E Conjecture and Criticism (David Deutsch)
While it is true that Deutsch rejects the hypothesis, it is more accurate to say that his program subsumes the hypothesis. He thinks that nature is far weirder than most of us can conceive.
For him, our universe is only the faintest wisp in an indescribably intricate structure called the multiverse. The multiverse is something like a network of all possible universes, except that network doesn't begin to capture its complexity. I'll return to the multiverse a bit later, but to get a crude feel for it, imagine that you set off in a car from Times Square in New York City headed for Los Angeles. You soon come to an intersection where the road splits, going right and left. In lieu of picking one, you, your car, and all of its contents are duplicated, and one copy of you goes right and the other left, into separate universes, each copy continuing along blithely unaware of the other. And this keeps happening. At every intersection, each copy of you and your car splits, budding off new copies in a fantastic amoeboid way. Imagine the unbelievable complexity by the time you get to the George Washington Bridge! To St. Louis! Yet the roads aren't becoming clogged by the trillions of your doppelgangers because, at every fork in the road your copies cruise off into different universes. The real multiverse is much more subtle and complicated than this, but you can see that Deutsch's thinking takes getting used to. I will skip most of his aggressive program for reorganizing our perceptions of science, philosophy, art, politics, etc., to home in on his vision of the roles of hypotheses, predictions, and testing in science. As you'll see, these concepts need serious revision or reevaluation if the multiverse is going to fit into science.In a nutshell, for Deutsch, scientific knowledge consists in explanations of natural phenomena that come from an iterative process of Conjecture and Criticism.
Now, since the procedure of falsification testing is a kind of formalized criticism, you might think that Deutsch's program is actually Karl Popper's Conjectures and Refutations in disguise. It's not. Deutsch has bigger ambitions than restating Popper for the twenty-first century; he wants to rescue science, beginning with theoretical physics, from what he sees as its impending identity crisis and set it on a new footing. Whereas Popper believes that the search for scientific Truth proceeds by proposing and testing empirical hypotheses, Deutsch wants to rewrite the rules and dispense with the hypothesis as we know it, and he adds a new element to the search for Truth. As empiricists, we are accustomed to thinking about Truth as representing agreement between our statements about the world and how the world really is. Deutsch plans to wean us away from our dependence on empiricism. He believes that good explanations can take its place.10.E.1 Good Explanations
Why do we have annual seasons? For hundreds of thousands of generations, humans made no progress in answering this or almost any significant question about nature. With a few transient exceptions (e.g., the “Golden Ages” of Greece and Florence), ignorance was the dominant state of affairs in Europe until the 1500s, when the explosion of knowledge called the Scientific Revolution overlapped the equally dramatic societal change known as the Enlightenment. Philosophers (remember this was before science emerged as a separate branch of knowledge) began to rebel against the authority of the Church and the iron legacy of the Greek philosophers which held that logical deductions from accepted facts, dogma, would lead to new knowledge. In the breakaway, philosophers began to accept observation and experiment as the way to settle questions of scientific truth—the doctrine of empiricism was born.
In David Deutsch's retelling, empiricism was a good start but, by itself, didn't enable us to make significant advances in understanding nature; we needed to develop a culture of criticism and to seek good explanations.
A good explanation is “hard to vary,” which means that every detail of the explanation is irreplaceable; if you change one detail, the explanation falls apart and no longer explains what it was supposed to. Bad explanations are “easy to vary,” and changing their details has little effect on their explanatory power.Deutsch's favorite example invokes the accounts of the annual earthly seasons given by ancient mythologies and the one given by modern astronomy. In the myths, some god might cause winter because she's sad and, as her emotions change, the other seasons follow. Because the details of the myths have nothing really to do with the seasons, mythological explanations like this are easy to vary. You can substitute one god for another, alter his or her circumstances, motivations, powers, objectives, etc., and generate equally good explanations for the seasons. Myths are bad explanations. Modern astrophysical theory attributes the seasons to the degree of tilt of the earth's axis of rotation with respect to its orbital plane around the sun. Winter happens in the hemisphere tilted away from the sun, while summer occurs in the other hemisphere. Change any detail of this explanation, say remove the tilt in the axis, and it no longer accounts for the seasons. The explanation is hard to vary. You might object that the modern astronomer has a lot more data to work with than the Greek mythologist did, and this is true. In fact, another way of thinking about the hard-to-vary criterion is that a good explanation is more closely connected to observational data, more tightly constrained by data, than are bad explanations.
