometimes speak, metaphorically, of the thermometer "knowing" what the temperature is. In some ways we are like elaborate thermometers. When we are in the presence of a red object with our eyes open and sufficient light available, we see a red object and believe that there is something red there. There is a causal process, initiating with light reflecting off the red object into our eyes, going through our perceptual system, and culminating in the belief that there is a red object present. The causal theorist's main idea is that when a fact in the world leads to a belief in that fact, it is a case of knowledge. When a person has a belief that is not causally connected with the associated fact, there is no knowledge.
This central idea needs elaboration and clarification, but it does seem to have some initial plausibility. We can initially formulate the idea this way:
C. S knows p iff S's belief in p is caused by the fact p.
The causal theory retains the TAK's implication that knowledge requires true belief. Suppose you believe p, but this belief is false. If p is not true, then there is no fact p, and hence it cannot be that the fact p caused your belief in p. And if you do not believe p, then, of course, it is not the case that the fact p causes your belief in p. So, the truth and belief conditions are retained by this theory. But the justification condition is replaced by the causal connection requirement.
Goldman concludes the essay in which he defended this theory by saying that it "flies in the face of a well-established tradition in epistemology, the view that epistemological questions are questions of logic or justification, not causal or genetic questions."3 His idea, then, is that whether you have knowledge depends not on what your reasons are but rather on what the cause of your belief is. A key aspect of the theory, then, is that it eliminates the TAK's justification condition and the fourth condition designed to deal with Gettier's examples and replaces them with a causal connection condition. As we examine this theory, we will look to see if this replacement is a good idea.
To see the merits of the causal theory, notice how well it deals with common examples. Consider first some cases in which a person has a true belief that is not knowledge. Suppose that you have to be out of town on the day of a scheduled picnic. At the time of the picnic, without having heard any reports about the weather back home, you form the belief that it is raining at home. You have this belief only as a result of your nasty disposition: you do not want the others to have a good time that you must miss. And suppose that you happen to be right. In this case, you have a true belief that it is raining in your hometown, but you lack knowledge. And, just as the causal theory would have it, there is no causal connection between the fact that it is raining and your belief. So the theory gets this case exactly right. If, on the other hand, you believed that it was raining because your friends called you and told you that it was raining, and they had seen the rain, then there would be a causal connection going from the rain, through your friends, and on to you. So you would have knowledge. Thus, the theory gets this case right as well. And, of course, if you had seen the rain yourself, there would have been a causal connection and you would have had knowledge. These simple cases seem to work out very well.
Even some more complicated cases go well for the causal theory. Consider Gettier cases such as Example 3.3, in which Smith believes that there is a sheep in the field. Smith's belief is true, but the presence of the sheep did not causally contribute to that belief. The belief traces back to the dog (or the statue, or, . . .). So, the causal theory seems to get this right as well.
B. Developing the Causal Theory: Goldman's Theory To appreciate the merits and the problems of the causal theory, it is best to examine a specific version of it. Here is what Goldman proposed:4 C*. S knows p if and only if the fact p is causally connected in an appropriate way with S's believing p.
There are silly, but convincing, examples that make it clear why Goldman includes the requirement that the belief and the fact be connected in an "appropriate way." This can best be seen by considering the following example involving an inappropriate connection.
Example 5.1: The Blow to the Head
Gerald has fallen down the steps and hit his head. The blow to his head scrambles his brain in such a way that he forms a variety of wild beliefs. Among other things, he believes that eating lettuce causes obesity, that the Chicago Cubs will win the World Series, and that he has just fallen down the steps. In fact, he has no recollection of the sensation of having fallen. This belief, like each of the other two just mentioned, is simply a direct result of the blow to the head.
In this example, Gerald believes
1. I, Gerald, have just fallen down the steps.
Gerald believes (1), and there is a causal connection between the fact that (1) and Gerald's belief that (1). Still, clearly, Gerald does not know (1) if he comes to the belief in this way. In something like what happens in Gettier cases, he has fortuitously (in some sense) come upon a true belief. Goldman says that although there is a causal connection between the fact and the belief in this example, the connection in this sort of case is not an appropriate connection. This example illustrates why (C*) is better than (C).
