Science

In the previous chapter, we discussed the criticism of skepticism as a view that blocks the endorsement of any view. We parried this attack by arguing that it rests on the false assumption that views are chosen by mere decisions, whereas beliefs are largely given.

The questions are important and for the following three reasons:

ι. Understanding the psychology of belief may help resolve dis­putes rationally.

2. Evolutionary considerations suggest that humans possess built- in biological mechanisms for the formation, maintenance, and modification of beliefs, mechanisms that are sufficiently effi­cient to permit the survival of sufficiently many societies that are sufficiently diverse. (Humans survive in India despite beliefs common in that country that prevent restricting the damage that cows and monkeys cause; treating tigers the same way would more significantly threaten their survival.) An evolutionary the­ory about this brings about automata emulating natural adapt­ability in order to learn and adjust to hostile environments.

3. Apsychological theory of the change of belief may help efforts to escape some limitation on belief, facilitate research, and make a difference in everyday activities, mainly between dogmatic and flexible patterns of thought and conduct.

Centuries of experience suggest that preaching is useless at best because it can do no more than expose the disadvantages of dogma­tism, which is rampant despite general familiarity. The recent theory of learning from experience is this: not all experience teaches but only those that happen during imprinting or when adrenaline flows excessively. This theory suggests that all animals are dogmatic and that learning by modification of imprinted information is unusual. This is so because dogmatism has a survival value, a virtue that we must acknowledge and see how it can be retained while dogmatism is eschewed. The chief advantages of dogmatism are that it is a mode of retention of possibly vital information and that following it ushers decisive action.^[28] The question, then, is: What alternative means are there for efficient retention and decisive action? The understanding of beliefs and how they operate may help develop new ideas about them, and training individuals accordingly will be the least costly and dangerous cure for dogmatism.

Underwhat conditions are erroneous beliefs acquired and sustained? Daniel Kahneman and Amos Tversky, to mention one famous exam­ple, described certain patterns of probability-assessment formation that are systematically and repeatedly erroneous.

The default popular view about beliefs is this: right beliefs are jus­tifiable; people should believe all and only what they find justifiable. This view suffers from three shortcomings:

Sharon Moalem andJonathan Prince, SurvivaloftheSickesfcA MedicalMaverickDiscovers Why We Need Disease.) Nevertheless, such avoidance does occur; therefore, learning, even if much too narrow, at times aids survival. Learning linked to survival options often occurs in significant situations, either by imprinting or in life-threatening situa­tions, wherein the neurochemistry of strong emotion imprints on long-term memory. In humans, we call the outcome of this kind of learning obsession. Under Freud’s influence, due to his faith in error avoidance, we misread all obsessions as neurotic (perhaps to the exclusion of taboos). There is an important truth in Freud’s confu­sion: humans are able to correct their errors; when they fail to do so, we suspect some obstacles on their way to improvement, usually fear. In accord with this, he observed, the intellectual ability to rectify error together with the affective inability to do so comprise a symptom of neurosis. He suggested that, invariably, the fear is rooted in the excitement that associates the acquisition of behavior patterns as obsessive. We suggest that this needs reinforcement - specifically, the fear of ridicule. It is possible to mitigate this fear by improving social norms by raising public awareness of the impossibility of avoiding all error. Whenever possible, the application of skepticism is helpful for neuroses because neurotics do behave dogmatically, as Freud noted.

ι. It rests on the presupposition that people choose the objects of their beliefs, at least if they are intelligent or well educated or rational. But this runs contrary to the common observation that beliefs are given.

2.      It rests on the presupposition that statements are justifiable, which is contrary to its well-known (skeptical) refutations.

3.      lang=EN-US>It rests on the presupposition that such justification generates unanimity among rational people. This runs contrary to the plain observation that disagreement is ubiquitous, especially among intelligent, well-educated, rational people.

We call attention to the following. Our skepticism does not deprive us of the use of the concept of belief the way the learned literature uses it. Although greatly varied, this literature is regularly in agree­ment about one important observation regarding belief: the discovery of a refutation of a statement taken to be true is a surprise.

The Critique of Empiricism

Empiricist philosophers demand belief in all and only those statements that are rational in which to believe. They assert that statements merit­ing rational belief are those that are reliable, credible, or believable. They are statements that logic or experience supports - directly or indirectly (namely, as corollaries from such statements). Empiricists deem unproblematic support from logic. Because they admit that observations are problematic, they limit the reliability or credibility of observation reports to direct perceptions, to observations that reflect directly perceived reality in order to ensure that they are infallible.^[30] The rules for inference that are used to yield reliable or credible con­clusions are those of logic, deductive or inductive. To deem deductive logic as unproblematic is insufficient, empiricists reluctantly admit, because they want an inductive logic that would permit the deduction of theories from observation reports. Remember that Hume refuted the claim that this is at all logically possible.

Their efforts are doomed to frustration, we say. We now return to the paradoxes of induction, not as refutations of empiricist epistemol­ogy but rather as refutations of the empiricist psychology, the empiri­cist theory concerning the formation and the fixation of beliefs, the view that rational people believe statements inductively deduced from direct observations.