The parable of the seasons illustrates the concept of hard to vary, but sheds no light on why Deutsch believes that science is outgrowing its dependence on empiricism and the hypothesis, which we'll take up next.
10.E.2 Rejection of Empiricism and Induction
Empiricists believe that we get our information about the world through sensory experience and that experience is the standard for assessing the validity of scientific truths.
Deutsch says that empiricism is a dead end, it is “the misconception that we ‘derive' all our knowledge from sensory experience.” According to his way of thinking, observations cannot provide the logically certain basis required for valid deduction; we can't reason from raw data to deeper truths about nature. Our sensory evidence that the sun comes up every 24 hours is plainly wrong: the sun does not rise; the earth rotates and makes it seem that the sun rises. We are continually deceived by what our senses tell us about the world, and, therefore, reasoning from naive experience inevitably generates error, not truth.What about induction? Inductivists claim that we get from observations to explanations through inductive reasoning, but Deutsch declares that inductivism is not merely wrong, it is another misconception: a nonexistent fictional process. He concurs with Hume's take-down of induction—(philosophical) induction is useless because its foundational assumption, the Uniformity of Nature principle, is unprovable—but Deutsch argues further that Hume let the inductivists off the hook too easily. A deeper problem is that induction is unavoidably grounded in empiricism; our sense impressions provide the purported regularities that inductivists reason from, and, since sensory experience is unreliable, inductive conclusions are not only logically weak, they lack any basis at all.
Most importantly, says Deutsch, we create the explanations that we seek; they do not arise spontaneously from observations. Instead, science makes good predictions because it develops good explanations, not because it has raw sensory data. Astrophysicists know a tremendous amount about stars and galaxies; they understand why stars are bright, and why, every so often, one of them blows up in a spectacular burst of energy called a supernova, but no one has direct experience of a star. Astrophysicists know about stars because they have good explanations for them and their properties.
Deutsch concurs with Popper in dismissing inductivists who do not admit that, before they can induce anything from phenomena, they first have to define the phenomena (i.e., they have to conjecture about what to count as phenomena in the first place).
Observations, measurements, experience, are all “theoryladen.” We “tacitly [rely] on explanatory theories to tell us which combinations of variables... we should interpret as being ‘repeated.’ ” Deutsch quotes the Greek philosopher Heraclitus to make the point: “No man steps into the same river twice, because it is not the same river, and he is not the same man.” You need theories of “river” and “man” before you can study rivers and men. Therefore, descriptions and categories do not precede theory: they follow theory (and may then, of course, lead to new theory).And induction comes up woefully short when science needs to make novel predictions. Our limited earth-bound experience of 24-hour sun cycles does not predict when an astronaut in a satellite orbiting the earth will witness sunrise; on the other hand, the astrophysical explanations of the properties of the solar system and of earth-orbiting satellites correctly predict that the astronaut will see a sunrise every 90 minutes.
A key concept for Deutsch is reach: the ability of a theory to account for observations other than those that it was constructed to explain. Reach is related to “non-obvious predictions (Chapter 1),” but for Deutsch reach is an expansive and fundamental concept. Reach figures prominently in his rejection of induction—induction goes from particulars to generalizations of the same kind; it is confined by such cases and cannot go beyond them. Reasoning from properties of apples might generalize to properties of oranges, but the reach of real scientific theories takes them beyond the obvious applications. Newton’s theory of gravity was constructed to explain the rotation of the moon around the earth. Its reach allows it to account for the paths of man-made, earth-orbiting satellites, as well as earthly ocean tides. Induction can’t do that.