Goldman gives the following examples of appropriate causal connections between facts and beliefs: perception, memory, and proper reconstructions of causal chains.5 The first two of these are relatively easy to understand: If seeing a tree causes you to believe that there is a tree in front of you, there is an appropriate connection and you have knowledge. If a memory causes you to retain that belief, you still have knowledge. Some cases are more complex, and they involve mental reconstructions of causal chains.
Example 5.2: The Missing Tree
Smith returns home and sees a lot of sawdust and wood chips on the ground where there used to be a large tree. He recalls that the tree was diseased and that he received a notice stating that the city tree service was going to cut it down. So he infers, and thereby comes to know, that the tree was cut down.
In this case, Smith did not perceive the tree being cut down and he does not remember that it was cut down. But he does know that it was cut down. Goldman's theory accounts for this fact by allowing for appropriate connections between a fact and a belief when the belief results from a proper reconstruction of the causal chain leading from the fact to the belief. Smith is able to reconstruct in a sensible way the causal chain involved and thus has knowledge. In this case, there is a causal chain leading from the cutting down of the tree, to the presence of wood chips and sawdust, to the belief that the tree was cut down. Because he can reconstruct the causal chain from the tree being cut down to his belief, Goldman counts this as a proper reconstruction of a causal chain and thus it makes for an appropriate causal connection. His theory implies, correctly, that there is knowledge in this case.
Some other properly reconstructed causal chains differ slightly from this one. Suppose Black sees a fire in the fireplace and believes that there is smoke coming out of the chimney. The smoke coming out of the chimney does not cause her belief that there is smoke coming out of the chimney. Rather, the fire causes both the smoke to come out the chimney and her belief that it is. In this case, then, there is something, namely the fire, that leads to both a certain fact—the smoke coming out—and her belief in that fact. In other words, there is a common cause of the fact and her belief in the fact. Goldman says that when a person reconstructs a causal chain like this correctly, the fact and the belief are causally connected in an appropriate way and the causal condition in his analysis is satisfied. Thus, the theory implies, correctly again, that she does know that there is smoke coming out of the chimney.
Goldman's idea is clever and innovative. It deals with many common examples effectively. It may also deal with many Gettier-style examples in a good way. Typically, in those examples, there is no causal connection, or no appropriate one, between the fact that p and the belief in p.
However, the causal theory is not without serious problems. We turn next to them.
Puzzles and Problems for the Causal Theory C1. Knowing Generalizations The causal theory works best when we restrict our attention to knowledge of specific observable facts in the world. But The Standard View implies that we know more than that. For example, it allows that we can have knowledge of generalizations such as
2. All men are mortal.
The problem is that generalizations do not seem to be causes. (2) does not cause your belief in (2), so there is no causal connection, and thus no appropriate causal connection, between this fact and your belief in it. It is belief in specific instances of mortality (and other information) that causes the belief. So, how could you know (2), according to Goldman's causal theory?6 C2. Overdetermination Cases
Example 5.3: Poison and Heart Attack
Edgar knows that Allan has taken a fatal dose of a poison that has no antidote. Enough time has passed so that Edgar knows that Allan is dead. But Allan didn't die from the poison; instead, so upset at what he had done, he had a fatal heart attack. Edgar's reconstruction of the causal chain is not correct.
This is a problem for Goldman's theory because Edgar does know that Allan has died but has not reconstructed the causal chain properly. In this case, Edgar knows something—that Allan has taken a fatal dose—that is causally sufficient for Allan to be dead. On this basis Edgar can know that Allan is dead. It is not required that this be the actual cause of Allan's death. You can know that some-thing happened even if you are mistaken about how it came about. You do not have to reconstruct the causal chain properly.7
This is a very serious problem for the causal theory. To repair the theory, it would be necessary to say that there is a proper causal chain between an event and a belief in cases in which a person believes that the event has occurred be-cause he justifiably believes that something causally sufficient for it has occurred. But this brings the idea of justification (and evidence) back into the picture in exactly the way Goldman wanted to avoid. The truth is that knowledge just does not require that there be a causal connection between an event and a belief in its occurrence. It requires justified belief.
C3. Perception and Evidence
8Example 5.4: The Trudy/Judy case
Trudy and Judy are identical twins. Smith sees one and, for no good reason, forms the belief that he sees Judy. It is true, and it is a case of perception. He reconstructs the causal chain between Judy's presence and the belief properly. He knows about Trudy, but rashly discounts the possibility that she is the one he sees.
