I certainly think working in terms of expected utility is sensible. But the person I was replying to didn't argue in terms of expectations, and it's not obviously mandatory to think in those terms. (For instance, you might assign more or less importance to very low-probability events than expected utility maximization would suggest. If you do that, then in the very very long run you lose; but none of us is likely to live that long, and I don't think it's clear that "if you live for a million years and consistently do X, then you can be Dutch-booked" is much of an argument against doing X.) So I didn't put my argument in terms of expectations, and I don't think it depends on using expectations.