10.E.3 Demotion of Testability, Prediction, and the Hypothesis
There is great overlap between Deutsch’s program and Popper’s Conjectures and Refutations, and Deutsch appears to be a sincere admirer of Popper.
Like Popper, Deutsch accepts that all our discoveries are ultimately uncertain and that it is unnecessary—in fact, it is a mistake—to try to justify our conjectures about the world. On the other hand, Popper believes that the testability of hypotheses is indispensable for making scientific progress, and Deutsch emphatically disagrees; after all, bad explanations may be testable. He points out that the mythological hypotheses for seasons, though scientifically nonsensical, were testable in principle. Consider. The myths tacitly predicted that all parts of the earth would experience the same season at the same time: winter in Greece would mean winter everywhere. The Greeks, however, had limited information about the shape and extent of the earth. Had they realized that it was spherical, they could have tested the universal season prediction by sailing south of the equator, at which point they would have discovered that their explanation was incorrect. Mythology could devise an endless series of testable stories without hitting on a good explanation. Deutsch concludes that testability alone is not a sufficiently rigorous standard for science.What about predictions? Deutsch's sees prediction as weaker than testability. The mythological framework made centuries of correct, but vacuous, predictions of seasonal sequences. Predictions are a dime a dozen; you make a prediction when you bet money that your favorite lottery number will be a winner. Lottery betting and the like are examples of “explanation-less” exercises; they are scientifically worthless because they are not part of attempts to explain nature. Many purportedly scientific predictions are little better. Instrumentalism, an offshoot of empiricism that emphasizes the predictive capabilities of science and regards successful prediction as a substitute for good explanation (as indeed QMB does; Section 10.B.) is an example of a failed approach. In sum, Deutsch thinks that testability and predictability are inadequate benchmarks for scientific progress.
In the same vein, Deutsch demotes the hypothesis to secondary status because hypothesis testing relies on empiricism, and, since empiricism for him is passe, the testable hypothesis cannot be the apex of scientific standards either. Experience and empirical testing are not irrelevant to science; they can help separate good explanations from bad ones at low levels of scientific analysis. What is of paramount importance though is that we can invent explanations that exist in domains of imagination far removed from anything we can conceivably experience, and that's where David Deutsch wants to go.
10.E.4 Conjecture and Criticism
“Conjecture” is a catch-all term, encompassing guess, explanation, hypothesis, theory, model, and law. For Deutsch, as for Popper, conjectures are products of the creative mind. Their origins don't matter; they are important only for what they say about the world. In Popper's program, we test conjectures and reject the falsified ones. In Deutsch's program we criticize conjectures and reject the bad ones. Testing is an empirical process and criticism is an intellectual one.
We need to evaluate explanations that are not amenable to experimental testing and to distinguish between better and worse conjectures based intellectual considerations alone.
For these reasons and more, Deutsch is relentlessly upbeat about progress. There will be no end of problems for science to tackle, but at bottom, all problems are attributable to lack of knowledge, so they will eventually be solvable. There are, literally, no boundaries to the growth of knowledge except those established by the laws of physics. His vision extends far beyond the scope of experimental science. Why shouldn't we, he asks, be able to take on notions such as “beauty” or “culture” and discover good explanations for them via Conjectures and Criticism?
So far, probably nothing about Deutsch's thinking strikes you as being excessively radical. Unusual, perhaps, but not outlandish, and you might be wondering why he believes that our most basic conceptions of science need major revisions. The answer is that he wants to take science into new dimensions.
10.E.5 Is Theoretical Quantum Mechanics Still Science?