This is also a serious problem for the causal theory. Here, there is a causal connection between a fact and the belief in that fact. Nevertheless, there is no knowledge. And the reason is that the reasons for believing that the causal chain is the way it is are not good ones. The mere existence of a causal connection is not good enough.
A causal theorist could develop the theory in a way that deals properly with this case. The idea would be to require that the believer be warranted or justified in reconstructing the causal chain in the proper way. Smith is not justified in this case, since he has no good reason to think he sees Judy rather than Trudy. But the belief in this case is just a perceptual belief—Smith just sees a person and forms the belief that it is Judy. That is just like a case in which he sees a table and forms the belief that it is a table. If we say that he needs warranted beliefs about the causal history in the Trudy/Judy case, then the same should be required in a case in which he forms a true belief that there is a table there. The result of these considerations is that any good version of the causal theory will not be different from the TAK in the way Goldman says it is.
There is a dilemma for causal theorists: either they do require "explicit evidence about the causes of perceptions or they do not. If they do not, then the Trudy/Judy case is a counterexample because there is a causal connection that the believer has reconstructed correctly (though unjustifiably). If, on the other hand, causal theorists do require explicit evidence about the causes of perception, then the causal theory turns out not to differ significantly from the TAK. The central difference is that it adds the causal connection condition, which Example 5.3 shows to be a mistake.
D. Conclusions on the Causal Theory The idea that knowledge can be analyzed in terms of causal connections rather than justification is an interesting one, but it seems not to work out very well when it is considered in detail. It is true that there is often a causal connection between the facts we know and our beliefs in those facts. But we can have knowledge without there being such a causal connection, as Examples 5.2 and 5.3 show, and we lack knowledge even when there is a causal connection, as Example 5.4 shows.
II. TRUTH TRACKING A. The General Idea Recall the idea used to introduce the causal theory: thermometers "know" what the temperature is since they respond to the temperature in a regular way. The causal theory did not make full use of this idea. What is true of the thermometer is not just that its being 50° causes the thermometer to display "50°." When it is that temperature, the thermometer says that it is, and when it is not that temperature, the thermometer says that it is not. Something analogous is true of people who know about the world around them. Consider, for example, a dog owner and her dog, Rex. When Rex is in the room with her, she believes that he is. Rex makes his presence felt, and she is receptive to his ways of doing this. When he is not in the room, she believes that he is not in the room. Her beliefs about this topic "track the truth," in somewhat the way the thermometer tracks the temperature.
One idea, then, is that knowers are "truth trackers." Robert Nozick was a leading proponent of this theory" The idea of truth tracking as an account of knowledge is this: a person who knows a proposition, a knower, is a person who tracks its truth. Just as a thermometer's readout tracks the temperature, a knower's attitudes toward a proposition reflect the truth value of the proposition. This idea is expressed in the following definition:
TT. S knows p iff (i) p is true, (ii) S believes p, (iii) S's attitude toward p tracks the truth value of p: When p is not true, S does not believe p; and when p is true, S does believe p.9 Like the causal theory, the tracking theory retains from the TAK the idea that knowledge requires true belief. But it replaces the justification condition and the Gettier condition with the tracking requirement.
The tracking theory works well for some simple cases. An office worker knows that his computer is now turned on, but if it were off he would believe that it is off. His beliefs about the computer's condition track its actual condition. He does not know whether or not his neighbor is home, though he thinks she is. But that thought is based on vague recollections of general patterns. Even if she were out, he would still think she is at home. His beliefs do not track her location.
Principle (TT) has the right results for many other cases. It seems to have the right results in the cases that caused problems for the causal theory. The theory can even handle some Gettier cases quite effectively as well. For example, in one of the cases Smith did not know that someone in his office owned a Ford. He based his belief on the Nogot evidence but his belief was true because of Havit. However, if his belief were false, perhaps because Havit sold his Ford, Smith would still have had the Nogot evidence and would still have believed that someone in the office owned a Ford. So, if the proposition that someone in the office owned a Ford were false, he would have believed it anyway. He was not a truth tracker for this proposition, and thus, according to (TT), did not have knowledge.