The first paradox, we may recall, was Goodman’s. Defining the word grue to mean that an object is grue if and only if it is green until 2100 and blue thereafter, the two statements, “all emeralds are green” and “all emeralds are grue,” are equally empiricallyjustified, although the first seems true and the second absurd.

We argued that Goodman’s paradox refutes the empiricist epis­temological view that statements are inductively justifiable; we now argue that it refutes the psychological empiricist view of belief. For all we know, all observed emeralds are green, so we do admit that. The inductive rules of procedure, whatever they may be, cannot determine a rational preference for the conclusion that all emeralds are green over the conclusion that they are grue. But, actually, we disbelieve the second conclusion. Empiricism allegedly explains the rational choice of beliefs, but as for the expectation of green over grue, it obviously does not.

We use the same strategy for the other famous paradox of induc­tion, Hempel’s paradox, which is that the empiricist theory implies the possibility of raising the conviction that all ravens are black by exhibiting both black ravens and non-black non-ravens, such as white shoes. Hence, Hempel refuted the empiricist’s psychological theory of rational belief.

We present additional refutations later. First, we should introduce the quantitative theory that replaces the instantiation theory of empir­ical support: the probability theory of rational beliefs.

The Probability Theory of Rational Beliefs

The probability theory of rational beliefs states that every theory is credible to some degree, a degree that the calculus of probability helps determine in light of the total available empirical evidence. Thus, changing the stock of empirical evidence can alter the probability of hypotheses, reducing the credibility of one and raising that of another. This theory seems to explain how scientific revolutions occur, that science may relinquish faith in phlogiston in favor of oxygen, faith in Newton’s gravitational force in favor of Einstein’s curved space. Thus, counter-examples undermine a hypothesis thereby reducing its probability, whereas consistent examples support it thereby raising its probability. (In truth, we often believe people reputed as experts even though most of us have hardly glanced at the evidence; however, for now, we will let this ride.)

The calculus of probability yields with ease the following theo­rem: evidence that undermines a theory reduces its probability to zero. Empiricists ignore this: they are interested not in probability but rather in the rescue of induction. The calculus of probability assigns to any two statements a number between zero and one: the proba­bility of the one (the hypothesis) given the other (the evidence) is a given fraction. What is given and how it is given is not the business of the calculus. It does not allow the computation of the probabil­ity of one assertion given another if the status of the evidence as given is tampered with. Therefore, empiricists have little choice but to demand that we do not consider as evidence any statement except an observation report that we deem certain. Otherwise, the calculus of probability does not apply. Such observation reports do not exist, of course.

The probability theory of rational beliefs is refuted by the following simple argument of Popper. Consider the brief time (between 1915 and 1919) when Einstein’s theory of gravity was already known but Newton’s theory of gravity was not yet refuted. According to the prob­ability theory of rational beliefs, because at that time a lot of evidence supported each theory and none undermined either, each had a high probability, credibility, or rational degree of belief - that is, higher than 50 percent. (A theory the probability of which is less than 50 per­cent is less probable than its negation.) Now, consider the probability of the view that one of the theories is true. According to the calculus of probability, it is the sum of their probabilities. Hence, it is more than 100 percent. This is absurd.

The probability theory of rational beliefs is also open to criticism that rests on Hempel’s paradox. Allegedly, the probability theory solves it. The frequency of non-black non-ravens is much higher than that of black ravens. The support due to an improbable observation is higher than that due to a probable one. Therefore, the support of observing black ravens in the theory “all ravens are black” is much higher than that of observing non-black non-ravens. But this does not eliminate the paradox because it still allows for the intolerable idea that all evidence is relevant to any hypothesis. How much irrelevant evidence does it render relevant? How many white shoes support “all ravens are black” as much as one black raven? Obviously, this is somewhat amusing; the irrelevance of some evidence for some hypothesis is vital for so many reasons.

The probability theory of rational belief is also open to criticism that rests on Goodman’s paradox (in regard to the theory, “all emer­alds are grue”). The considerations are the same as those stated in the previous paragraph, except that the relief that the probability theory allegedly offers here is from the claim that the new alterna­tive that Goodman conjured is very improbable. Again, we may ask: How much do we have to boost the empirical support of the new hypothesis to compensate for this defect? Also, because the disjunction between many alternative Goodman predicates has an ever-increasing probability, this alone suffices to cancel any relief that the solution from probability may conceivably offer.

To add insult to injury, not only do optical illusions refute, even silly conjurer’s tricks do as well. Consider the theory, “Rabbits cannotjump out of empty hats.” Conjurers produce observations that refute it. We simply disbelieve the observation, not the theory - for the best reasons available or for no reason at all, as you like it. (The enjoyment of con­jurers’ tricks rests on the titillating seeming credibility of what runs against our deepest convictions.) The reasons, as it happens, refute the probability theory of rational beliefs, all protest to the contrary notwithstanding, because this theory leaves no room for false observa­tions. Sir Francis Bacon knew this: first, he limited the set of scientific observations to those attested to (but, to no avail, for the conjurers stand at the ready); and, second, said Bacon, the dilution of false observations will render them harmless. The scientific tradition soon limited the set of scientific observations to those that are repeatable. This renders scientifically irrelevant the justification of generalizations about unique observations from the start. Moreover, because conjur­ers’ tricks are repeatable, they are scientific. Their explanation is in the same style as the explanation of optical illusions.

