Epistemology: The Theory of Knowledge
Epistemology is the philosophical examination of human knowledge. One of the central problems that faces an epistemologist, a philosopher engaged in the examination of knowledge, is how to refute the epistemological skeptic. Such a skeptic should be clearly distinguished from an ontological skeptic.
The former claims that we have no knowledge of the external world, the past, the future, or the minds of others. The ontological skeptic, however, goes beyond this- his skepticism is about whether anything exists other than himself. He has doubts about whether there is a world or any mind external to himself, whether anything happened before a few minutes ago, and whether anything will happen in the future. The epistemological skeptic can consistently accept our usual beliefs about past, present, and future. Knowledge, however, is more than mere belief, and this skeptic's claim is that we lack knowledge about these things.
He can, unlike the ontological skeptic, even be convinced that our usual common sense beliefs are true, because, as Plato points out in his Socratic dialogue Theaetetus, knowledge is more than true judgment or belief. Often we have beliefs that luckily turn out to be true, but knowledge requires more than good luck. Socrates' example about lawyers in the Theaetetus shows this. A lawyer who is a skillful orator can persuade a jury to believe his client innocent merely by the skill of his speech. Sometimes it is indeed true that the defendant is innocent, and so the jurors, by luck, have true beliefs but certainly not knowledge. Consider another example of two people who believe there will be an eclipse on a certain day. Their beliefs turn out to be true, but the first person holds the belief because he dreamed the eclipse was going to occur, while the other believed it on the basis of precise astronomical measurements. While both people have true beliefs, only the second could be said to have knowledge. The epistemological skeptic holds that all our beliefs are like those of the first person. Some are true, but which are true is a matter of luck rather than knowledge.
It seems, then, that besides true belief we must have sound evidence or reason for our belief to obtain knowledge. That is, knowledge is justified true belief. In this way lucky true beliefs would not qualify as knowledge. But we still have not done enough to discover what is needed for knowledge beyond true belief. To see that more is needed, consider the example of someone who every day for years has seen a painting of his mother on the wall of her living room. He enters the room again one day and on looking at the wall believes, among other things, that the picture is still in the living room. He is surely justified in this belief, but what is on the wall is a very realistic photograph of the original painting, which has been hidden behind a table in a corner of the room. Thus he has a justified true belief about the painting's being in the room, but lacks knowledge. Again, it is merely luck that his belief is true. Unfortunately, after many attempts by many persons, no one has yet uncovered the needed fourth condition to add to justified true belief to reach knowledge. How leniently or strictly the fourth condition is construed has important consequences for the skeptic. Indeed, one thing we shall try to discover is we proceed is just how lenient he can be and till justify his thesis.
Consider first a very strict view of what is needed for knowledge beyond justified true belief. It might well be that a belief must be evident or certain or indubitable to be knowledge. But if this is so, then a skeptic can argue that no evidence for beliefs about the external world, the past, the future, and other minds is sufficient for certainty, and thus there is no knowledge of these. And if there is no such knowledge, then each person is limited to knowledge of himself and his own present sensations and thoughts. This does not lead to solipsism, which is the view that only I and my own thoughts and sensations exist, but instead leads to the view that I only have knowledge of myself and my own sensations and thoughts.
Descartes' Method of Doubt
We can see this skeptic's point by turning to the work of Rene Descartes, who employed what he called "the method of doubt" in order to establish an indubitable or certain foundation of knowledge. It was his opinion that all previous systems of knowledge were seriously deficient because they neither required that the evidential basis of knowledge be certain nor provided a way to extend certainty beyond the initial evidential base. As what is taken to be knowledge is extended from the initial base to include more and more about increasingly complex matters, no knowledge claim is better justified than the evidential foundation of all knowledge. But if the foundational evidence is not certain, then it can be false and the whole structure of knowledge claims erected upon it can also be incorrect. Without certainty at its base and without some way to extend certainty to other more rarified areas, the cooperative, continuous human endeavor to extend the frontiers of knowledge might be completely misguided because it would then continually lead to falsehoods.
