Here are two ways to positively evaluate something, e.g. a cup of coffee: On the one hand, we might think that the cup of coffee issufficiently goodfor some purpose, e.g. for having something tasty, or for causing us to wake up properly. On the other hand, we might think that the coffee is a betterin some respect than some relevant alternatives. For example, we might think that the cup of coffee in front of us is tastier, or more a more effective stimulant, than any other cup of coffee within a 10 mile radius. These are quite different ways to evaluate a cup of coffee: On the former, it does not matter how the coffee compares to something else – the coffee is evaluated positively just in case it meets some demands we make for cups of coffee. Call this asatisficingevaluation.19 On the latter kind of evaluation, it matters how the coffee compares to something else – the coffee is evaluated positively just in case it is better than any relevant alternative. Call this anoptimizingevaluation.
Note that we can have satisficing and optimizing evaluations of the same thing us- ing the same standards of evaluation: The coffee might be sufficiently good in that it is quitetasty, and also better than any alternative in that it istastier than any alternative. In both cases the evaluation of the coffee depends on tastiness and tastiness alone. So what separates satisficing and optimizing evaluations is not that there are different standards for evaluation. Rather, it is thestructureof the evaluation that is different between satisficing and optimizing evaluations: Satisficing options aregood enough, while optimizing options are better than alternatives. Note also that an option might be satisficing without being
optimizing andvice versa, even relative to the same standards: The cup of coffee might not be the tastiest coffee available, but still be sufficiently tasty. Conversely, the coffee might indeed be the tastiest available, but still not sufficiently tasty.
What has this distinction between satisficing and optimizing evaluations got to do with explanatory acceptability? Well, notice that the Threshold View is a satisficing view of one kind of evaluation, viz. epistemic evaluation of a hypothesis for the purposes explanation. The standard of evaluation isprobability, of course, since it is the probability of a hypoth- esis that must be sufficiently high in order for it to be acceptable on the Threshold View. As I’ve mentioned, one might be tempted to give up on probability-based views altogether given the shortcomings of the Threshold View. But we are now in a position to realize that the Threshold View is not the only possible way of explicating acceptability in terms of probability: We can instead develop anoptimizingview of acceptability using probability as our standard for evaluation.
To see the promise of such an account, consider the following piece of scientific fic- tion. Suppose Darwin is trying to explain the fact that organisms evolve over time, and that there are 100 potential explanations for this fact to which Darwin assigns non-zero prob- abilities. H1 is Darwin’s hypothesis of natural selection, H2 is Lamarck’s hypothesis of inheritance of acquired characteristics, andH3-H100are some other hypotheses that would, if true, explain biological evolution. Now suppose thatH1 andH2 are much more proba- ble for Darwin thanH3-H100. For definiteness, let’s suppose thatp(H1) = p(H2) = 0.4, whereasp(H3) = ...= p(H100) = 0.2/98≈ 0.00204(wherep(·)is Darwin’s probability function). Now, in a nearby possible world, Darwin’s near-identical counterpart, Darwin*, is in almost the same situation. It’s just that for Darwin*, the probabilities of the 100 hy- potheses are very different (because Darwin* has access to other evidence). Let’s suppose that Darwin*’s probabilities are as follows: p*(H1) = .4, p*(H2) = · · · = p*(H100) =
0.6/99≈0.00606(wherep*(·)is Darwin*’s probability function).
the same for Darwin and Darwin*, Darwin and Darwin* differ in that only for Darwin* is H1 is significantly more probable than any other relevant hypothesis. Thus, a natural thought is thatH1is acceptable for Darwin* but not for Darwin. After all, the thought goes, H1 is 66 times as likely to be true as any other hypothesis that would explain biological evolution for Darwin, whereas for Darwin* there is an equally likely rival in Lamarck’s theory of acquired characteristicsH2.20 Now, admittedly, we should not take our intuitions about probabilities too seriously, for as we have seen the sense of “probability” we are employing here is not the layperson’s sense that it’s natural to think is influencing our intuitive judgments. But this suggests a promising line of thought, viz. that a hypothesis is acceptable just in case it is much more probable than any other hypothesis what would explain the same things. The next section spells out a precise account based on this idea, and section 6 shows that the account avoids the double trouble of the Preface and Lottery paradoxes in a satisfying manner.