The obvious response to our criticism is to flout this explanation or any other reasonable theory that exposes fake evidence as likewise fake. But, then, this theory may unwittingly allow for arbitrary immu­nization of any theory to all criticism by simply allowing the view of all refutations as fake. As long as this option is even faintly permissible, it renders error avoidance impossible. If, however, there is a guarantee that the proscription of fake evidence is not open to this abuse, then this is a guarantee that all genuine evidence is true. Such a guarantee is a theory of evidence. Does this theory rest on evidence? And is this evidence genuine or fake? It is all too easy to marshal fake evidence to support it.

Indeed, our argument itself smacks of a Sophist conjurer’s trickery, and perhaps it is. Should we proscribe it? On what grounds? If no grounds are available, then the theory will disallow this proscription, no matter how sane the proscription may be. Ifit has grounds, then the question reappears: Is that ground solid or are we victims of another (higher level?) conjurer’s trick? This, then, is a first step in an infinite Hume-style regression. Of course, in daily life, people rightly ignore such deliberations because they do not seek a promise that they will be infallible. Rather, they only do their best to behave as intelligently as they can and then they hope for the best. However, it never occurs to them, not even in their wildest dreams, that they can avoid all error and all optical illusions. Even the Pope’s infallibility, in which all Catholics must believe, is not purported to make him infallible except when he speaks ex cathedra (and then he is - or should be - careful not to make any refutable assertions). Why are everyday guarantees not enough for philosophy? Because philosophy has made a profound observation, that we prefer our own culture to all other cultures, and that we do so uncritically and often to our great loss. This observa­tion leads to a salient question: Why prefer my culture to yours? But philosophy gave this question the wrong answer: because my culture is scientific and so it is infallible. Some philosophers or, rather, some anti-philosophers, especially Wittgenstein, tried to proscribe the ques­tion. But the profound observation does raise the question, and it is salient even though most people do not ask it - at least not in daily life - and most philosophers still cling to the wrong answer thereto. The observation was that our preference for our views of the world rests on the sheer accident of our birth. The question was: What is the rule for the choice of a better view? And the erroneous answer was: choose only the views that you can prove. But this is too stringent a demand. Suffice it to say that we agree to seek explanations of what we know and discuss them critically and try to improve them in light of the criticism that we manage to elicit. On the whole, we suggest a simple policy: try to demand as little as you can and see if it delivers the goods. When you are disappointed, you better ask: What else should we demand? How then can we strengthen our demands minimally? Remember, it is easy to demand the moon; it is wiser to demand only what is vital or come as close to it as we can. Tradition prefers the bigger demand to the smaller one, on the suggestion that it is safer. This suggestion is amply refuted. We have no criteria for all situations about which demand is the safest. And when the bigger demand deliv­ers observed harm and only promised benefits, said Popper (The Open Society and Its Enemies), those who suffer the harm in question deserve the benefit of doubt.

The Minimum Unexpectedness Hypothesis on Beliefs

We finally turn now to our answer to the psychological question: Under what conditions does one believe that a certain statement is true?

To reiterate, it is generally agreed that a sense of surprise accom­panies the shattering of a belief: the discovery of a refutation of a statement believed to be true is a surprise. Let us see whether we can make do with this minimal supposition. Of course, we would like to assume that people seek the truth, but this is excessive, especially because, regrettably, many people are willing to settle for much less. So let us take one psychological characteristic of the possession of a familiar truth: the absence of surprise at its reappearance. This is not all, especially because it is not sufficient to characterize the relentless search for the truth that is such a characteristic of science. Moreover, even without science, quite possibly life with no surprise is intolera­bly boring (we return to this issue in Chapter 6). Nevertheless, let us examine this idea to see how far it can take us.

To summarize, then, our tentative (and possibly refuted) hypoth­esis is this: the function of ordinary belief (to exclude momentarily religious belief and its like) is to reduce surprise. Rightly or not, people tend to believe in a given statement if it reduces the unexpectedness of the apparent world, if it makes us expect our experiences as we live them:

1.      People assign probabilities to events in some vague sense of probability, where the probable is the expected or the unsur­prising.

2.      Surprise or unexpectedness or improbability of events is relative to given beliefs. Events surprise if they are experienced more frequently than extant beliefs warrant, especially if those beliefs rule them out.