Descartes employed his method of doubt to find what is initially or noninferentially indubitable or certain, and then proposed that knowledge be extended by deductive inference because deduction preserves certainty. His method of doubt is, basically, to doubt everything that there is some reason, however minimal, to doubt. Thus, if a situation possibly will occur, and if it were to occur there would be some reason to doubt some knowledge claim, then that claim is not noninferentially certain and it is not to be part of the foundation. Let us apply this method of doubt to the four kinds of claims that our skeptic denies are known.
First, we claim to know much about the world around us, at least at the times we seem to see, hear, feel, taste, and smell various parts of it. We surely seem to see flowers, touch paper, hear music, taste sugar, and smell perfume. At the very least, each of us is sure that he perceives parts of his own body. Consequently, we claim to have perceptual knowledge of our own bodies and the external world. However, by using Descartes' method of doubt the skeptic can show that no perceptual claim is initially or noninferentially certain because there is always some reason to doubt it. It is possible that, even when I seem to be seeing my own hand directly in front of my face, it is lying by my side; I am sleeping and dreaming that I see my hand. Further, it may seem I awake when, in fact, I merely dream I awake. The point is not that all my experiences may forever be really dreams, but only that for each one it is possible it is a dream. Thus while some experiences may be true perceptions of the external world, none is immune from doubt. Those I feel most sure about may be among those about which I am mistaken. Feeling sure is no guarantee of either indubitability or truth. No claims about the external world are initially certain.
Descartes argues, however, that some claims about the present are initially indubitable. While it is possible there are situations, such as dreaming, that would make it reasonable to doubt that I see my hand, there is no possible situation that would justify my doubting that I am doubting, or that it seems to me I see a hand, or that I, who seem to see a hand, exist. Thus, even if I am always mistaken about the external world, it is initially indubitable for me that I, who am mistaken, exist and that I have experiences and beliefs. Furthermore, it is initially certain that I have feelings such as pains and tickles and sensations of touch and taste. While my beliefs may be mistaken, and I may misconstrue what causes my feelings and sensations, it is indubitable that I have them at the same time I seem to have them. Here, then, are some claims each person can place in the category of what is initially certain for him.
Second, after I have had a belief, feeling, or sensation, my memory can be mistaken about it, and so any of my attempts to remember a belief, feeling, or sensation may lead me astray. I can always be mistaken about what is not present and manifest to me now. As Bertrand Russell pointed out, it is possible that the world was created just five minutes ago and that everything at that time was just as we remember it now. It is possible each person came into existence at that time with a bodily state that seems to result from his existing before five minutes ago and with all his seemingly correct but actually mistaken memories of a past before five minutes ago. This is a possible situation, and if it occurred this situation would make knowledge claims about the past incorrect. For this reason, no claims about the past are noninferentially or initially certain.
Third, the situation is no better for claims about the future. No belief about the future is certain unless it is based on some previous evidence. My beliefs that the sun will rise tomorrow and that I will die some day are as close to certainty as any claims about the future, but even if they are certain they are at most inferentially certain. That is, knowledge of the future depends on being inferred from some previous evidence. If I were the only living thing ever on the earth and had never seen the light of day until just yesterday, it would not be at all reasonable for me to believe that the sun will rise tomorrow or that I will die some day. It is only the vast accumulation of past evidence that makes some beliefs about the future even approach indubitability.
Fourth, beliefs about the minds of others face a similar problem. It is possible that I am dreaming or hallucinating when I seem to see someone else who is behaving in a way that leads me to believe that he is in pain, or is thinking, or feeling some emotion. But even if I were to know I was not dreaming or hallucinating, it is still possible that he is feigning, that he expresses his inner, mental life in behavior quite different from the way I express my own, or that what I see is not a man but a skillfully made automaton.
