The Scientific Attitude

by Lee McIntyre

  Doesn’t this undermine the idea that a theory explains because it tells us why things happen? This is controversial. Some of the most famous scientific explanations could not say at the time they were offered what causal powers were responsible for the patterns they explained. (Another outstanding example here is Darwin’s theory of evolution by natural selection, which awaited Mendelian genetics before it could say why evolution occurred.)15

  This raises the question of whether scientific theories are just instruments for prediction. Whether they are a mere shorthand account of patterns in our experience that—given the limits of scientific theorizing—can never offer a definitive answer to the mechanism behind them. It is generally thought that this is not enough for scientific explanation—that scientific theories must try to say not just that something happened but why. The answers don’t have to be immediate, but a good theory should give some promise that answers will be available upon further empirical investigation. The importance of this is illustrated by what happens when we do not really have a theory at stake: when all we have are beautiful predictions, but no explanation for why the predictions are fulfilled.

  “Bode’s law” is one of the most dramatic illustrations in the history of science of just how far you can push things (and how quickly they can come crashing down) if you have good fit with the evidence—and even a few good predictions—but no theoretical support. In 1772, after studying the distance between the planets for what must have been a very long time, Johann Bode noticed a startling correlation. If one takes the common doubling series {0, 3, 6, 12, 24, 48, 96, 192, 384, 768}, adds 4 to each number, and then divides by 10, one gets {0.4, 0.7, 1.0, 1.6, 2.8, 5.2, 10.0, 19.6, 38.8, 77.2}, which is almost identical to the distance of the planets from the Sun measured in astronomical units (an astronomical unit is defined as the distance between the Earth and the Sun). In 1772, there were only six known planets, which had the following distance from the Sun: Mercury (0.387), Venus (0.723), Earth (1.0), Mars (1.524), Jupiter (5.203), and Saturn (9.539). At first, no one seemed bothered by the lack of an explanatory mechanism; hadn’t Newton after all famously “framed no hypotheses” about gravity? Still, there were questions. What about the “gap” at 2.8? And what about the rest of the series? These were taken to be “predictions” that would indicate where further planets might be. When the planet Uranus was discovered nine years later at 19.18 astronomical units, people were agog. Twenty years later, when scientists came to believe that there had once been a planet (which they posthumously named “Ceres”) between Mars and Jupiter, that had broken up and formed the asteroid belt at 2.77 astronomical units, Bode’s law was hailed as a momentous scientific achievement. Although it explained nothing (because it had virtually no scientific theory for its predictions to confirm), it was taken seriously because it had successfully predicted two new planets. When Neptune was discovered at 30.6 astronomical units in 1846, followed by Pluto at 39.4 astronomical units in 1930, things began to fall apart. Bode’s law was finally accepted as nothing more than a remarkable artifact of naive correlation.16

  Compare this with something like string theory, which in its most modest rendering is a theory of gravity at the microscopic level and at its most ambitious is a complete theory of everything in the universe.17 Entire books are devoted to the technical details of this fiendishly difficult subject, but the short story is this. Einstein’s general theory of relativity proposes to explain the largest things in the universe (stars and galaxies), while quantum mechanics proposes to explain the smallest (molecules and atoms).18 Both theories are incredibly well-supported by the empirical evidence, but there is a problem: they are fundamentally incompatible with one another. To put it bluntly, they cannot both be right. It may, however, be the case that while neither one is completely correct, both theories are special cases of some larger theory that subsumes and explains the phenomena covered by each of them. One candidate for such a theory—known as the Standard Model in physics—has done a good job of accounting for all of the fundamental forces in the universe but one: gravity. This has led physicists on an ambitious search for a quantum theory of gravity, of which string theory has been held to be the most promising (but not the only) candidate. But there is another problem: string theory has absolutely no empirical support to suggest that it is right.

