Nozick devotes the first subsection of Chapter 9 (pp. 277-279) to analogical reasoning and to what I like to call “the mental contamination problem.” According to Nozick, one of the problems with reasoning by analogy is our inability to keep an open and uncontaminated mind. Why does this problem occur? Because whenever we are presented with a hypothetical example or with a body of data, we will invariably be tempted to evaluate that example or data through the filter of our theoretical priors, or as Nozick puts it (p. 277, emphasis in original):
“Suppose that you are trying to convince me to change my evaluation of a case [via analogical reasoning]. If your parallel example is not close, I can accept your evaluation of it while maintaining my original evaluation of the case in question. The closer the parallel example, the more prone I will be to see it through the filter of my original evaluation. (‘That’s not so bad after all, for it’s just like . . . .’)”
There are, of course, other ways of reasoning (see table below, left), but Nozick’s focus here is on analogical reasoning (and on the problem of mental contamination) because that is the method he will use in the remainder of Chapter 9. As an added bonus, however, Nozick explains why deductive reasoning is also so rarely able to change minds. Simply put, the main problem with deductive methods of reasoning is that one can always reject the premises on which a deductive argument is based. (For my part, I suspect it is this ability of philosophers to reject each others’ premises at will that explains why there is so little progress, if any, in the field of philosophy. See generally David J. Chalmers’s excellent essay “Why isn’t there more progress in philosophy?,” available here.)
So, how does Nozick solve the mental contamination problem? He doesn’t. Instead, he merely exhorts his readers ahead of time to evaluate each of his hypothetical examples (which he will present in the next subsection of Ch. 9) on its own individual merits. In other words, although Nozick doesn’t use Bayesian terminology, he is effectively begging us to update our priors each time we are presented with a parallel example! Or in plain English: don’t let your evaluation of the first example in a chain of reasoning be contaminated or otherwise unduly influenced by the conclusion you know that is coming at the end of the chain. It remains to be seen, however, whether this exhortation will prove effective. In any case, since the next subsection (pp. 280-292) is one of the longest subsections in the entire book, we will review it carefully over the next few days.