3.      When people consider a new idea, they also consider the ratio­nale for its being proposed in the first place, and if that rationale depends on some events, then people also consider the question of whether these events are to be expected.^[31]

Let us expand on this a little further. The improbable is surprising. Let us assume, then, that in some loose sense, surprise and improba­bility increase or decrease together. And, to repeat, what we consider probable and not probable depends on the theories that we entertain. Thus, to take the easiest - although not the best - example, assuming a coin to be fair leads to an obvious expectation: flipped repeatedly, it will come up heads equally as often as tails - more or less, that is (because small deviations are repeated experiences). Great deviations from this or systematic results are surprising. A theory that the coin is biased in a way that fits the observed systematic deviation from the expected result alters the expectation and reduces the level of expe­rienced surprise. To take an extreme example, the observation of a series of heads only leads to the suspicion that both sides of the flipped coin are heads.

To reiterate, people believe a new idea if it raises the expectation of given experiences that are otherwise surprising. For example, the surprise that certain events regularly occur is reduced by postulating (spuriously or seriously, as the case may be) regularities that govern them. Thus, the supposition that all ravens are black renders quite expected the unexpected observations and initially surprising obser­vation that ravens are regularly black (because without the theory, all colors may be expected). It is the theory that alters probabilities - quite in accord with the calculus of probability - and with them, the theory alters surprise values: in this light, the observation of a non­black raven would surprise and the observation of black ravens would not.

Now, note the following:

1.      This hypothesis presents a sufficient condition for belief but not a necessary one. We say that explaining surprising events tends to create a belief; however, some beliefs are obviously due to different factors, often also of quite different sorts.

2.      Like the probability theory of rational beliefs, this hypothesis assigns some probabilities to events, and both theories state that an explanation of an unexpected event raises our tendency to endorse the explanation. However, contrary to the probability theory of rational beliefs, this hypothesis does not assign prob­abilities to explanations; rather, it refers only to probabilities of events. As mentioned previously, the idea that evidence can raise probabilities of theories is a version of the theory of induc­tion and, hence, a refuted one.

3.      The hypothesis is easily extendable to explain increases and decreases of degrees of belief or, rather, degrees of confidence: lowering the estimated probabilities of observations raises the degree of belief in their explanations. This idea is attributed to Ronald Fisher, who said that a hypothesis is likely in light of given evidence to the extent that it raises the probability of that evidence. That is, the likelihood of a given hypothesis in light of some evidence is the probability of that evidence relative to the probability of that evidence without that hypothesis.^[32] For example, the greater the number of observable ravens, the lower the prior probability that all of them will turn up black. Therefore, the more observed ravens turn out to be black and none turn out to be non-black, the higher the confidence in the theory that all ravens are black because it makes this observation much more probable. Nevertheless, to repeat, this degree of belief is not a probability measure. Moreover, any correlation remains at best qualitative, not quantitative. There is no way and no need to render it quantitative.

We now apply the hypothesis and resolve the empirical versions of the paradoxes of empiricism. We resolve them regarding only psychol­ogy, ignoring epistemology, because we do not demand infallibility of accepted belief.

Consider the question that Goodman’s paradox raises: Why does the theory that emeralds are grue have no adherent although so many philosophers know about it? Following the minimum unexpected­ness theory, the question can be answered as follows. The statement, “all emeralds are grue,” is arbitrarily constructed; one cannot escape notice of its arbitrariness: the moment in time for an emerald to change and the change itself that it has to undergo to qualify as grue are intentionally determined quite arbitrarily. The record of arbitrary theories is not good. The theory, “all emeralds are grue,” thus does not reduce unexpectedness but rather increases it via the implication that the strange event following from a theory invented arbitrarily will turn out to be true.

To illustrate the dependence of our incredulity toward arbitrari­ness, consider a case in which the date of conversion of the color of emeralds is not selected arbitrarily. Imagine the idea that all emeralds change color in time, so that all emeralds are green until they age, when they turn blue. Imagine further that we have concluded this idea from our theory of the molecular structure of solids. Imagine, finally, that it so happens that all emeralds we can lay our hands on just now are young. There is nothing paradoxical about these assumptions. (This would be intuitively more acceptable if we speak about one of the synthetic gems that have recently entered the market rather than about emeralds.) We would then find nothing paradoxical about the need to consider both hypotheses: that our emeralds will remain green and that they will become blue when they age according to certain specifications. Inductivists may take this up and say that our claim that some assumptions are not arbitrary is our admission that some theories are inductively supported. This is not so. First, the sup­positions that we discuss may be false. Thus, suppose that as it happens, there are different kinds of emeralds, and our experience is limited to one kind. Or perhaps the process of gem pigmentation decay accel­erates or decelerates during some later phase, due to some hitherto unobserved specific phenomena. Broadening our horizons will make us realize our mistake. Inductivists say that this is why we should not jump to conclusions. Thus, when an inductivist theory is refuted, its adherents are quick to turn it into a demand. The hypothesis that we discuss here is descriptive, be it true or false. As to the distinction it makes between the arbitrary and the non-arbitrary, again, induc- tivists contrast the arbitrary with the justified; this hypothesis takes any reason that people admit as good enough because it is descriptive. Specifically, and to repeat, the hypothesis takes any explanation to be good enough if and to the extent that it reduces surprise. This is psychology. The methodological parallel to it is the demand to explain repeatable observations. This chapter is psychological, not methodological. Inductivists are mistaken when they ignore this: even if psychology could substantiate methodology, the two would remain distinct.