In trying to solve certain otherwise intractable problems, some theoretical physicists have devised such fantastically exotic hypotheses that they are beyond the possibility of empirical testing.27 This development has become a source of anxiety in physics. In the words of two physicists,28 “As we see it, theoretical physics risks becoming a no-man's land between mathematics, physics, and philosophy that does not truly meet the requirements of any.” What's going on is a “battle for the heart and soul of physics.” It may be disconcerting to imagine physics, the “queen of sciences” as leaving (or being banished from?) science, but concerns along this line have been growing for years. Deutsch rises to the challenge of rescuing theoretical physics with his new approach for assessing scientific Truth. Weighty issues are at stake, so it's worthwhile for nonphysicists to peek into Deutsch's world to get a sense of what he's trying to do. Don't worry if you find the following sections somewhat confusing; they are meant as a sketch of scientific problems that have not yielded to conventional problem-solving strategies. The details are not critical.
We'll start by returning to the concept of the multiverse. There are many versions of multiverse theory and Deutsch favors a “many worlds” interpretation; that is, the universe that we perceive is only one of a large, potentially infinite number of universes diverging from ours in degrees ranging from imperceptible to enormous. The differences among universes are not arbitrary though. There are rules: the laws of physics hold everywhere, and the universes cannot communicate with each other; each is an isolated entity. A conceptual hurdle for many of us is that we're not supposed to imagine the universes as stretching out contiguously in space. Rather the universes “overlap” and occupy the “same” space. I'll have more to say about this later, but it helps to keep in mind the recommendation of the poet Samuel Taylor Coleridge29 who said that readers should adopt a “willing suspension of disbelief” in order to grasp the “semblance of truth” that a writer (in this case, me) is trying to express by conjuring up weird and insubstantial images.
Why would anyone waste time on such a crazy idea as the multiverse? As strange as it is, the multiverse conjecture might help rid physics of snags that lurk within the conventional theories, such as randomness and the dual “waveparticle” nature of tiny, fundamental fragments of physical reality. For instance, the physicist Werner Heisenberg theorized that we can never in principle determine the precise position of an electron in an atomic nucleus; it is strictly indeterminate. At best the electron has a random probability of being in one place or another. Though it is integral to the standard interpretation of quantum mechanics, the existence of randomness in physical law has long bothered physicists. Einstein rejected it with the comment that “God does not play dice with the world.”30 Resorting to abstract probabilistic calculations to describe the electron's location represented a cop-out, he implied; if we truly knew what we were talking about, we wouldn't need such calculations. David Deutsch agrees; we shouldn't surrender to randomness.
Physicists like Einstein and Deutsch believe that physical laws are rigorously deterministic, and, at least for Deutsch, the multiverse solves problems associated with randomness and indeterminacy. For a crude physical picture of how this could work, suppose there were a huge national lottery where hundreds of millions of numbered tickets were sold and a scrupulously fair drawing held. The winning number, we would normally say, was picked “at random.” But this would be an illusion of our cramped point of view. What we experience as a chance event, when seen from a perspective that takes in the multiverse, is an absolute certainty. The winning number is the one that had to win in our universe; its selection was unalterably fixed by the myriad details of its history, from the raw materials and manufacturing of the little numbered balls bouncing around in the lottery machines, to the conditions and precise timing of the drawing. Everything that could possibly have influenced which balls the machines spit out was the result of a definite physical event; no randomness allowed. Variance in any of the details would have resulted in another number's popping up. The most bizarre corollary of multiverse reasoning, however, is that every number must be a winner in some universe (even your number, somewhere, but probably not here).
Back to physics: each tiny physical entity, such as an electron, is like a lottery number and has a defined position, speed, direction—actually, ranges of them—and each parameter is realized in a different universe. There is no shortage of universes. And it's mind-boggling scale is not the most intellectually taxing aspect of the multiverse. In Deutsch’s hands, the concept of the multiverse is the tool that rationalizes physics by eliminating the eerie principle that a bit of matter such as an electron is both a wave and particle simultaneously, that it has wave-particle duality.