Thus the skeptic can conclude that no claims about the external world, the past, the future, or other minds are noninferentially or initially indubitable or certain. While this establishes one point he needs to justify his thesis, more is needed because it may be that some initially indubitable claims can be used as ultimate reasons to extend knowledge to the areas of initial dubitability. Indeed, though Descartes agreed that all initially indubitable knowledge of what exists or occurs is limited to those beliefs each man has about himself, and his own unique beliefs, feelings, and sensations, Descartes thought he could avoid skepticism by providing a way to extend knowledge from this very limited base.
Deduction and the Extension of Certainty
Any extension of knowledge from the ultimate evidential base involves inference, and if knowledge requires that we preserve not only truth but also certainty, we must use only deductive inferences. Basically, this is Descartes' full method: begin with certainty and extend knowledge by a method of inference that guarantees to extend certainty. There are, however, problems here that also play into the skeptic's hands.
A deductive argument, in barest form, consists of statements, one of which is the conclusion and the others of which are the premises. The premises are used to justify or provide evidence or reasons for the conclusion. The purpose of a deductive argument is to establish two relationships. The first is between the premises and the world—that is, they should be true of the world. The second is a relationship between the premises and the conclusion whereby it is necessary that if the premises are true then the conclusion is true. When the latter relationship is established the argument is valid. And when the premises are also true, the argument is sound because the two relationships together guarantee the truth of the conclusion and thus extend truth. Furthermore, if all the premises are certain, then so is the conclusion.
The question is whether there are valid arguments where the premises are noninferentially certain, and the conclusions, although not initially certain, obtain inferential certainty through the arguments. If we can find such arguments we will have extended certainty, truth, and knowledge as Descartes hoped to do. There is, however, an immediate problem. Generally, deductive arguments lead to conclusions that are implicit in the premises and thus seem unsuited for any important extension of knowledge. If we limit the premises to those that are only about one individual's beliefs and immediate sensory experience, it would seem that no conclusions about the past, future, external world, and other minds would be deducible.
However, we can try, as Descartes did, to find some general principles that relate beliefs and sensations to external objects, other times and minds, and that are initially certain.
Such principles would be conceptual truths that differ in important ways from statements that assert the existence of particular things distinct from a person and his own experiences. For example compare the general statements that all bachelors are unmarried with one that all bachelors are unhappy. Both are general because they mention all entities of a certain sort. However, while the first can be known to be true a priori—that is, known independently of any experiential or empirical evidence—the second requires an empirical investigation of actual bachelors to discover whether any are happy. The second, then, is known at best a posteriori. It is dependent upon empirical evidence gathered from observation or experimentation. The first is known once the concepts of being a bachelor and being married are understood, and this knowledge is noninferentially certain because it need not be based on or inferred from any other knowledge or evidence. We can, then, expand the list of our ultimate, or initially certain premises to contain a priori conceptual truths, including many with terms such as "bachelor" and "unhappy" that do not refer merely to one person or his own thoughts and sensations.
There are still problems, however, because each person needs conceptual truths relating himself and his own experience to other places, minds, and times. The problem can be illustrated by concentrating on extending knowledge from the Cartesian base to the present external world. We want general conceptual truths that relate persons, beliefs, and sensations to the existence of external objects in such a way that each of us can use additional premises about himself and his own experiences to deduce validly something about the existence of external physical objects. In order to do this we must take into account the ways experiences can mislead us about the external world. The sense experience of someone who is under the influence of alcohol or drugs is not a guide to the existence of external objects. The same is true of abnormal conditions such as optical illusions, which can result from special lighting conditions and unusual relationships among objects. However, there are general conceptual principles that are noninferentially certain and that relate sense experience to external objects, such as the first premise in the following argument:
1. If a normal perceiver, under normal conditions, were to have an experience of a red object, then he would perceive a red object (and so there would be a red object).
2. I now have an experience of a red object.
3. The relevant conditions and I are normal now.
4. I now perceive a red object (and so there is a red object now).
This is a valid argument, and let us agree that the first principle is initially certain. But while it is also initially certain for me that I am now having an experience of a red object, it is far from certain, let alone initially certain, that the relevant conditions and I, or anyone else, are normal at any particular time. Thus the third premise is not part of the Cartesian base of knowledge and cannot be used to extend knowledge from the base.