Consider, then, the empirical psychological variant of Hempel’s paradox. His instantiation theory of credibility made him declare all evidence relevant to any given hypothesis. Let us report our experi­ence: people - including philosophers - deem most available evidence as irrelevant to any given hypothesis. Our empirical variant of the para­dox takes care of this. The irrelevance of a randomly given observation to a given hypothesis is their logical independence of each other. Their conjunction is not surprising in the least: the one does not exclude the other. And, as long as it does not render the observation unexpected, there is surprise. As it happens, cases of non-black non-ravens are not considered unexpected.^{color=black face="Book Antiqua">[33]} Hence, white shoes do not convince us that all ravens are black.

Take note of this. Here, we are speaking of the psychological hypothesis of the change of beliefs and, by our tentative hypothe­sis, this change of beliefs is prompted by its ability to reduce surprise. If that reduction is achieved by a new explanation, then people find that explanation credible. Whether they then change their belief or suspend judgment for a while longer is a different matter altogether. They may be dogmatists who cannot change their mind and who post­pone decision until they forget. It is only if they act that their action is explicable. So, we claim that the hypothesis is limited to changes of opinion; it asserts that change is prompted by surprise; it does not assert that all surprise causes change of opinion.

So how does this psychological hypothesis apply to scientific re­search? This is still a complex matter. Inductivists identify the psycho­logical and methodological or epistemological theories of the change of belief and then they have to limit their discussion to rational belief (which renders their descriptive theory prescriptive, we remember), taking it for granted that any received scientific theory is credible. This runs contrary to the notorious truth that science has sufficient room for some conflicting theories, such as that of Newton and Einstein. To accommodate for that, inductivists do not view as rational the belief in any scientific theory but rather only in the received one. The question becomes, then: What makes a researcher believe or disbelieve the latest theory? This question rests on the assumption that when researchers declare a new idea scientific, they believe in it. This is contrary to much historical evidence that makes much sense: researchers may assume that a theory has sufficient merit to draw attention without consulting their intuition about how credible they find it. They often postpone deliberation as long as they seriously consider the theory and its merits, and they offer their initial assessments as recommendations to exam­ine the new theories for their merit and for their ability to withstand severe tests.^[34] Indeed, researchers proliferate all viable hypotheses they can think of simply for the sake of the process of elimination in the hope of creating points of departure that bring forth new and hitherto unimagined hypotheses.

Thus, although our hypothesis applies to dogmatists and resear­chers alike, it does not apply to scientific research for the simple reason that researchers do not trust their own intuition; indeed, most great discoveries surprised their discoverers.^[35]

Researchers, then, are seldom concerned with the credibility of the ideas that they investigate. Still, given that people do find irrelevant to a given hypothesis most of the information that they know, researchers into the psychology of the change of belief might ask: Does our hypoth­esis imply that any observation alters the credibility of any theory, even if only to a very small degree? No. The observation of an unsurprising event invites no additional theory to render it unsurprising; therefore, by that hypothesis, it plays no role in the formation of a new belief because it says that an increase in the credibility of a given theory is due to its successful reduction of unexpectedness.

Yet, our hypothesis does explain, for example, the limits of the applicability of our variant of Hempel’s paradox. The paradox does not apply if, instead of referring to common characteristics (e.g., not being a raven and not being black), the hypothesis in question refers to uncommon characteristics. Consider, for example, the theory, “All bodies that fall are heavy.” One might be convinced that this theory is true by observing light things that do not fall, such as airborne bal­loons. Common things are both heavy and tend to fall to the ground. The characteristics, “a body that does not fall” and “not heavy,” are uncommon; therefore, observations of light bodies that do not fall are unexpected. As a result, one may enlist them to convince any hypo­thetical individual who does not believe the theory that all bodies that fall are heavy. One may object to this example by asserting that being heavy is simply the disposition to fall, just as being light is sim­ply the absence of this disposition. Consider another example,^[36] the theory that “all the objects above the size of a molecule are colored” (where items ordinarily deemed transparent are also recognized as however subtly tinted). Contrary to our variant of Hempel’s paradox, it is possible to convince someone that this theory is true by referring to submolecular objects (e.g., an electron): they have no color. Once again, both characteristics - “being below the size of a molecule” and “not being colored” - are uncommon, at least to direct experience; therefore, examples of things that are below the size of a molecule and not colored do convince that all objects above the size of a molecule are colorful.

