“If in every event which can possibly occur the consequence of action I is not preferred to that of action II, and if in some possible event the consequence of II is preferred to that of I, then any sane preferer would prefer II to I.”
This sentence, in a letter from Savage in 1950, finally persuaded Samuelson that rational choices must obey expected utility. He had been skeptical – thinking that expected utility was just a simple approximation, like exponential discounting or like separability. Marschack and Savage and Friedman all wrote letters trying to persuade him, and though they were right, they kept using bad arguments, and Samuelson disposed of them.
Savage and Friedman wrote a paper saying that, because people buy both insurance and lottery tickets, expected utility implies that the utility-of-money must be concave then convex. Samuelson saw that this was ridiculous, & said “there’s as much to be learned about gambling from Dostoyevsky as from Pascal.”
But the sentence from the letter above finally persuaded Samuelson.
This is all from Ivan Moscati’s “How Economists Came to Accept Expected Utility Theory”.