  At this point, string theory is a mathematical model that many physicists hope is correct, for the simple reason that there are few alternatives. But this raises an important question: without empirical support, is string theory even science? Isn’t it “just a theory”?19 We here face a situation that is the direct opposite of the one just confronted with Bode’s law: instead of an explanation that has amazing fit with the data, but no theory behind it, we instead have an incredibly complex and fruitful theory with absolutely no empirical support. But doesn’t this violate our earlier criterion that a scientific theory must be vetted against some sort of evidence?

  Precisely this question was considered at an academic conference entitled “Why Trust a Theory? Reconsidering Scientific Methodology in Light of Modern Physics,” held at the Ludwig Maximilian University in Munich, Germany, in December 2015. This unusual conference brought together both physicists and philosophers to consider whether it was possible for there to be a new way of doing science. One such proposal was made by the physicist-turned-philosopher Richard Dawid, whose book String Theory and the Scientific Method was sanguine about the idea that, given the difficulty of gathering empirical support for string theory, we needed to turn to alternative modes of “non-empirical theory assessment” such as explanatory coherence, unification, fruitfulness, and even the aesthetic criteria of “elegance” or “beauty.”20 This was shot down by several of the scientists in attendance, who argued that even though string theory may not have any currently testable empirical consequences (because of the enormous practical limitations on building the apparatus to test them), it does make predictions that are in principle testable.21

  Not all physicists would agree with this. Some have even held that there are sociological factors at work such as “groupthink” and the pressures of tenure, career advancement, and grant money at stake, which—even absent empirical support—have made string theory “the only game in town.”22 To say such things, however, seems different from saying that string theory is not testable. If one reads these kinds of criticisms closely one finds careful phrasing that string theory “makes no predictions about physical phenomena at experimentally accessible energies” and that “at the moment string theory cannot be falsified by any conceivable experimental result.”23 But these are weasel words, born of scientists who are not used to taking seriously the distinction between saying that a theory is “currently” testable versus whether it is “in principle” testable. The practical limitations may be all but insurmountable, but philosophical distinctions like demarcation live in that difference.

  Perhaps the critics are right and string theory gets far too much attention given its current lack of empirical support. Almost certainly they are right that it would be absurd to “redefine science” in order to accommodate this one theory. Time will tell whether the paradigm of string theory has enough going in its favor to survive—as a practical matter—without the buttress of empirical support. But, on the question of whether it is science, I come down on the side of those who draw a distinction between saying that something is not testable “now” versus saying that it is not testable “in principle” (for instance, if it made no empirical predictions). Which means that—like so many other theories before it—string theory may be scientific, even if it later turns out to be wrong.

  What does all of this mean for the more general issue of the importance of theories for science? Having a theory may be essential for science, but is it enough? Doesn’t there have to be some evidence in support of a theory—or at least some possible evidence that could support it—before we can claim that it is scientific? If not, how will we ever be able to say why nature works as it
does, which seems necessary for scientific explanation? As we saw with Ptolemy and Newton, a theory can be scientific and false, but it must seek to fit with the evidence and make some attempt to explain it. We do not need to take the drastic step of rejecting every false theory as unscientific. What makes a theory scientific isn’t that it is true, it is that it says something—even if it is just a promissory note to be cashed in the future—about whether there is a mechanism behind the theory that supports its predictions and is consonant with empirical evidence down the line. Are we there yet with string theory? Will we ever be? Scholars disagree. But we may nonetheless appreciate that without a theory, we would not even be having this conversation. This is why Bode’s law failed and string theory may yet succeed.

  Why then do some critics complain that everything in science is “just a theory”? Is it because they believe that every scientific theory is as controversial as string theory? Or is it that they just don’t understand (or don’t want to understand) how powerful it is to say that we have a good scientific theory?