This illustrates the general problem. Any sound deductive inference from indubitable statements about some person's sense experience and indubitable general truths to conclusions about the existence of external physical objects requires some premise about the person who has the sense experience and the conditions in which he has them, if the inference from the particular sense experience he has to the existence of external objects is to be valid. But none of these additional premises about people and conditions are either initially certain general principles or statements merely about one person's sense experience. Thus they are not part of the Cartesian base.
Similar problems arise for the three other cases. Concerning the past, there are two different sorts of knowledge claims. One kind is a present memory claim about a past event, and one is about something not presently remembered but about which there is supposedly present evidence. The problem facing memory claims is that someone's believing that he remembers something is not enough to guarantee correctness of memory, even if it is only a claim about some inner experience he has had, such as an experience of pain. In addition, it is more difficult to find an indubitable general principle to link the present belief about the past with -what actually occurred. There is no analogue to normal observers and conditions for remembering.
Where a claim about the past is not based on memory it must be justified by testimony of others, relics and documents from the past, or scientific theories that have implications about the past. For example, someone can examine the pigment of a painting to date its past origin, or use a theory about decay of radioactive particles to determine the age of certain material. However, these means of obtaining knowledge of the past depend on present perception of external objects. If, as previously argued, we cannot extend certainty to claims about present external objects, we cannot extend it beyond them to implications of present testimony, relics, documents, and scientific theories for the past. But even if we could reach inferential certainty about the present facts, additional general principles are needed to obtain certainty about the past. Again, however, none of the needed principles is initially indubitable. No statement relating the past to present pigments, radioactive material, testimony, or documents is initially certain. Each is at most probably true relative to present empirical evidence, and thus it is not a conceptual or a priori truth.
The same problems arise for extending certainty to claims about the future and other minds. Claims about both of these would obtain certainty only by inference from indubitable claims about the present. But for other m*nds the relevant claims about the present concern presently observed behavior, such as screaming or writhings. For the future, present claims with implications for the future are needed. These can be general principles, such as the claim that night follows day, or particular statements, such as the claim that this piece of radioactive material will become lead in so many years. But there is no initial certainty about any of these claims because each depends on present observations of the external world. As with the past, we have found no way to extend certainty to specific claims about present observations, let alone extend it from them to claims that go beyond present observations by implying something about the future.
An epistemological skeptic can use Descartes' method of doubt to show that initial certainty accrues only to certain conceptual principles and to each person's present claims about his own existence, beliefs, sensations, and feelings. The skeptic can also show that Descartes' method of extending certainty and knowledge from this base (the method of deductive inference) does not succeed. For many skeptics this is enough to establish their thesis, but others, such as David Hume, wish to go even further. Hume, contradicting Descartes, claimed that no one has knowledge about himself as something over and above his own sensations and ideas. Hume claimed that whenever he introspected to discover or experience himself, he always "stumbled on" some particular ideas or perceptions, and so he never found anything distinct from them. For Hume, then, there is initial certainty only about ideas and sensations. Any other knowledge, even about oneself, involves inference from ideas. And, as with all the other cases, some of the premises needed to link claims about presently perceived ideas to claims about the nature of perceivers are far from initially certain. If Hume is right, we are driven to a most radical skepticism, which leaves as knowledge only claims about certain ideas.
Skepticism, Probability and Induction
All the skeptical conclusions we have reached depend on two assumptions: that there must be ultimate, noninferentially certain, evidence for empirical knowledge and that knowledge is extended from this base only by extending certainty. This double requirement of certainty gained plausibility because it may well be that indubitaility must be added to justified true belief to obtain knowledge. Then, given the restriction of noninferential certainty to a Cartesian base, and the extension of certainty only by deductive inference, the skeptic's point is made. However, even if we grant the need for an indubitable base, the claim that all knowledge requires certainty seems too strong. It seems that knowledge is extended by induction as well as deduction, and induction does not extend certainty but only provides probability. We shall find, however, that an epistemological skeptic can succeed even if he allows inductive inferences from the indubitable base to what is justified as merely probable.
