Hallazgos preliminares

Reformulating the justification game to conform to this format, it would run as follows:

Claimant: I see a zebra.

Skeptic: People sometimes dream that they see zebras. Claimant: It is as reasonable for me to accept both that people sometimes dream that they see zebras and that I am not dreaming as to accept the former alone. (My information that I am not dreaming is trustworthy and the information that people sometimes dream that they see zebras is misleading with respect to the question of whether I see a zebra. Conjoining that I am not dreaming produces a result that does not compete with my claim that I see a zebra.) A claimant wins a round in the justification game just in case she can beat or neutralize the competitor produced by the skeptic. A claimant wins the justification game, showing that she is justified in accepting that p, just in case she wins every round in the justification game starting with the claim that p. We may put this more formally in terms of a general definition of justification as follows:

S is justified in accepting that p at t if and only if everything that competes with p for S on X at t is either beaten or neutralized for S on X at t.

Personal justification based on the acceptance system of a person may then be defined as follows:

S is personally justified in accepting that p at t if and only if everything that competes with p for S on the basis of the acceptance system of S at t is beaten or neutralized on the basis of the acceptance system of S at t.

This completes our account of personal justification.

-126-Reasonableness and Probability

Coherence and personal justification have been defined in terms of reasonableness and acceptance directed at obtaining truth and avoiding error. Reasonableness has been taken as primitive. This has two constructive advantages. First of all, we acknowledge the normative aspect of justification. One ought not to accept something, if one is epistemically rational, when it is more reasonable to accept its denial. At the same time, we leave open the question of whether this notion of reasonableness is reducible to some naturalistic or nonnormative conception. Second, we allow for a plurality of factors to influence the normative evaluation. We have left it open what considerations might make it more reasonable to accept one thing rather than another in the quest for truth on the basis of one's acceptance system. We leave open the question of what sort of factors are relevant to obtaining truth and avoiding error. Most of the defenders of the coherence theory, Quine, Sellars, Harman, Aune, Rosenberg, and Bonjour have, in one way or another proposed that multiple factors determine whether a belief

coheres with some system. ³ They have, however, differed among themselves as to what factors are relevant.

Is some naturalistic reduction of reasonableness possible? The simplest reduction would be to equate reasonableness with probability. For such an account to be naturalistic, we would have to be sure that the notion of probability was itself free from normative definition. This condition is not satisfied in those notions of probability which impose normative constraints on the assignment of probabilities. The equation of reasonableness with probability fails for other reasons, however. One needs to consider more than probability to decide whether it is more reasonable for a person to accept one thing than another. Since the equation of reasonableness with probability has so much intuitive lure and traditional backing, it is worth considering briefly why it should be rejected.

The basic reason for rejecting the equation of probability and reasonableness is that probability is only one factor relevant to deciding what is reasonable to accept in the interests of obtaining truth and avoiding error. One can see this from a simple

example. Compare the following two claims:

It looks to me as though there is a computer in front of me. There is a computer in front of me.

How would one compare the reasonableness of accepting each of these statements with the objective of obtaining truth and avoiding error? The first statement is less risky, but it tells us less. The second statement is

-127-a bit more risky, but it is more inform-127-ative. As -127-a result, we c-127-an s-127-ay th-127-at the risk of error is greater in accepting the second than the first, but the gain in obtaining truth is greater in accepting the second than the first. This example reveals that the objectives of obtaining truth and avoiding error are distinct and may pull in opposite directions.

The more informative a statement is, the more it tells us about the world, the greater our gain in accepting it if it is true and the greater our risk of error. The probability of a statement tells us what our risk of error is, but it tells us nothing about how much

we gain in accepting it when it is true.

Major scientific claims, those concerning galaxies, genes, and electrons, for example, though among the most important things we accept and claim to know, are less probable than either of the cautious claims articulated above. The reasonableness of accepting such claims is influenced by our interest in accepting claims which, if true, are important general truths about ourselves and our universe. Given the past history of scientific claims, even those put forth by scientists of great genius, we must

concede that the risk of error, the probability that these claims are false, is far from negligible. We can easily see that the interest in accepting what is true and the interest in avoiding accepting what is false may pull in opposite directions by

considering the results of aiming at one to the exclusion of the other. If a person were only interested in avoiding error and indifferent to accepting truths, total success could be attained by accepting nothing at all. If, on the other hand, a person were only

interested in accepting everything that is true and indifferent to accepting falsehoods, total success could be attained by accepting everything. The problem is to accept what is true while at the same time seeking to avoid error.

Reasonableness and Expected Utility

The foregoing ideas can be summarized in a simple mathematical representation.

Suppose that we could specify what value or, more technically, the positive utility we assign to accepting some specific hypothesis, h, if h is true, and represent that by 'Ut(h)', and, similarly, the negative utility we assign to accepting h, if h is false, and represent that by 'Uf(h)'. Now, if we ask ourselves how reasonable it is to accept h in the interests of accepting what is true and avoiding accepting what is false, we must take into account the values of both Ut (h) and Uf(h). There are just two outcomes of accepting h, that h is true and that h is false, and we must take into account what value we attach to each of those outcomes. There is, however, another factor to consider, namely, how probable each of those outcomes is. So, if we let 'r(h)' represent

-128-the degree of reasonableness of accepting h, and 'p(h)' represent -128-the probability of h being true and 'p (-h)' as the probability of h being false, we obtain the following formula for computing the reasonableness of accepting h:

r(h) = p(h)Ut(h) + p(-h)Uf(h). ⁴

The reasonableness of accepting h, r(h), the sum of two products, may be called the epistemic expected utility of accepting h. It is equal to the sum of one's positive expectation of accepting h when h is true and one's negative expectation of accepting h when h is false.

This formula clarifies precisely the consequences of identifying probability and

reasonableness. To do so is mathematically equivalent to assigning a value of one to Ut(h) and a value of zero to Uf(h), no matter what h is. In that case, no matter what claim h might be, r(h) = p(h). It is absurd, however, to regard all truths as equally worth accepting. Some truths are more valuable than others measured strictly in

terms of obtaining truth about the world because, as we have noted, some truths tell us much more about the world than others. Therefore, the equation of reasonableness and probability must be rejected. Its rejection does not, of course, mean that

probability is irrelevant to reasonableness. On the contrary, we can see that if the utilities of accepting two claims are the same, then the comparative reasonableness of accepting one in comparison to the other will be determined solely by the probabilities.

This is more important than it might at first seem, because some competing claims satisfy the constraint of having the same utilities.