Here’s the description of wave-particle duality for us lay people: if you send a beam of light (a stream of photons) through a single narrow open slit in an otherwise opaque screen, the pattern on the wall behind the screen where the beam hits is a single fuzzy bar of light; it’s what you’d expect if photons were minute solid particles. If you repeatedly fired a BB gun at a screen with an analogous (bigger) slit, the BBs that got through would hit the wall behind in a similar fuzzy-bar pattern. If you shot BBs at a screen with two slits, you’d see two parallel fuzzy bars on the other side, just as you’d expect.
On the other hand, if you send a beam of photons through two identical narrow slits parallel to each other on the same screen, then you don’t see a pattern of two fuzzy bars on the other side. Instead, you see a regular array of alternating light and dark bars; an interference pattern. This is eerie because we just said that when you have one slit, a photon beam behaves like a stream of miniscule BBs, and now we’re saying that when you have two slits, it doesn’t.
The interference pattern that you get with photons is like the interference pattern of waves that you see when a single ocean wave passes through two narrow openings in a stone jetty at a beach. Instead of two simple waves emerging on the other side, there is a complex pattern of wavelets as the wave goes through both openings. The wavelets coming out interact by alternately cancelling and reinforcing each other (i.e., interfering with each other). The big deal is that waves create interference patterns, particles do not. To recap: when a photon goes through one slit, it behaves as a particle; when a photon goes through two slits, it behaves as a wave. So which is it? This is the infamous double-slit experiment, and it embodies the paradox of wave-particle duality. (There are many excellent videos depicting wave-particle duality and the double-slit experiment online—here’s a good one for non-physicists: https://www.youtube.com/ watch?v=fwXQjRBLwsQ.)
Everyone agrees on the observations; the question is, what do they mean? One fundamental difficulty is logical as well as physical: a wave is an extended thing—it has volume, it occupies space. An electron has virtually no volume and occupies virtually no space. When waves collide, they interfere with each other, meaning they partially cancel and partially reinforce each other; when particles collide, they bounce off each other. It appears that wave-particle duality is logically self-contradictory, and hence, accepting it means giving up on logic or ignoring the self-contradiction. So much the worse for logic, says the physics Old Guard: don’t worry about “understanding” the duality, just “shut up and calculate.”31 If the calculations predict future events reliably, that's all you can hope for.
A paradox like wave-particle duality is anathema to Deutsch. It's selfcontradictory nature is a signal that we don't properly understand nature. The very name, “wave-particle duality,” is merely a description, it doesn't explain anything. The way out of the maze is through the multiverse, where wave-particle duality does not exist. There, photons are always and only particles; however, what seems to be a single photon is actually many simultaneous, overlapping “instances” of photons that can diverge and merge under various circumstances. Each instance of a photon is in a different universe that “overlaps” with ours (i.e., the only one that we're aware of). The many overlapping instances of a given photon are interchangeable and indistinguishable. They are “fungible” in the way that dollars in your bank account are; neither an individual photon nor a dollar in your account have unique identities. No matter how pitifully few dollars might be in your account, the notion of an individual dollar is meaningless; the ones from your paycheck are inextricably intermingled with the ones that your grandmother gave you on your birthday. You couldn't tell them apart if you wanted to. On the other hand, if you withdrew some bills, you could note their serial numbers, draw a mustache on George Washington, etc.—they would become individuals under certain circumstances. Then, if you redeposited them into the bank, they would return to fungibility and lose their identity again.
Deutsch thinks that photons are sort of like that. They may go off and lead differentiated, individualized existences in different universes or remain in a fungible state where universes intersect and repeatedly pop out in one, or in many more than one, universe. Or, like dollars that you hid inside your mattress and forgot about, some photons can remain individualized and never return to fun- gibility. Permanently individualized photons will have experienced subtly different histories and have acquired different properties; they may be traveling in an array of slightly different directions, for example.