An inductive argument differs from a deductive argument in that, unlike a valid deductive argument, it is possible for its premises to be true and its conclusion false. Thus justification by means of induction guarantees neither certainty nor truth. Nevertheless, if inductive arguments can be said to confirm or support claims to a degree that makes them probable, then perhaps such arguments can be used to extend knowledge. In general, induction is used to make inferences from a number of individual observed cases to universal or statistical laws, or to individual unobserved cases. For example, from the claim that observed changes in temperatures of gases vary with changes in the volume and the pressure of the gas, we can inductively infer the perfect-gas law. This states that the temperature of a gas is proportional to the product of the pressure and volume of the gas. We can also infer the prediction that the next gas to be examined will have the same relationship between its temperature and its volume and pressure. And from the claim that 45% of an observed sample of American voters favor a certain candidate for president, we can infer that probably the candidate will receive about 45% of the actual vote. In each case the argument is from an observed sample of some group to a claim about the whole group or about some member of the group that has not been observed.
The question at this point is whether a person can start with his Cartesian base of indubitable knowledge and extend knowledge probabilistically by inductive inferences so that it includes claims about a present external object, about another person's beliefs, feelings, or sensations, about the past, or about the future. Again, the easiest case seems to be that of a presently existing external object, because we need only find something to make it probable that someone is perceiving such an object. We need some set of indubitable statements about observations, which can be used to infer inductively that someone is perceiving an external object. For example, can we inductively justify that someone is perceiving a cherry from the fact that he is having a visual experience of something small, round, red, smooth, and plump, and then a taste sensation of something tart? I might try to do this in my own case by using the following argument:
1. All (most) times when I have had sensations of something small, round, and so
forth are times when I have perceived a
2. I am having such sensations now.
3. I am perceiving a cherry now.
We can grant that the second premise is initially certain. However, the first premise is not certain because it is clearly possible that some, even most, of my previous experiences occurred while I dreamt or had hallucinations. Furthermore, if the first premise is supposed to be based on past observations, it requires remembering or correctly recording a significant number of past instances. As we have already seen, claims about the past are not initially certain. Thus knowledge of the present external world is not to be obtained in this way.
The consequences of this for our claims to know something of the past, the future, and other minds is also disastrous. Without knowledge of the present we cannot get knowledge of other minds or the past, and without knowledge of the present and past, there is no knowledge of the future. No claim relating my present sensations or beliefs to the past or to mental facts about others is initially certain. We might try thf premise that most times when I believe I remember something are times I correctly remember it, or the premise that most times I have a sensation of someone screaming in pain are times someone really is in pain. While it surely seems that both premises are reasonable, they are not initially nor noninferentially certain. Thus neither premise (nor, it seems, any other that would warrant the desired inductive inference) can be ultimate premises in an inductive argument to extend knowledge from a Cartesian base. Consequently, neither deduction alone nor the addition of induction provides a way to extend probabilistic knowledge, let alone certain knowledge, from the base. Epistemological skepticism is still secure.
Skepticism and the Initially Reasonable
In our attempt to avoid skeptical conclusions we have rejected the requirement that knowledge requires indubitability, but this relaxation is not enough. We have, however, been implicitly making three other assumptions that can be questioned: namely, that only the Cartesian base is noninferentially certain, that a noninferential basis of knowledge must be certain, and that at most deduction and the kinds of inductive inferences previously stated can provide inferential justification sufficient for knowledge. Perhaps we can justify rejecting some of these assumptions, and, if so, perhaps that will enable us to refute skepticism.
The first assumption has been rejected by those known as "common sense" philosophers, in particular, Thomas Reid (1710-1796), and, more recently, G. E. Moore (1873-1958). According to these philosophers, some statements about the external world are noninferentially certain and thus can be used to extend knowledge. Moore claimed that when someone confidently believes he sees his own hand in front of his face, it is initially certain that he sees a hand, even granting that it is possible he is mistaken. We have been equating initial epistemological certainty with indubitability and that in turn with there being no possible situation that would make the claim unreasonable. If Moore is right, this is too restrictive a conception of initial certainty. Indeed, it might be argued, the mere possibility of being unreasonable should not count against a claim being certain, if there actually is no other claim more reasonable to believe. If a claim is as reasonable as any other, then no more reasonable claim would be available to justify its rejection, and rejection would never be justified. Such certainty might seem to provide a sufficient basis for knowledge.