We hardly need to say that we are not surprised by the failure of empiricist philosophers of science to make sense of what people find credible and what they refuse to believe. They seek epistemological grounds for the preference of theories and identify the view of a theory as credible with lending credence thereto. This is a dual error. First, such grounds do not exist because all statements are doubtful. To this, they respond by observing that belief is repeatedly observed as people do believe this or that theory, and some are more reasonable about it than others. Indeed, there is interesting psychological knowledge about beliefs. People do find theories more or less convincing, more or less credible. But, here is their second error, and it is traditional as well as modern: it conflates epistemology and psychology so, with the defeat of empiricist epistemology, empiricist psychology is defeated as well. Here, we discuss only the alteration of credence and what people find more or less credible than before (namely, psychology). We do not trouble ourselves in this context about the question of the rationality or the validity of accepted or unaccepted beliefs. Indeed, our tentative hypothesis applies to magically and scientifically minded people, to open-minded people, and to dogmatists. We do admit that people have beliefs. And we declare that the strength of our hypothesis is that it not only explains the change of beliefs that even dogmatists undergo but also the trouble they have in their struggle to maintain their beliefs, in their effort to fend off criticism of their views, whether or not empirical. They usually do so by the use of ad hoc hypotheses, but that is simply because these are more readily available than testable hypotheses. And what is most amusing about dogmatists is that they do not like ad hoc hypotheses any more than their critics, but they often find it necessary to learn to live with them.

Again, the empiricist philosophers of science claim that irrational beliefs do not interest them, that they study rational belief. This ren­ders their views prescriptive rather than descriptive, we remember. To this, they may respond with the assertion that they consider scientific theories credible, so that their theory is descriptive. To reiterate, it is then refuted by observations of the attitudes of researchers toward theories.

More on the Paradoxes of Empiricism

There remains the question, then: What are the processes that govern the endorsement of the hypotheses in which we believe? There are a few candidates here, and we present them in the order in which we examine them: the absence of ad hoc adjustments, simplicity, Occam’s razor, predictability, and tradition. In addition to these topics, all sorts of combinations are possible, of course, which we presented in the previous chapter when discussing the refutation of the empiricist epis­temology. We now refer to these criteria as factors that partake in the psychological process of changing beliefs.

Theories Adjusted Ad Hoc

Let us begin then with ad hoc adjustments of theories. To reiter­ate, when a researcher continuously modifies a theory after finding a counter-example to it, the modifications that come most easily are those that exclude the counter-examples from its domain of appli­cability. The modified theory then encompasses increasingly fewer events and in ad hoc moves. These repeated corrections produce a series of theories with increasing ad hoc modifications. Credence in them diminishes, but rules of induction make no sense of this; to the contrary, it forces us to see nothing wrong with these moves and even to commend them because the known observations still support the modified theory in its ad hoc modifications no less than before and even more so. Inductivism thus supports most theories that fit infor­mation in as ad hoc a manner as possible - namely, those theories the specific corollaries of which are all predicatively confirmed - so that the cases to which they apply are already examined and harbor no surprise. The famous philosophers at the turn of the twentieth cen­tury, Ernst Mach and Pierre Duhem, understood all theories to mean exactly this. Duhem stated that when a theory is applied to a new case, the domain of applicability of that theory is extended only after the venture is a success. Look back, never forward.

Our hypothesis explains the incredulity of ad hoc as follows. Usu­ally, we do not believe in a theory that underwent too many ad hoc modifications because it involves the highly unexpected event that all exceptions to the initial theory have already been accidentally observed. The probability that this is so is very small. Thus, the theory that has been repeatedly modified ad hoc has not reduced the element of surprise.

Of course, dogmatists do believe in ad hoc theories - and not only dogmatists but also many adherents to Newton’s theory decades after Einstein won the consensus. They have to endorse ad hoc correc­tions to their views, no matter how reluctantly. Even scientific theo­ries often undergo ad hoc corrections. Our hypothesis explains this reluctance.^[37] Some people are dogmatic, some are credulous: the for­mer find it difficult to depart from some of their beliefs, paying for it with a picture of the world that is less orderly than that of the less dogmatic; the latter fall for theories more easily than the less credulous so that they too may believe in the less credible theories. Nevertheless, according to our tentative hypothesis, the function of the endorse­ment of belief even for the dogmatists and the credulous is the same as for all of us: the reduction of surprise. Yet, to explain the conduct of the dogmatists and the credulous, we need additional factors. The dogmatist finds it difficult to depart from past beliefs; the credulous is eager to fall for new ones. This is not new: Sir Francis Bacon, father of the modern theories of rational belief, did not try to explain beliefs but rather to advocate the disbelief in every nonscientific idea on the assumption of the almost total absence of scientific theory in his day. He explained this absence as the outcome of the scarcity of rational beliefs and found it natural to explain this scarcity by assuming the prevalence of dogmatism and credulity. He viewed these qualities the way we imitate them here; therefore, our hypothesis is descriptive and partial. Is it refutable? We return to this question later. Let us say now, however, that come what may, we should not view it as a principle of induction because it is a tentative hypothesis that invites tests - and we wish to render it testable, of course.

We may first extend our tentative hypothesis to add another empir­ical suggestion to it: researchers who cling to any theory after it was modified ad hoc do so because they hope to show that the exceptions are not accidental, that they are not due to accidental observations. Thus, our hypothesis goes very well with the familiar and obvious observation that it is difficult to refute a well-tested unrefuted theory. Hence, as David Bohm observed, when reputed researchers suggest ad hoc modifications to their theories, their peers extend the courtesy of waiting to see if they can render them less ad hoc before they decide to ignore them.