  The theory of evolution by natural selection is “just a theory,” but it is instantiated in virtually everything we believe in microbiology and molecular biology, from the cellular level up to species. It has been rigorously tested for a hundred and fifty years. It accounts for the data, makes predictions, and is completely unified with Mendelian genetics, which is the mechanism behind it. Evolutionary theory is the absolute backbone of scientific explanation in biology. Indeed, the famous evolutionary biologist Theodosius Dobzhansky was taken by many to speak for the profession when he said: “Nothing in biology makes sense except in the light of evolution.”24 The alleged holes in Darwin’s theory of evolution by natural selection are in some cases nothing more than misunderstandings by the layperson of what biology is about. And any actual holes are nothing more than the sorts of research problems that one would expect scientists to be working on in such a mature science. As Kuhn taught us, in any open-ended enterprise you will never have explained everything. One must keep pushing forward.25 But make no mistake. Scientists have accounted for the complexity of the eye. They have even found a candidate for the “missing link.”26 The sort of nonsense one sees put forth by creationists in an attempt to discredit evolution are not the sorts of criticisms that scientists make to one another; they are the stuff of ideologues and conspiracy theorists.27

  Gravity too is “just a theory.” So is the germ theory of disease. And the heliocentric solar system. Indeed, as we have seen, everything in science is “just a theory.” But this does not mean that we have no reason for belief. Having a good theory is the foundation for science. We don’t need deductive certainty for a theory to be scientific or for it to be believable. The notion here is a delicate one but nonetheless important: we are entitled to believe a theory based on the evidence that supports it, while knowing full well that any future evidence may force us to give up our beliefs and change to another theory. In science we must simply do the best we can with a rigorous analysis of the data that we have available. Our beliefs can be justified, even if we cannot (and should not) quite bring ourselves to maintain that they are true.

  Thus in some sense the critics are right. Science cannot prove anything. And everything that science proposes is just a theory. When we are at the mercy of future data, this is the situation that all empirical reasoning must contend with. And, unfortunately, this is the basis on which some members of the public—most particularly the ideological critics of science—have misunderstood how science works. It is true that science must contend with the open-ended nature of empirical reasoning, yet it is also rigorous, meticulous, and our best hope for gaining knowledge about the empirical world. But how can this be?

  The Role of Warrant

  It is now time to introduce the concept of warrant into the debate about whether we are justified in believing empirical theories. Even though they are not provably true, certain, or (even in principle) more likely to be true, there is a sense in which we would be foolish to ignore the idea that scientific theories are believable precisely because they have positive empirical support.

  Here a crucial distinction must be drawn between truth and warrant. Even if a scientific theory is not technically, logically, more likely to be true once it has survived a number of rigorous tests, the question arises “aren’t we nonetheless justified in believing it?” And here I think one may plausibly say that the answer is yes. Despite the logical problems presented by induction, verification, and confirmation (and seeing how easily Popper’s practical concession of corroboration can slip into relying on positive instances), there is something deeply important about the success of empirical tests that scientists seem justified in wanting to hang on to. The credibility of a scientific theory does seem to increase once it has survived a number of rigorous tests. Indeed, even most philosophers of science—who are very familiar with all of the problems of inductive logic—understand that it would be rash to claim that just because positive evidence cannot be used to prove that a theory is true, this means that the theory is not believable.28

  There is a subtle distinction between saying that a theory is true and saying that we are justified in believing it. The idea goes back to Socrates. Perhaps we may never reach the truth in our lifetimes, but can we make progress toward it? Can we at least eliminate false claims to knowledge? To say that a theory has warrant is to say that it has a credible claim to believability; that it is justifiable given the evidence. This is to say that even if a theory later turns out to be mistaken—like Newton’s theory of gravitation—one may still maintain that given the evidence at the time, scientists were rational to believe it. Why is this important? Precisely because—given the way that science works—one expects that in the long run virtually all of our empirical theories will turn out to be false.29 But this does not mean that we are unscientific for believing them, or that it would be better to withhold all belief until the “rest of the evidence” is in. Indeed, given how science operates, the rest of the evidence will never be in!

