CONCLUSIONS
In this chapter, I have examined two important claims about simplicity. One is that nature is simple, and the other is that simplicity is an epistemic virtue. “Global” arguments for the claim that nature is simple are either nonexistent or are unconvincing.
I favor “non-global” arguments, but these will show only that in certain respects, and to a certain extent, nature is (representable as) simple, and in other respects and to a certain extent nature is (representable as) complex. This can be supported by scientific investigation of parts of nature, not by metaphysical, theological, or even empirical speculations about the totality of nature. Similarly, I reject “global” arguments in favor of the claim that simplicity is an epistemic virtue—arguments that appeal to the historical success of simple theories, or to concepts of evidence, or to assumptions about prior probabilities or likelihoods. If you are not justified in making an inference to your hypothesis from the available empirical evidence, then either you don't have enough evidence or you don't have the right kind of evidence. What you need is not simplicity but more or better evidence. To be sure, when you don't have sufficiently good evidence you can always invoke the simplicity of the hypothesis. This may cause you and others to come to believe the hypothesis. (To misquote Samuel Johnson, “Simplicity [as an epistemic defense] is the last refuge of a scoundrel.”) But then you still don't have sufficiently good evidence for your belief; you just have a simple hypothesis, which you may regard as a good thing for other reasons. On the other hand, if you are justified in making an inference to your hypothesis from the available evidence, then you don't need simplicity at all for this purpose.You may reply that simplicity does carry at least a little epistemic weight, especially in those cases in which you have some empirical evidence that is almost but not quite enough to make you justified in believing the hypothesis. An appeal to simplicity will just be enough to bring you over the hump epistemically.
Suppose that a drug company has tested a drug for reducing certain symptoms. '1 he company has tested it on groups of patients with those symptoms and so far has had good results. But suppose the groups tested are almost but not quite large or varied enough to justify the inference to the conclusion that the drug is effective, although the preliminary results are promising. Now, suppose also that the hypothesis that the drug is effective in reducing symptoms is simpler than competing hypotheses being considered—e.g., that it is not the drug that is producing the reduction in symptoms but a set of other things, any one of which could reduce the symptoms and at least one of which was operating in each of the cases in which the patient had reduced symptoms. The drug company then invokes the simplicity of its hypothesis in its report. If the groups tested are almost but not quite large or varied enough to justify the inference to the hypothesis that the drug is effective, should an appeal to simplicity do the job? The FDA would, I hope, reject this idea. What the drug company needs in order to show the effectiveness of the drug is to produce more tests—on larger and more varied patients. Simplicity here should carry no epistemic weight. Only empirical evidence—results of testing—will.My reply is similar in the case of the weaker Bayesian idea that simplicity increases the (epistemic) probability of a hypothesis and thus counts as Bayesian B-evidence for it. Suppose, as in the example above, that the groups tested are almost but not quite large or varied enough to justify the inference to the conclusion that the drug is effective. The FDA demands stronger evidence for approval. If the drug company replies that the simplicity of the hypothesis makes the evidence stronger, the FDA should reject that response. Again, what is wanted is more testing.