Simplicity and Occam's Razor

Our hypothesis explains the phenomenon as follows: the linear function is the least arbitrary.¹¹ Once we allow any function, we have too many candidates. Therefore, the assumption that by sheer luck we hit on the right one does not reduce but rather raises the level of unexpectedness.

A similar analysis applies to Occam’s razor, which is the demand to eliminate unnecessary entities. Our psychological hypothesis is that people do not tend to believe in theories that theyjudge as multiplying entities needlessly. The vagueness of all this is deadly for methodology but not necessarily for psychology because, in many situations, we have fairly durable ideas about what is and is not necessary, and then we judge the unnecessary as arbitrary.

Our hypothesis explains this psychological phenomenon in the same way that we explained simplicity: keeping stubbornly to defunct entities is, at best, a mere guess, and the assumption that such guesses are a reliable source of information increases rather than reduces the level of unexpectedness of experiences.^{[38] [39]}

In the name of Occam’s razor, all sorts of entities were eliminated that no one misses, but some thinkers have also used it in support of their demand to eliminate all matter (idealism), or all minds (mate­rialism), or God (atheism). All these moves were subject to intense philosophical debate - perhaps they still are. This is due to the absence of a criterion and even for a feel for what is and is not necessary, or even what the necessity was of which Occam spoke. Let us add that a debate rages over the very question: Do researchers endorse ad hoc hypotheses? Popper stated that they should not; Thomas Kuhn, Paul Feyerabend, and Imre Lakatos stated, as they so often did, that it is a mistake to tell them not to. We consider this dispute misguided: researchers do not trust their beliefs but rather try out diverse options, just as detectives may examine evidence even when they are convinced that the suspect is not guilty or that the evidence is planted. Finally, one can always let the beard grow for a while longer to see how it feels and only then consider shaving it.

Predictions and Explanations

Let us now turn to the difference between predictions and explana­tions. To reiterate, contrary to the standard version of empiricism, predictions that come true are more convincing than explanations of known events.

Our tentative hypothesis explains why. It is usually easier to cre­ate a new theory that successfully explains given events by offering generalizations than to predict new events correctly. Therefore, the event of devising a theory that provides a new true prediction is more unexpected than devising a theory that provides a new explanation of past events. Consider an explanatory theory that also provides testable predictions that turn out to be true. We do not expect a conjecture to be successful in predicting unexpected events. Hence, when a the­ory successfully predicts them, the assumption that the theory is true drastically reduces the level of unexpectedness.

Our tentative hypothesis explains more: when a researcher comes up with a surprisingly strong theory, we may examine its structure and see if its strength is due to some verbal trick or if it expresses some new intelligent idea or remarkable corollary; the credibility of the theory increases even before any prediction. This is important for cases in which predictions are difficult to come by: it takes some convincing to persuade a researcher to prepare a new experiment.^[40]

Also, in the unusual case in which two equally explanatory theories compete, we do demand them to compete by predictions to break the tie between them. This is a crucial experiment that makes no sense

to empiricists, all protests to the contrary notwithstanding.^[41] Popper proved impossible the traditional empiricist theory that deems ration­al belief as abiding by the calculus of probability. He suggested that if one insists on rational belief, one should consider the failure to refute a theory a better measure. For our part, we carefully avoid recom­mendations. We assert that often researchers do follow his sugges­tion and as an instance of the more general case of the reduction of unexpectedness.

Thus, we explain the change of mind that researchers may experi­ence as a group after certain spectacular and crucial experiments. This psychological phenomenon we explain by our psychological hypoth­esis that is more general because the psychology of researchers is not general, and even researchers do not all follow scientific criteria except in research. They then often rely on tradition.

Antiqua">Thus, the situation is rather unexpected. Popper’s critics are in error because they habitually and traditionally confuse psychology with methodology. Popper claimed that significant research is expli­cable as results of the application of his methodology. We add to this the observation that the psychology of research appears distinct from methodology, and we have managed to bring them closer together than Popper noted even though we used the psychology that runs par­allel or analogous to his methodology. But, to repeat, we do not wish to run this parallel or analogy too closely: psychology does differ from methodology, and then our parallel or analogy breaks down. To see that, we continue with our examination of the criterion that conflicts with this parallel or analogy: the use of tradition as a criterion for credibility. This too is not what we recommend: we recommend no belief, least of all in research.

Tradition

Tradition influences belief. This is a repeated, unrefuted observation. It is very much the expected situation. Normal education conveys traditional theories, leading to the expectation that they are true. Therefore, when certain observations are explained by two competing theories, only one of which is traditional, the tendency is to endorse the traditional one a priori.^[42]

Our tentative hypothesis, you may recall, is partial. It presents a suf­ficient condition for the endorsement of a belief, not a necessary one. This applies also to the role of tradition. Our tentative hypothesis does not refer to the difference between slow and fast learners. It also does not consider the prevalence of differences of opinion within the same community despite its members sharing the same fund of knowledge and more or less the same tradition. Moreover, some people do not find it possible to deviate from tradition and they may even be unable to envisage such a deviation. These phenomena are very interesting; they seek a psychological explanation, and we recommend trying to develop one. We do not do so here, however, because it is beyond the scope of this study.

