Labyrinths of Reason by William Poundstone

is how we distinguish reasonable hypotheses from unreasonable ones. You might answer, “By doing experiments, of course!” That is one way of distinguishing, but scientists cannot test every hypothesis, good, bad, or indifferent.
    “Research is the process of going up alleys to see if they are blind,” joked biologist Marston Bates. The role of scattershot research is severely limited, however. Philosopher of science Hilary Putnam illustrated this point with his “demon theory.” The theory (really a hypothesis) is this: A demon (maybe Descartes’s) will appear before your eyes if you put a flour bag on your head and rap a table 16 times in quick succession. Now, of course, this is stupid, but it
is
a hypothesis and
is
capable of being tested. It can be tested a lot easier than most scientific hypotheses can be tested.
    The above is Demon Theory No. 16. There is a Demon Theory No. 17, which is the same except that there have to be 17 raps, and a Demon Theory No. 18, and a Demon Theory No. 19, etc. There is an infinity of demon theories. Obviously, said Putnam, scientists have to be selective about the theories they test. You could spend your life testing dumb theories and never get anywhere. It is vital that you winnow out the “possibly true” hypotheses from the “not worth bothering with” ones before getting as far as experimentation.
    Unlike Putnam’s demon theory, most hypotheses are motivated by experience. A snowflake falls on your sleeve. It has six sides. A reasonable hypothesis is “All snowflakes have six sides.” But why that and not “All snowflakes that fall on Tuesday have six sides” or “All things have six sides” or “Everything that melts has an even number of sides” or “All hexagonal objects have six sides?” Moreimportant, why do we even think that something about the shape of snowflakes is generalizable? The very fact that there is a word “snowflake” presupposes the common knowledge that there is a class of tiny, cold, white objects that fall from the sky and that may have other properties in common. Without the implicit hint given by the word, one might grope for hypotheses such as “Everything in the class consisting of (this white thing on my sleeve, Queen Victoria, lasagne, and all the beach balls in the Southern Hemisphere) has six sides.”
Anything Confirms Anything
    Bad hypotheses have a way of subverting evidence. An example is the paradox often called “anything confirms anything”—a paradox that probably occurs, in the form of fallacious reasoning, more frequently than any other of those discussed in this book.
    It is reasonable to think that something that confirms a hypothesis will confirm any necessary consequence of that hypothesis. If man is descended from apes, then undeniably Darwin is descended from apes. A fossil that confirms the hypothesis that man is descended from apes must also confirm that Darwin is descended from apes. So far so good.
    Take a compound statement such as “8497 is a prime number and the other side of the moon is flat and Elizabeth the First was crowned on a Tuesday” (this example from Goodman). To test this, you check 8497 for divisors and conclude that is it prime. This discovery confirms the compound statement, and one consequence of the compound statement is that the far side of the moon is flat. The fact that 8497 is prime confirms that the moon is flat!
    Of course, the compound statement could lump together any propositions at all. Replace them with propositions of your own choice and make your own paradox. Anything can be shown to confirm anything else.
    Evidently it is easier to join hypotheses logically than to be sure that there is valid reason for linking them. This link is essential for valid confirmation. Goodman’s sentence is patently a hodgepodge, but it suggests the wide-ranging consequences of any powerful theory. Many proponents of pseudoscience use the “anything confirms anything” argument. To give just