Unfortunately, for any claim someone makes that implies that some external object exists, it seems there is always one that is more reasonable, namely, that he believes the object exists. This statement is from the Cartesian base, and even with our liberalization of the conception of certainty we seem driven back to the core of indubitability. It might be objected that we should relax our restrictions on certainty even more, but instead of arguing that point, let us consider whether certainty is needed for a noninferential basis of knowledge.
Surely a claim that one's own hand exists is initially reasonable, even if not certain. Why can we not extend knowledge from such a base? This at least lets us escape to the present, perceived external world, and perhaps from there we can extend knowledge further. Of course, many skeptics would reject this move, but before doing so they should see how much their thesis would be affected. It is clear they would have to admit that there is knowledge about some presently perceived and perhaps very recently perceived external objects. However, they need not retract their views about unperceived presently existing objects, other minds, the past, and the future.
Consider first unperceived presently existing external objects. To show that the hand I just saw exists now when I do not perceive it, we need an inductive argument with a premise such as:
- Most objects exist at a time they are not perceived, if they have been perceived just before that time.
Similar premises are needed for knowledge of the past, such as:
- Most memories of past sunrises correspond to actual sunrises.
The skeptic can grant that if we could compile evidence for such premises from repeated particular observations in the present, then we could use the premises to extend knowledge to the unperceived present and past. However, no one can observe now that something previously perceived exists now when it is not perceived, or that something has happened in the past that corresponds to a present memory. Consequently, there is no way to use these premises to extend knowledge from initially reasonable present observations to the unobserved present or past.
We have seen that knowledge of the future depends on knowledge of the present and the past, and so without knowledge of the past there is no knowledge of the future. However, with knowledge of the past and induction we could get knowledge of the future. For example, if we knew that the sun has risen every day for centuries, then we could infer that, probably, the sun will rise tomorrow.
It might also seem that knowledge of other minds awaits only knowledge of the past, but a special problem arises. Even allowing induction and knowledge of the past, only the person who has a sensation or thought can observe that it accompanies certain behavior of his body. Thus no one can observe that most bodily writhings and screams have been accompanied by feelings of pain, because his observations of others are restricted to their behavior. The strongest claim he could verify by observation is that his own writhings and screamings have been accompanied by pain. But as a skeptic can point out, this is too weak a premise to justify inductively that unobserved pains of others accompany similar bodily behavior. In general, observations, even repeated observations, of only one entity, provide too weak a basis for inductive justification of a claim about other entities. In the case of other minds we can do no better. Thus, even granting knowledge of presently perceived external objects and the past, we cannot extend knowledge from the initial base to other minds.
Skepticism and Indirect Justification
Rejecting the first assumption of the Cartesian method and two additional assumptions has resulted in such a minimal extension of knowledge that an epistemological skeptic might well agree to their rejection. It is not clear, however, that he can safely give up the last assumption, namely, that at most deduction and the preceding forms^ of induction provide inferential justification sufficient for knowledge. Insofar as there is justification of scientific theories that explain what is observable by postulating unobservable entities, such as atoms, electrons, and neutrons, there is a third form of justification. The general statements of these theories are not deducible from either noninferentially certain or initially reasonable statements about particular observations. And they are not inductively derivable from such statements, because there is no way to observe the behavior of some of the postulated entities to justify the general statements inductively from particular observations. For example, the observable relationship between the temperature of a gas and its pressure and volume expressed in the perfect-gas law is explained by the atomic theory of gases. This theory postulates that a gas consists of atoms in motion that, as the gas temperature rises, hit the sides of whatever contains the gas with higher velocities and thereby increase the pressure on the container. This theory is not justified by observing the particular atoms in motion, but primarily by its explanatory and predictive power and its capacity to be tested experimentally by its observable consequences, such as increases in pressure. This is, then, a method of justification that differs considerably from the other two.
