Caledonian, if utility is any function defined on amounts of money, then if you are maximizing expected utility, you cannot fall prey to the Allais paradox. You can define a utility function on gambles that is not the expected value of a utility function on amounts of money, but then that function is not expected utility, and you’re outside of normal models of risk aversion, and you’re violating rationality axioms like the one Eliezer gave in the OP.
Caledonian, if utility is any function defined on amounts of money, then if you are maximizing expected utility, you cannot fall prey to the Allais paradox. You can define a utility function on gambles that is not the expected value of a utility function on amounts of money, but then that function is not expected utility, and you’re outside of normal models of risk aversion, and you’re violating rationality axioms like the one Eliezer gave in the OP.