Induction and Science

[list]

• The principle known as the uniformity of nature holds that nature will behave in the future as it has in the past.
• The only evidence for the truth of this principle is that, so far, we have observed that nature has behaved in the present as it had in the past.
• It is impossible to observe in the present what will happen in the future.
• Therefore, specific observations that nature has behaved in the present as it had in the past are the only way to justify the generalized conclusion that nature will behave in the future as it has in the past.
• Therefore, the uniformity of nature principle can be derived by no method other than inductive reasoning.

The deductive reasoning above can be applied to prove the inductive nature of other scientific laws, theories, or models that claim to predict what will happen in the future based on what has been observed in the past.Edited by Stephen Push, : No reason given.

You stated what for you would constitute evidence for induction in science, and I provided precisely what you requested.The first sentence I quoted is clearly a specific observation. Note the use of the past tense, singular nouns, and terms such as "this phenotype" and "that population." In the second sentence, the authors alert us that they are proposing a generalized principle by starting off with the clause "More generally," by switching to the present tense, and by writing about plural "innovations" rather than this innovation or that innovation.Is the second sentence a proposed hypothesis? It is only if you are using the broad sense of the word "hypothesis," which is synonymous with "generalized principle." It is not a proposed hypothesis in the narrow sense of a proposition that can be tested by observation or experiment. The latter use would require specific cases of the generalized principle.Edited by Stephen Push, : No reason given.

Salmon’s paper is a refutation of Popper’s view of induction in science. Salmon shows that induction is necessary for rational prediction — at least in practical decision-making and probably in theoretical work as well.I’m a fan of Popper’s view of the value of falsifiability in the demarcation of science from other fields. But I do not believe that induction and falsifiability are mutually exclusive or that induction-free science is necessary to defend Popper’s view of demarcation.Edited by Stephen Push, : No reason given.Edited by Stephen Push, : No reason given.Edited by Stephen Push, : No reason given.

I agree inductive methodology does not distinguish science from non-science. But that doesn't mean science must be induction-free to be distinguished from non-science.Do you believe that acknowledging a role for induction in science would refute Popper's view that falsifiability is the hallmark of science?Edited by Stephen Push, : No reason given.

Welcome to the fray, RAZD. If there could be A outside B in the diagram, then "all A is B" would be an induction, not a deduction. Since "all A is B" is based on induction, I don't see any logical difference between that conclusion and the conclusion "all B is A."I second Panda's request for a real-world example. I believe the hypothetical you have introduced is different from the types of problems generally encountered in science. In the real world, we wouldn't be able to study every A and every B. We wouldn't even know whether the elements of A and B that we measured are all of the elements that exist. Science often uses inference to the best explanation (IBE), which is a non-deductive method. For example, Darwin used IBE when he marshaled information about many plants and animals to make the case for common descent.

Thanks for the real-world example, RAZD. That is technically a deductive argument, but it conceals a lot of non-deductive work. The first two lines of your syllogism rely on the work done by Darwin and others -- using the non-deductive IBE method -- to establish the theory of evolution. To know whether the third line is true, scientists would have to use IBE to determine where this new animal fits into the tree of life.After non-deductive methods have done all of the heavy lifting, the remaining deductive argument is trivial. Rather than a "best guess," that sounds like a shot in the dark. I find it hard to imagine how such a line of reasoning could be useful in actual scientific research.

In my opinion, that is an eminently sound bit of reasoning. But it is not deductive, because the conclusion is not a necessary consequence of the premises. As I said in my previous post, your dog-canine-mammal syllogism is deductive. But it is not very helpful because it presupposes common decent, a concept which I believe was developed largely by non-deductive means. If we were to presuppose special creation rather than common decent, it would not be an observed fact that dogs are canines or that canines are mammals. Precisely. It is a true but trivial deduction.That's not to say that I deny the value of deduction in science. Deduction is useful in deriving testable hypotheses from theories. Mathematical deduction is useful in models and statistical analysis.Edited by Stephen Push, : No reason given.Edited by Stephen Push, : No reason given.