B. Developing the Tracking Theory: Nozick's Theory Nozick argued that a theory along the lines of (TT) is not quite right as it stands and that it needs some "refinements and epicycles."10 He is surely right about that. He presents examples such as the following to bring out the reason for this:
Example 5.5: Lucky Knowledge
Black is hard at work in her office. From time to time she looks up from her desk and computer to stretch her neck. On one such occasion she happens to glance out the window toward the street. Just at that moment she sees a mugging on the street. She has a clear view of the event. She is a witness.
In this case, Black knows that a mugging has occurred. Yet the tracking condition is not satisfied. It could easily happen that Black does not glance out the window at the crucial moment. In that case, the mugging still takes place, yet she does not believe it. She is not a truth tracker for the proposition that a mugging is taking place. So Example 5.5 is a counterexample to (TT). Black does have knowledge, but she is not a truth tracker.
Another example illustrates a similar point.
11Example 5.6: The Grandma Case
Old frail Grandma sees grandson Johnny playing happily before her. She knows that Johnny is well and playing happily. But suppose Johnny were sick. The family would tell Grandma that Johnny is well and playing happily, but that he's doing this at a friend's house. They do not want to make Grandma worry. So, if Johnny were sick, she would still believe that he is well.
Grandma does not track the truth. Yet she has knowledge, since she actually does see him.
Nozick has a proposal on how to fix things up.12 He suggests that what goes wrong in these cases is that there is a switch in the methods by which the per-son comes to the belief. Simplifying slightly, we can attribute the following view to him:
TT* S knows p iff (i) p is true, (ii) S believes p, (iii) S used method M to form the belief in p, and (iv) when S uses method M to form beliefs about p, S's beliefs about p track the truth of p.13 The idea is that to know a proposition, you must be a truth tracker for the proposition when you stick to the same method for forming a belief about it. In the Grandma Case, she actually uses "looking at him" as her method. As the example is told, if he were not well and she had been told that he was well, she would then be using a different method. The fact that this other method would lead her to a false belief does not matter when (TT*) is applied to the case. To apply (TT*), we must consider what would happen if he were not well and she were to use the method of looking at him to come to her belief. Presumably, under those circumstances, she would see that he is sick and would not believe that he is well. Thus, if she sticks to this method, she gets things right. So, this deals with Example 5.6.
C. Problems for The Tracking Theory C1. Tracking Is Not Necessary for Knowledge Consider again Example 5.5, the Lucky Knowledge case. This is an example in which Black does have knowledge, but she could just as easily have failed to have knowledge because she could easily have glanced out the window at a slightly different time. In this case, condition (iii) of (TT) is not satisfied. Defenders of the theory might think that (iv) of the revised analysis, (TT*), is satisfied. However, this does not seem to be right. If the method in question is "glancing out the window," then the revised analysis is no better than the original. The key question is whether condition (iv) is satisfied in this case. That condition, applied to the example, yields:
When Black uses the method of glancing out the window to form beliefs about whether thereis a mugging occurring there, her beliefs track the truth (that is, if there were a mugging, she typically would believe that there is; and if there were no mugging, she typically would not believe that there is one.)
If we just add a minor detail to the story, it will be easy to see that this condition need not be true even if Black does actually know that the mugging has happened. Suppose that Black has a view of only part of the street outside her window. There is a large part she cannot see. In our example, Black's method, looking out on the street, enabled her to know that there was a mugging. But it could quite easily have failed. That is, it could have led her to believe that there was no mugging even if there had been one. It would have done this if the mugging had been in a slightly different place, just out of her view. So the method she used—looking out on the street—is not all that reliable for forming a correct belief about whether or not there is a mugging on the street.14 She does not track the truth even when she uses that method. So the theory implies that she does not know that there is a mugging going on even when she sees it clearly. This is a mistaken outcome.
This is a forceful objection to the tracking theory of knowledge. It shows that one can have knowledge without being a truth tracker, even if one sticks to the same method for forming a belief about that proposition. Perhaps the example is not entirely decisive since it is less than clear what counts as using the same method. Perhaps a defender of the theory could say that when Black looks out on the street and does not see a mugging, then the method of belief formation differs from the method used when she does see the mugging. Lacking any clear idea of exactly what counts as using the same method, it is difficult to be sure about this. At the very least, then, this objection reveals that the theory is unclear in this crucial way.