  The doctrine of fallibilism accepts that we can never be certain of any empirical belief, yet maintains that it is unreasonable to think that all knowledge requires certainty.30 Yes, the problem of induction undermines both certainty and probability, but what sort of epistemological stance is appropriate in the face of this? Should we give up all nondeductive beliefs? Even where the empirical evidence is strong, should we refuse to say that we know anything? This seems absurd. The fallibilist accepts that—outside deductive logic and mathematics—we can never achieve certainty. But this does not mean that we must foreswear all claims to knowledge. Not everything that is true is necessarily true.31 And at least some of these nonnecessary truths are surely worth pursuing. Instead of giving up on vast tracts of possible knowledge, perhaps we should enlarge the notion of knowing so that it includes the idea that we can have justified empirical beliefs, even if we also understand that some of these beliefs may later turn out to be false. The doctrine of fallibilism, therefore, is as much an attitude as a set of principles: it tells us that it is all right to feel comfortable with the idea that some of our beliefs are justified based on their fit with the evidence, even if in the long run they may turn out to be mistaken.

  Of course, it is important not to be gullible or overambitious. We must avoid thinking that just because we have good current evidence for an empirical belief, it is probably true; that although we may not know that we know something, the strength of the evidence suggests that the underlying theory must be pretty close to reality.32 Yet, in compensation for this epistemological equanimity, we are saved from having to retreat into a sterile skepticism, where we take the problem of induction so seriously that we can believe in nothing, because it might someday be overthrown by better evidence. The idea of warrant can be squared with reliance on empirical evidence, even if fallibilism also requires that no accumulation of evidence will ever amount to certainty.

  We cannot hold reasoned belief hostag
e to certainty. Despite Popper’s and others’ best efforts, the process of scientific reasoning is just never going to be deductively valid. Scientists are right to want to rely on the encouragement they get from positive instances, as long as they do not go overboard and assert that their theory is true, or indulge in the temptation to overlook negative evidence. But what about the problem of induction? I may not be the first philosopher to think this, but I will take the risk of saying here what I have heard many philosophers say in private: to hell with the problem of induction. The problem of induction was never meant to substitute mental paralysis for justified belief. Even David Hume seemed to recognize that in some sense induction is wired into how humans reason:

  Most fortunately it happens, that since reason is incapable of dispelling these clouds, nature herself suffices to that purpose, and cures me of this philosophical melancholy and delirium, either by relaxing this bent of mind, or by some avocation, and lively impression of my senses, which obliterate all these chimeras. I dine, I play a game of backgammon, I converse, and am merry with my friends; and when after three or four hours’ amusement, I would return to these speculations, they appear so cold, and strained, and ridiculous, that I cannot find in my heart to enter into them any farther.33

  Both our brains and our instinct tell us that positive instances count. As human beings, we cannot help ourselves in relying on induction to learn about the empirical world.

  Yet might this be appropriate after all? One of the most fascinating responses to the problem of induction was given by Hans Reichenbach, who made the case that even if induction could not be logically verified, it could at least be vindicated.34 His argument is intriguing. Suppose that the world is disordered: that there are no empirical correlations between anything. In this situation, no method would be able to account for or explain the regularities in our experience, because there would be none. Consider, on the other hand, a world more like our own, in which there were correlations in our experience and we wished to employ a method that would pick up on them. What method should we use? Arguably, the best choice would be induction. While it is possible that other methods might work, none would work better than induction. Like science itself, induction is responsive to patterns in our experience, flexible enough to change its conclusions based on new evidence, and capable of abandoning any individual hypothesis and starting again if the data warrant it. Even though induction may sometimes lead us into error, so would any other method of reasoning. And, although some other method may work, it will never surpass induction. Reichenbach thus concludes that induction will do at least as well as any other method in discovering the regularities in the empirical world. Thus inductive reasoning is “pragmatically vindicated.”35