7 - A Conversation About Justification

# 7 - A Conversation About Justification Popper showed that [Induction is Impossible](Induction%20is%20Impossible.md). Inductive generalization from observations is impossible, and inductive [Justification](Justification.md) is invalid. Inductivism rests upon the mistake idea of science as seeking predictions on the basis of observations, rather than as seeking explanations in response to problems. Science actually makes progress by conjecturing new explanations and then choosing between the best ones by experiment. This is accepted by most philosophers of science. What is not accepted by most philosophers is that this process is *justified*. Science seeks better explanations. A scientific explanation accounts for our observations by postulating something about what reality is like and how it works. But why should a better *explanation* be what we always assume it to be a token of a *truer theory*? Why, for that matter, should a downright bad explanation necessarily be false? There is indeed no logically necessary connection between truth and explanatory power. A bad explanation (such as [Solipsism](Solipsism.md)) *may* be true. Even the best and truest available theory may make a false prediction in particular cases, and those might be the very cases in which we rely on the theory. No valid form of reasoning can logically rule out such possibilities, or even prove them unlikely. But in that case, what *justifies* our relying on our best explanations as guides to practical decision-making? More generally, whatever criteria we used to judge scientific theories, how could the fact that a theory satisfied those criteria today possibly imply anything about what will happen if we rely on the theory tomorrow? This is the modern form of the problem of induction. Again, most philosophers today accept that scientific progress is made through conjectures and refutations, and that theories are accepted when their rivals are refuted, not when they had numerous confirming instances. They accept that the knowledge obtained in this way *tends to be reliable*. The problem is that they don't see why this should be the case. > **Problem:** Why should a theory (selected via conjecture and refutation) be a *reliable basis for action*? Philosophers yearn for this missing *justification*. In a sense, [Justification Is Seeking A Foundation](Justification%20Is%20Seeking%20A%20Foundation.md). Let us redefine an "inductivist" as someone who believes that the *invalidity* of inductive justification is a problem for the foundations of science. We can in fact justify our reason to act based on our best theories. This justification—always tentative—comes from the explanations provided by the relevant scientific theories. To the extent that those explanations are good, it is rational to rely on the predictions of the corresponding theories. [Explanations Are Justified By Their Superior Ability to Solve Problems They Address](Explanations%20Are%20Justified%20By%20Their%20Superior%20Ability%20to%20Solve%20Problems%20They%20Address.md). [Argument Is How We Achieve Justification](Argument%20Is%20How%20We%20Achieve%20Justification.md). ## The Updated Problem of Induction The problem of induction is not how to justify or refute the principle of induction, but rather, taking for granted that it is invalid, how to justify any conclusion about the future from past evidence. According to Popperian methodology, one should in these cases rely on the _best-corroborated_ theory — that is, the one that has been subjected to the most stringent tests and has survived them while its rivals have been refuted. In these cases it would be *justified* to rely on the theory. The process of corroboration has justified the theory, in the sense that it's predictions are most likely to be true than the predictions of other theories. However, not just *any* other theories—specifically, *rival* theories. To be clear, the justification does *not* come from the "evidence". The evidence—all experiments whose outcomes the theory correctly predicted in the past—is consistent with an infinite number of theories, predicting every possible logical future outcome. ## Justification Comes From Argument In that case, what exactly about choosing the best-corroborated theory is justified, if not due to the evidence? The answer is [Argument](Argument.md). Only [Argument](Argument.md) ever [Justifies](Justification.md) anything—tentatively, of course. We are [Fallible](Fallibilism.md) and thus all theorizing is subject to error. But still, argument can sometimes justify theories. That is what argument is for. A theory will not be justified by *pure* argument, as in the case of a mathematical theorem. The evidence will play *some role*. When dealing with empirical theories, according to the Popperian scientific methodology, crucial experiments play a pivotal role in deciding between it and its rivals. The rivals were refuted, it survived. Now at this point an inductivist may wish to say "in consequence of the refuting and surviving, all of which happened in the past, using the theory in the future is now justified". But it is misleading to use "in consequence of" when