The Ramsey Test

How do we decide whether or not to accept an indicative conditional? Frank Ramsey suggested the following:

"If two people are arguing 'if A will C?' and both are in doubt as to A, they are adding A hypothetically to their stock of knowledge and arguing on that basis about C.... We can say they are fixing their degrees of belief in C given A." (Ramsey 1929:143)

There are several strategies for extending Ramsey's suggestion into a more precise theory of conditionals.

Ramsey Test as Minimal Belief Adjustment while Beliving A

In order to evaluate whether or not you accept A→C, you update your beliefs to accommodate A, and see if you accept C. A→C is assertable if the probability of the consequent on the antecedent is high.

But this can't be quite right. Consider the conditional, "If my boyfriend is cheating on me, I will never know." This conditional may well be fine, but, it will come out false on the aforementioned analysis. If you accept the antecedent of this conditional, the probability of the consequent is going to be very low indeed.

Ramsey Test as Minimal Belief Adjustment while A

A suggestion discussed by Jonathan Bennett goes as follows. In order to evaluate A→C, I should take the set of probabilities that constitutes my current belief system and add to it a probability of 1 for A. Then, I should allow this addition to update the rest of my system in the most natural, conservative way. Finally, I should assess whether this updating results in my assigning a high probability to C. On Bennett's view, subjecting indicatives to the Ramsey test does not involve supposing that you believe the antecedent, but rather, supposing that the antecedent is true. This reformulation of the test seems better to capture Ramsey's intent.

Most theorists contend that the Ramsey Test is a means of determining whether indicative conditionals are assertable. The same is not true of subjunctive conditionals. To see why, consider the following subjunctive conditional, "If George Bush had not won the 2000 election, then John Roberts would not have been a nominee for the supreme court." This conditional seems perfecly assertable but if fails the Ramsey Test. Assigning the antecedent a probability of 1 does not result in my assigning the consequent a high probability (despite the fact that assigning the antecedent probability of 1 is likely to make substantial changes to my system).