I agree. The point I am trying to make is that this evidence-based and highly tested theory has been developed largely with non-deductive methods. Your deductive syllogism seemed to take that prior work for granted and thus appeared to give deduction credit for knowledge that had been discovered non-deductively. Imagine that one of these dog-like foxes escaped and you found it without knowing what it was or where it came from. How would you determine its origins? Could you do so by a completely deductive method? I think not. You might look at morphology, behavior, biochemistry, and genetics. You might try to breed it with other canids to see with which species it produced viable, fertile offspring. I submit that you would be using induction and/or inference to the best explanation more than deduction in this research. There it is again. A facile deductive argument is given credit for the really interesting work that was done to establish (a), (b), and (c) using mostly non-deductive methods. Please forgive me if I didn't explain my position clearly. I never meant to imply that presupposition is a good way to do science. By "presupposes" I meant that your deductive argument assumed the truth of the premises without explaining why that assumption was justified.I'm not trying to nit-pick. I believe it is important to disclose the justifications for the premises because I think doing so will reveal that this supposed deductive reasoning is basically non-deductive reasoning in disguise. Once we have established -- largely though non-deductive methods -- that dogs are canines, canines are mammals, and this new mystery animal is a kind of dog, the deductive argument seems to me to be nothing more than a tautology.Edited by Stephen Push, : No reason given.

When I first read this statement, I thought you had constructed a formally correct but trivial deduction. I guess I did that because I consider dogs to be canines, and canines to be mammals, by definition. If these phylogenetic relationships are definitive, then there is no possibility of finding a dog that is not both a canine and a mammal.On re-reading your post, however, I now realize that you apparently were not considering these relationships to be definitive. You seem to be saying that all dogs observed so far are canines, allowing for the possibility that one might find a dog that is not a canine or a canine that is not a mammal. (By the way, I don't know why you included the third premise. It seems unnecessary.)If the latter is what you meant, then your argument is inductive rather than deductive, because you have drawn generalized conclusions from specific observations. The terms "any new dog species" and "any new canine species" may seem specific, but those terms are synonymous with "all dog species" and "all canine species."I apologize if I misinterpreted your post.Edited by Stephen Push, : No reason given.Edited by Stephen Push, : No reason given.

Yes, that's essentially my position, although I would prefer to say that a theory is never more than probably true. Much of the discussion in this thread has been driven by the assertion advanced by nwr that induction is never used in science. I appreciate that you are bringing to this discussion a more nuanced examination of how both induction and deduction may be used in science. I think the words fact and true should be used with great care in science. Scientific knowledge is perpetually tentative. Scientists sometimes use deductive reasoning. And of course deductive conclusions are necessarily true when the premises are true. But I believe that the premises are often, if not always, subject to error and revision. I agree with you, because I’m a scientific realist. An instrumentalist, however, would claim that there is no way to know whether we are getting closer to reality. If I understand your position, you are saying we can make a prediction from a tentative theory and then measure how an actual experimental result differs is from the prediction. We can then modify the theory, make a new prediction, and see if we have gotten any closer to reality.A problem I see with that approach is that, by adding post hoc theory modifications, even a theory that is a poor representation of reality may become more accurate for the purpose of making predictions. By adding epicycles, for example, the Ptolemaic model became better at making predictions, but it never got any closer to reality. I can see how deduction plays a role. For example:

• Mammals have hair, produce milk, etc.
• New animal has hair, produces milk, etc.
• Therefore, new animal is a mammal.

But in the dog-like fox hypothetical, I think your conclusion is an inference to the best explanation rather than a deduction. In fact, you refer to it as the most likely answer. While descent from the current wild fox population is the most likely explanation, based on the evidence, you would not be able to rule out completely other possibilities. For example, the foxes may have descended from a dog-like fox ancestor that is now nearly extinct. I agree the latter explanation would be highly unlikely, but given the available evidence, it is not a logical impossibility.Doesn’t your use of the term parsimonious suggest the use of non-deductive reasoning?I agree that cladistics is less subjective than some other taxonomic systems, but I don’t think it is any more deductive. Nicely done! I can see you have used deductive reasoning to show that, if the evidence supports common descent, whether that process is "natural" or God-directed is irrelevant. Nevertheless, the premise that the evidence supports common descent was derived non-deductively. You have again invoked parsimony in defense of the premise.Edited by Stephen Push, : No reason given.