we are not talking about a logical [Deduction](Deduction.md). But in that case, *what sort of consequence was it*? Both *argument* and *outcomes* of experiment justified the theory. In other words, the outcomes of past experiment in conjunction with argument justified the theory. But then what exactly was it about those actual past outcomes that justified the prediction, as opposed to other possible past outcomes that might have justified a contrary prediction? It is that the actual outcomes refuted all rival theories, and corroborated the prevailing theory. Notice that the actual outcomes did not refute *all logically possible theories* (of which there are infinitely many). Rather, it refuted all *rival* theories, which are those proposed in the course of a rational controversy. If you think about this for a moment it may seem rather strange that the reliability of a theory depends on the accident of what *other* theories—false theories—have been proposed. Shouldn't a theories validity depend on it's own *content*, and on the experiment evidence? Well a bit more thought will show that this isn't so strange after all. Consider the classic, common sense view of induction. It may speak of a theory being reliable or not *given certain evidence*. In that case, the theory's reliability is conditional on the evidence that is available. In the Popperian Paradigm a theory's reliability is conditional on the *problem situation*. The most important features of a problem-situation are: what theories and explanations are in contention, what arguments have been advanced, and what theories have been refuted. ‘Corroboration’ is not just the confirmation of the winning theory. It requires the experimental refutation of rival theories. Confirming instances in themselves have no significance. So we could say: under inductivism, observation was supposed to be primary. One imagined a mass of past observations from which the theory was supposed to be induced, and observations also constituted the evidence which somehow justified the theory. In the Popperian picture of scientific progress, it is not observations but problems, controversies, theories and criticism that are primary. Experiments are designed and performed only to resolve controversies. Therefore only experimental results that actually do refute a theory — and not just any theory, it must have been a genuine contender in a rational controversy — constitute ‘corroboration’. And so it is only those experiments that provide evidence for the reliability of the winning theory. The ‘reliability’ that corroboration confers is not absolute but only relative to the other contending theories. That is, we expect the strategy of relying on corroborated theories to select the best theories from those that are proposed. That is a sufficient basis for action. We do not need (and could not validly get) any assurance about how good even the best proposed course of action will be. Furthermore, we may always be mistaken, but so what? We cannot use theories that have yet to be proposed; nor can we correct errors that we cannot yet see. ### Thought Experiment Suppose that a theory has passed through this whole process. Once upon a time it had rivals. Then experiments were performed and all the rivals were refuted. But it itself was not refuted. Thus it was corroborated. What is it about its being corroborated that justifies our relying on it in the future? The answer is that since all its rivals have been refuted, they are no longer rationally tenable. The corroborated theory is the only rationally tenable theory remaining. But doesn't that only shifts the focus from the future importance of past corroboration to the future importance of past refutation? The same problem remains. Why, exactly, is an experimentally refuted theory ‘not rationally tenable’? Is it that having even one false consequence implies that it cannot be true? In terms of the future applicability of the theory, that is not a logically relevant criticism. A refuted theory may not be true *universally*—in particular, it cannot have been true in the past when it was tested. But it could still have many true consequences, and in particular it could be universally true in the future. Another way to put this is that although a refuted theory is strictly false (because it makes some false predictions), all its predictions about the future might nevertheless be true. In other words, a *different theory* which makes the same predictions about the future but different predictions about the past, might be true. So a better phrasing of the question could be: why does the refutation of a theory also render untenable every variant of the theory that agrees with it about the future—even a variant that has not been refuted? It is not that refutation *renders* such theories untenable. It is just that sometimes they already are untenable by virtue of being a bad explanation. And that is when science can make progress. For a theory to win an argument, all rivals must be untenable, and that includes all the variants of the rivals which anyone has thought of. But remember, it is only the rivals *which anyone has thought of* that need be untenable. ### A