Is our psychological hypothesis testable, then? Following what we said previously, one way to refute it is to find a case in which certain repeatable observations are explained by two competing theories, one of which is more in line with tradition than the other and, contrary to our hypothesis, the endorsed explanation is the non-traditional. This, however, is questionable because we can explain the anti-traditional preference as the eagerness to endorse radical tendencies, and we may explain these and other tendencies by different ideas. Still, this additional explanation may be testable and thus restore the testabil­ity of our initial hypothesis. This is the norm in the social sciences, where liberty with added parameters is open to further discussion on the condition or in the hope of success in rendering an untestable hypothesis testable.

Hence, all we can say now is that our hypothesis allows for tradition to play a role. This is not much, but it is more than empiricism allows because it rejects all traditions except the scientific tradition and it does so on the (erroneous) understanding that the scientific tradition is utterly radical - namely, that the scientific tradition is free of all views except the ones that are proven.

Applications

We turn now to practical applications of our hypothesis, as follows:

ι. One obvious application refers to a practical way for convincing people, which is not very appealing but which in some cases is necessary, such as the search for a partner in business or matri­mony or political exchanges of all sorts. For, in any case, to con­vince is to influence the endorsement of ideas, whether concern­ing the suitability of prospective partners or the practicality of a political program. Indeed, we contend, efforts to create con­viction in a certain idea lead to listing conceivable events for the idea in question to explain and to evaluating the degree of unexpectedness of each of them. Realizing this improves per­formance.

2.      This hypothesis is also applicable to software programs that issue classifications or predictions for new cases (e.g., computerized diagnosis).^[43] Some of these programs issue the classification by applying rules: a certain syndrome renders likely a certain diagnosis. Now, during a medical examination, we may face a situation in which different rules lead to different diagnoses and prognoses (predictions of sorts) in regard to the case under clas­sification. Following our hypothesis, emulating human psychol­ogy, such cases can find their resolution by calculating which prediction implies fewer unexpected events, although the exact method is beyond the scope of the current work.^[44]

3.      When building a model to explain given data, we should check the model’s ability to reduce the level of unexpectedness of events - namely, its ability to explain. This is not news. Also, a model that does not reduce the level of unexpectedness, because it entails no prediction, is no better than any alternative model. We naturally tend to endorse what seems to us closer to common sense. This is how tradition comes into play unless we are critically minded. But different traditions offer entirely dif­ferent ideas as to just what common sense is, even in regard to the place of sheer coincidence and outright notions of magic.

4.      Pursuant to building explanatory models, when seeking an ex­planation, we should consider the possibility of a fluke. Dar­winian evolutionism can serve as an example: prior to Darwin, leading naturalists considered the variety of species inexplica­ble or explicable, if at all, as expressions of divine intervention (which amounts to the same). The Darwinian theory (as well as the neo-Darwinian theory) offers a pattern of explanations as processes of accidental changes, only the fit of which sur­vive long enough to breed. Following this pattern, we may hope to render explicable a greater part of the variety of observed species. So, the theory reduces at once the level of unexpected­ness of this variety. Note that, as a scheme, Darwinism is hardly refutable. Still, many of us endorse it because it reduces the level of unexpectedness and squares biology with the naturalist view of the world that is at the base of the scientific venture, at least better than its earlier competitors.

5.      Finally, our hypothesis explains the generally expressed com­plaint that the social sciences are trivial. Many of the obser­vations that the theories in the social sciences explain are expected; as a result, these theories look trivial. There are excep­tions, of course, but those who complain do so under the overall impression that most theories of the social sciences fail to reveal unexpected discoveries. The practical instruction is, therefore, as follows: look for unexpected general observations, try to cre­ate a theory that explains them, and then try to test it by refer­ence to still fewer expected observations.

The last point is, of course, an echo of Popper’s theory of scien­tific method. He excluded psychology, especially any psychological hypothesis regarding the change of beliefs. Without failing to admire Popper’s tenacity, we reintroduced a psychological hypothesis regard­ing changes of beliefs - as a separate hypothesis to be tested and improved upon. Now, assuming that our hypothesis is true, do we like it? Do we wish our beliefs to behave as it describes? If not, what can we do about it? Moreover, can we change and improve the psychological mechanism of our beliefs? These are good questions, and we can offer only outlines of answers. We assume that no rational person willingly desires to believe falsehoods. And so, it is worthwhile to study the question: What is the mechanism for self-deception? Also, we assume that repeatable refutations impose the relinquishing of the refuted views. Inasmuch as we can control our psychology, we may then train ourselves to open up to all the alternatives that we can find, and do so critically. This is easier said than done: it often takes more courage than is readily available.