To defeat the skeptic we need justified postulations that have implications about unobserved presently existing external objects, other minds, the past, and the future. Scientific theories are not only models for this kind of justification, but they also provide the implications we want. The explanation of presently observed phenomena, such as clicks on a Geiger counter, are explained as the results of radioactive decay that has occurred over many years. This implies not only that certain events have occurred over a long period of time in the past, but also that unobservable events are happening now. Astronomical theories and laws that explain presently observed configurations of astronomical bodies, also imply, given these present configurations, claims about the future, such as precise predictions about eclipses.
The case for knowledge of other minds is made by turning to psychological theories where presently observed behavior is explained by means of unconscious beliefs or repressions, feelings of guilt or inferiority, desires for love or attention, and sensations of pain or pleasure. These and many other scientific claims about the past, future, and other minds are justified in this indirect postulational way.
It is clear how the skeptic would reply. He has been lenient in allowing us to include among our basic evidential statements those that are merely noninferentially reasonable and in allowing us to use induction to extend knowledge from this base. Here, however, his leniency ends. As we shall see, there is some doubt about whether inductive inference from observed instances provides justification that is strong enough for knowledge, but even granting that it does, the skeptic would argue that the indirect justification of scientific theories is clearly not strong enough for knowledge. Many of the fundamental statements of scientific theories are not generalizations inductively derived from particular observations. In the case of atomic theory, for example, no entities to which such statements apply are observable. With no observed cases on which to rely, verification is scanty at best. It is true that theories can be falsified by finding crucial predictions that fail, but to withstand such falsification as it explains and predicts does little to approach what is needed for knowledge.
Skeptical Doubts About Induction
It is difficult to answer the skeptic at this point, especially because the fourth condition of knowing remains undiscovered. Will it require certainty or at least justification so strong that only deductive inference can be used to extend knowledge? Some people working in this area find themselves inclined more and more in this direction, or at least toward a condition that would eliminate this third kind of justification. Some have even argued for a condition that would require such a high degree of probability for knowledge that in many cases induction would not provide an extension of knowledge. Furthermore, some skeptics would also claim that no form of induction should be used to extend knowledge, because, unlike deduction, no way has been found to justify its inferences. This is the view of Hume, whose skeptical conclusions about minds we have already seen.
Briefly, Hume's point is this. We cannot justify the accepted forms of inductive inference by showing that true premises in arguments of these forms always lead to true conclusions, because sometimes false conclusions result. For example, at one time there was an enormous amount of evidence that made it highly probable that all swans are white. When Australia was explored, however, it was found that some swans are black. The most we can hope to show is that these inductive forms lead from true premises to true conclusions most of the time. But how are we to establish this? We might grant that these forms have been mostly successful in the past, but even if we allow that as fully justified, we cannot deduce that they will continue to be successful most of the time. Such a conclusion involves the future as well as the past, and we have seen that no statement solely about the future is deducible from one solely about the past. We could add an additional premise that the future will resemble the past in relevant respects, but that premise is not initially reasonable and must also be justified in some way. However, it does not seem deducible from initially reasonable premises, and we cannot justify it inductively from past observations unless induction is justified independently. Thus unless some other form of justification succeeds, the problem of justifying induction first clearly enunciated by Hume, like the problem of defining knowledge, which dates back to Plato, remains unsolved.
The consequences of these two unsolved problems for our present concern are crucial. Without a justification of induction or some alternative and without a clear understanding of the fourth condition for knowing, the epistemological skeptic remains both unrefuted and with many arguments to bolster his views. If knowledge requires certainty or at least a high degree of probability, and there is no way to justify some nondeductive form of inference, then the epistemological skeptic will be able to show that his thesis is more reasonable than that of his opponents. As a consequence, our usual convictions about the external world, the past, the future, and other minds will turn out to be not knowledge but merely strongly felt beliefs.