Logically Consistent Theory That is a Bad Explanation Now consider an extension of the previous thought experiment. We are conversing with [Mr Witt](Mr%20Witt.md): > [!quote] > "Consider the theory of gravity. This will be our prevailing theory. Now, consider a "rival" theory—namely that when you jump from a tall building, you float. Other than that, the prevailing theory holds universally. One might argue that all past tests of the prevailing theory were tests of this rival theory (for you have never jumped from a tall building, and other than that single additional postulate, the theories are identical). Thus the prevailing theories refuted rivals are also this new challenger's refuted rivals. This new rival theory is exactly as corroborated as the prevailing theory. How can this rival theory then be untenable?" To which we can answer: this new rival theory has effectively every single fault in the Popperian book! Mr Witt's theory is constructed from the prevailing one by appending an unexplained qualification about me floating. That qualification is, in effect, a new theory, but he has given no argument either against the prevailing theory of my gravitational properties, or in favour of the new one. He has subjected his new theory to no criticism (other than what I am giving it now) and no experimental testing. It does not solve — or even purport to solve — any current problem, nor has he suggested a new, interesting problem that it could solve. Worst of all, his qualification explains nothing, but spoils the explanation of gravity that is the basis of the prevailing theory. It is this explanation that justifies our relying on the prevailing theory and not on his. Thus by all rational criteria your proposed qualification can be summarily rejected. Mr Witt may then respond: > [!quote] > But could't I say exactly the same thing about the prevailing theory? It differs from mine only by the same minor qualification (name, that you will fall if you jump from this tall building). You argue that I ought to have explained my additional qualification—but why are our positions not symmetrical? The reason is that this rival theory does not come with an explanation of its predictions, but the prevailing theory does. This has nothing to do with which theory was proposed first, by the way. Any rational person who was comparing Mr Witt's theory with the prevailing one would immediately reject the rival theory. For the fact that Mr Witt's theory contains an unexplained modification of another theory is manifest in the very statement of it. Specifically, it is of the form 'such-and-such a theory holds universally, except in such-and-such a situation', but he never explains why the exception holds. Depending on how Mr Witt phrased this (and even potentially what language he expressed this in), the unexplained postulate may be explicit or implicit. But regardless, we are not concerned with the *form* of the theory; we are concerned with the *substance*. Mr Witt's theory solves nothing and only worsens the current problem-situation. It asserts the existence of a physical *anomaly* which is not present according to the prevailing theory. ### Theories Postulating Anomalies We are not saying that it is a [Principle of Rationality](Principles%20of%20Rationality.md) that a theory which asserts the exist of an objective, physical anomaly, is less likely to make true predictions than one that doesn't. What we are saying is that theories postulating anomalies without explaining them are less likely than their rivals to make true predictions. More generally, it is a principle of rationality that theories are postulated in order to solve problems. Therefore any postulate which solves no problem is to be rejected. That is because a good explanation qualified by such a postulate becomes a bad explanation. It is in this way that there really is an objective difference between theories that make unexplained predictions and those than don't. ### But wait, is this a principle of induction? We have just justified a theory about the future (the prevailing theory of gravity) as being more reliable than another theory (Mr Witt's rival), even though they are both consistent with all currently known observations. Since the prevailing theory applies both to the future and to the past, did we not just justify that (in regards to gravity) the *future resembles the past*. In order to go from corroborated to reliable, we examined the theories explanatory power. So, did we not show that a "principle of seeking better explanations", together with some observations and arguments, *imply* that the future will, in many respects, resemble the past? This seems to imply a principle of induction! Let's slow down and tread carefully—we are currently navigating some tricky terrain. While we have implied in some sense that the future resembles the past, this is rather vacuous. *Any* theory about the future will assert that it resembles the past in some sense. However, this does not allow us to derive or justify any theory or prediction about the future! For example, we could not use it to distinguish the rival theory of gravity and the prevailing one, for they both say, in their own