Interference is what we perceive in our universe when two or more separate instances of the same photon cancel each other and return to the fungible state. In the double-slit experiment, some non-individualized photons encounter selfinstances and engage in mutual cancellation, some will take slightly different paths through the two slits in the screen and remain individualized. The sum total of all of those cancellations and path deviations produces the net interference pattern that we observe. Apparent wave-particle duality, like randomness, is an illusion created by our limited field of experience; a photon remains a particle, it is never a wave. The wave-like interference pattern is a population phenomenon that indirectly reveals the multiverse-level behavior of photons, and Deutsch believes that the multiverse is in fact “the only possible explanation” for quantum interference.
Although superficial, this overview of reasoning from the standpoint of the multiverse should make it obvious why Deutsch must dispense with empiricist standards and the hypothesis: the multiverse theory cannot be directly tested; it is not falsifiable. We're faced with a choice: either concede that some aspects of nature are forever beyond the scope of rational investigation or redefine rational inquiry to include theories that are hard to vary, though unfalsifiable.
Doesn't this redefining of scientific reasoning throw open the doors to limitless fanciful rubbish? No, says Deutsch, because strict rules remain in place. '1 he multiverse theory is a hard-to-vary explanation for the occurrence of significant physical phenomena. It is severely constrained by observation and other good explanations: the laws of physics are obeyed in all universes, interuniverse communication is not allowed, randomness is not allowed, photons are purely particles, and so on. The theory is specific and can be subjected to criticism, error-detection, and correction. Deutsch admits that the multiverse is a powerfully strange picture of reality but he argues that we can make scientific progress by substituting the goal of explanatory understanding through Conjecture and Criticism for experiential testability.
10.E.6 Summary and Critique
Deutsch's revision of the scientific enterprise makes critics uneasy, although they find his novel ideas thought-provoking.32 Some conclude that Deutsch is a hard-core Popperian3x however, as I've tried to show, Deutsch's program extends well beyond Popper's. The two are in sync when it comes to fallibilism, rejection of induction, and insistence on creativity as the source of hypotheses, and in their vision of science as a means of approaching Truth about nature. I doubt very much that Popper would have considered Deutsch a Popperian, though it is worth recalling that Popper did not believe that unfalsifiable ideas were necessarily worthless33
A key unanswered question for Deutsch is how we should undertake the search for good explanations in the context of everyday experimental science. '1 he hard-to-vary criterion may be hard to apply when trying to choose between better and worse explanations except in extreme cases, such as when comparing mythology with modern astronomy.
Maybe the rule should be, “If you can test a hypothesis experimentally with the empirical falsifiability standard, then do it.” Deutsch might agree with this rule. “Experience is indeed essential to science,” he says, “its main use is to choose between theories that have already been guessed.” '1 his process is, of course, central to Popper's philosophy. But if we carry out conventional testing whenever possible and reject explanations that fail the tests as bad explanations, then does the hard-to-vary framework add anything? How does it enhance the usual process of seeking scientific Truth through observation and experiment? Perhaps, except in special cases (e.g., advanced theoretical physics), science doesn't need to abandon its customary practices of hypothesis construction and testing.
What about intellectual areas apart from science? I have skipped over the application of his philosophy to nonscientific fields including esthetics, political science, and cultural studies, but he is confident that Conjecture and Criticism, unencumbered by the demands of empiricism and testability, can make strong contributions to these fields as well. Indeed, The Beginning of Infinity can be read as a book-length defense of the proposition that Conjecture and Criticism is the only possible explanation for progress of all kinds. Still, Deutsch is a committed fallibilist who believes that knowledge is always incomplete and that our best hypotheses are perpetually subject to change without notice. Could there be a better explanation for progress than Conjecture and Criticism? He must be keeping an open mind as to the possibility.
10.