way, that the future resembles the past. Could we then say: "if, in the present, there is no explanatory theory predicting that a particular anomaly will happen in the future, then that anomaly is unlikely to happen in the future". We could say that! And I believe it to be true. However, it is not of the form "the future is likely to resemble the past". And, in trying to make it look like that, we have specialized it to specific cases such as "in the present", "in the future", and in the case of an "anomaly". But, in reality, it is just a general statement about the efficacy of argument. In short, if there is no argument in favor of a postulate, then it is not reliable. Past, present or future. Anomaly or no anomaly. Period. Nothing in the concepts of ‘rational argument’ or ‘explanation’ relates the future to the past in any special way. Nothing is postulated about anything ‘resembling’ anything. Nothing of that sort would help if it were postulated. In the vacuous sense in which the very concept of ‘explanation’ implies that the future ‘resembles the past’, it nevertheless implies nothing specific about the future, so it is not a principle of induction. There is no principle of induction. There is no process of induction. No one ever uses them or anything like them. And there is no longer a problem of induction. ## Reconsider this 'rival' theory a bit more closely... Let us consider this conversation further, looking a bit more closely at Mr Witt's theory. We might respond: > [!quote] > As we have agreed, your theory consists objectively of a theory of gravity (the prevailing theory), qualified by an unexplained prediction about me. It says that I would float, unsupported. ‘Unsupported’ means ‘without any upward force acting’ on me, so the suggestion is that I would be immune to the ‘force’ of gravity which would otherwise pull me down. But according to the general theory of relativity, gravity is not a force but a manifestation of the curvature of spacetime. This curvature explains why unsupported objects, like myself and the Earth, move closer together with time. Therefore, in the light of modern physics your theory is presumably saying that there is an upward force on me, as required to hold me at a constant distance from the Earth. > > But where does that force come from, and how does it behave? For example, what is a ‘constant distance’? If the Earth were to move downwards, would I respond instantaneously to maintain the same height (which would allow communication faster than the speed of light, contrary to another principle of relativity), or would the information about where the Earth is have to reach me at the speed of light first? If so, what carries this information? Is it a new sort of wave emitted by the Earth — in which case what equations does it obey? Does it carry energy? What is its quantum-mechanical behaviour? Or is it that I respond in a special way to existing waves, such as light? In that case, would the anomaly disappear if an opaque barrier were placed between me and the Earth? Isn’t the Earth mostly opaque anyway? Where does ‘the Earth’ begin: what defines the surface above which I am supposed to ‘float’? > > For that matter, what defines where I begin? If I hold on to a heavy weight, does it float too? If so, then the aircraft in which I have flown could have switched off their engines without mishap. What counts as ‘holding on’? Would the aircraft then drop if I let go of the arm rest? And if the effect does not apply to things I am holding on to, what about my clothes? Will they weigh me down and cause me to be killed after all, if I jump over the railing? What about my last meal? > > I could go on like this ad infinitum. The point is, the more we consider the implications of your proposed anomaly, the more unanswered questions we find. This is not just a matter of your theory being incomplete. These questions are dilemmas. Whichever way they are answered, they create fresh problems by spoiling satisfactory explanations of other phenomena. > > So your additional postulate is not just superfluous, it is positively bad. In general, perverse but unrefuted theories which one can propose off the cuff fall roughly into two categories. There are theories that postulate unobservable entities, such as particles that do not interact with any other matter. They can be rejected for solving nothing (‘Occam’s razor’, if you like). And there are theories, like yours, that predict unexplained observable anomalies. They can be rejected for solving nothing and spoiling existing solutions. It is not, I hasten to add, that they conflict with existing observations. It is that they remove the explanatory power from existing theories by asserting that the predictions of those theories have exceptions, but not explaining how. You can’t just say ‘spacetime geometry brings unsupported objects together, unless one of them is David, in which case it leaves them alone’. Either the explanation of gravity is spacetime curvature or it isn’t. Just compare your theory with the perfectly legitimate assertion that a feather would float down slowly because there would indeed be a sufficient upward force on it from the air. That assertion is a consequence of our existing explanatory theory of what air is, so it raises no new problem, as your theory does. The crux of our argument to Mr Witt is that his additional postulate is not just *superfluous*—it is *positively bad*. It predicts an unexplained observable anomaly, and by doing so it *solves nothing* and *spoils existing explanations*. This is worth pausing on. In [Logic](Logic.md), [Deduction](Deduction.md) has [Logical Consequences](Logical%20Consequence.md). I have also talked about how [Explanations Imply Consequences](Explanations%20Imply%20Consequences.md). The mental model I have here is that both deduction and explanation have a structure through which truth can *flow* or *move around*. Part of the structure that an explanation has *links it to other explanations*. By adding an additional postulate (I will float) to an existing explanation, that has *consequences* that flow to many different places. In our example, a consequence is that we have worsened our explanation of gravity (of which we are actively dealing with), but we also have worsened our explanations of fields, waves, the speed limit imposed by light, and so on. These are *consequences* of the new rival explanation. And it is these consequences that clearly make it *worse* than the prevailing theory. ## A Final Argument From a Defeated Mr Witt Now Mr Witt may respond with one final, albeit deflated counter argument: > [!quote] > Consider the raw materials for the argument. Initially, I assumed that past observations and deductive logic are our only raw material. Then I admitted that the current problem-situation is relevant too, because we need justify our theory only as being more reliable than existing rivals. And then I had to take into account that vast classes of theories can be ruled out by argument alone, because they are bad explanations, and that the principles of rationality can be included in our raw material. What I cannot understand is where in that raw material — past observations, the present problem-situation and timeless principles of logic and rationality, none of which justifies inferences from the past to the future — the justification of future predictions has come from. There seems to be a logical gap. Are we making a hidden assumption somewhere? To which we can respond: No, there is no logical gap. What you call our ‘raw material’ does indeed include assertions about the future. The best existing theories, which cannot be abandoned lightly because they are the solutions of problems, contain predictions about the future. And these predictions cannot be severed from the theories’ other content, as you tried to do, because that would spoil the theories’ explanatory power. Any new theory we propose must therefore *either* be consistent with these existing theories, which has implications for what the new theory can say about the future, or contradict some existing theories but address the problems thereby raised, giving alternative explanations, which again constrains what they can say about the future. Again, notice the use of the word "severed". This is referring to a strong, clearly connected structure that you have no control over! It is implied by your theories, just as consequences are implied in classic logical deduction. You cannot arbitrarily disconnect from this structure. Note that this is similar to the consequences implied in the case of [Reach](Reach.md). So we have *no universal principle of reasoning* which says that the future will resemble the past, but *we do have actual theories which say that*. In other words we have *specific theories*, our best explanations, that may say that the future will resemble the past. It is these specific theories that we are right to rely on. Notice what we have *not said*—we have not said that we have actual theories which imply a limited form of an inductive principle. That is *not* that case. Our theories simply assert something about the future. Vacuously, any theory about the future implies that the future will ‘resemble the past’ in some ways. But we only find out in what respects the theory says that the future will resemble the past after we have the theory—meaning that we couldn't use this *ahead of time* to make predictions about the future, we need the theory first! You might as well say that since our theories hold certain features of reality to be the same throughout space, they imply a ‘spatial principle of induction’ to the effect that ‘the near resembles the distant’. Let me point out that, in any practical sense of the word ‘resemble’, our present theories say that the future will not resemble the past. The cosmological ‘Big Crunch’, for instance (the recollapse of the universe to a single point), is an event that some cosmologists predict, but which is just about as unlike the present epoch, in every physical sense, as it could possibly be. The very laws from which we predict its occurrence will not apply to it. ## Argument Justifies the Principles of Rationality Mr Witt may reasonably then ask: > [!quote] > We have seen that future predictions can be justified by appeal to the [Principles of Rationality](Principles%20of%20Rationality.md). But what justifies those? They are not, after all, truths of pure logic. So there are two possibilities: either they are unjustified, in which case conclusions drawn from them are unjustified too; or they are justified by some as yet unknown means. In either case there is a missing justification. I no longer suspect that this is the problem of induction in disguise. Nevertheless, having exploded the problem of induction, have we not revealed another fundamental problem, also concerning missing justification, beneath? What justifies the principles of rationality is [Argument](Argument.md), as usual. As an extreme example, what justifies our relying on the laws of deduction, despite the fact that any attempt to justify them logically must lead either to [Circularity](Circular%20Argument.md) or to an [Infinite Regress](Infinite%20Regress.md)? They are justified because no explanation is improved by replacing a law of deduction. This may not seem to be a very secure foundation for pure [Logic](Logic.md). And it is not perfectly secure! We are [Fallible](Fallibilism.md). And we should not expect it to be perfectly secure, for logical reasoning is no less a physical process than scientific reasoning is, and it is inherently fallible. The laws of logic are not self-evident. There are people, the mathematical ‘intuitionists’, who disagree with the conventional laws of deduction (the logical ‘rules of inference’). They cannot be proved wrong, but we can convincingly *argue* that they are wrong. Thus there is not "problem of deduction". There is no problem with any of the usual ways of justifying conclusions in science, philosophy or mathematics. However, it is an interesting fact that the physical universe admits processes that create knowledge about itself, and about other things too. We may reasonably try to explain this fact in the same way as we explain other physical facts, namely through explanatory theories. An argument can be made that the [Turing Principle](Turing%20Principle.md) is the appropriate theory in this case. It says that it is possible to build a [Virtual Reality](Virtual%20Reality.md) generator whose repertoire includes every physically possible environment. If the Turing principle is a law of physics, as I have argued that it is, then we should not be surprised to find that we can form accurate theories about reality, because that is just virtual reality in action. Just as the fact that steam engines are possible is a direct expression of the principles of thermodynamics, so the fact that the human brain is capable of creating knowledge is a direct expression of the Turing principle. How do we know the Turing principle is *true*? We don't, of course! But wait, if we can't justify the Turing principle, then have we just lost our justification for relying on scientific predictions? No! We have now moved on to a completely different question! We are now discussing an apparent fact about physical reality, namely that it can make reliable predictions about itself. We are trying to explain that fact, to place it within the same framework as other facts we know. I suggested that there may be a certain law of physics involved. But if I were wrong about that, indeed even if we were entirely unable to explain this remarkable property of reality, that would not detract one jot from the justification of any scientific theory. For it would not make the explanations in such a theory one jot worse. ## Middle Out There is a misconception that is present throughout this entire note (and chapter 7 of [Fabric of Reality](Fabric%20of%20Reality.md)). The misconception is about the very nature of [Argument](Argument.md) and [Explanation](Explanations.md). Mr Witt seemed to be assuming that arguments and explanations, such as those that justify acting on a particular theory, have the form of mathematical proofs, proceeding from assumptions to conclusions. You look for the ‘raw material’ (axioms) from which our conclusions (theorems) are derived. Now, there is indeed a logical structure of this type associated with every successful argument or explanation. But the process of argument does not begin with the ‘axioms’ and end with the ‘conclusion’. Rather, it starts *in the middle*, with a version that is riddled with inconsistencies, gaps, ambiguities and irrelevancies. All these faults are criticized. Attempts are made to replace faulty theories. The theories that are criticized and replaced usually include some of the ‘axioms’. That is why it is a mistake to assume that an argument *begins with*, or is justified by, the theories that eventually serve as its ‘axioms’. The argument ends — tentatively — when it seems to have shown that the associated explanation is satisfactory. The ‘axioms’ adopted are not ultimate, unchallengeable beliefs. They are tentative, explanatory theories. Argument is not the same species of thing as deduction, or the non-existent induction. It is not based on anything or justified by anything. And it doesn’t have to be, because its purpose is to solve problems — to show that a given problem is solved by a given explanation . --- Date: 20250401 Links to: [Fabric of Reality](Fabric%20of%20Reality.md) Tags: References: * []()