Anthony DiGiovanni 🔸 comments on Consequentialist Foundations for Expected Utility

[Anthony DiGiovanni 🔸]

Thanks Vasco. I think it would help a lot if you spelled out the premises more, because they’re quite opaque to me as written. E.g. I don’t know what “the norm reveals a choice function” means. (I think this kind of use of jargon without giving context is a common failure mode of current LLM summaries.)

Also, if I understand correctly, “behaviour maximizes a family of preference orderings...” means that the result only shows that we can represent an agent’s behavior as satisfying completeness. But my unawareness argument isn’t about what our behavior can be represented as. The question is: When we’re comparing our options when making decisions in the first place, should we have complete preferences? Cf “Winning isn’t enough”:

But what these arguments really show is that you are disposed to playing a dominated strategy if we cannot model your behavior as if you were a Bayesian with a certain prior and utility function. They don’t say anything about the procedure by which you need to make your decisions. I.e., they don’t say that you have to write down precise probabilities, utilities, and make decisions by solving for the Bayes-optimal policy for those.

(But let me know if I’ve misunderstood the result.)

[Vasco Grilo🔸]

Thanks, Anthony.

I think it would help a lot if you spelled out the premises more, because they’re quite opaque to me as written. E.g. I don’t know what “the norm reveals a choice function” means. (I think this kind of use of jargon without giving context is a common failure mode of current LLM summaries.)

I asked Claude to update the post to address your comment. There is now a section with the setup of the theorems, and clearer premises. Are they sufficiently understandable now?

Also, if I understand correctly, “behaviour maximizes a family of preference orderings...” means that the result only shows that we can represent an agent’s behavior as satisfying completeness.

I only briefly skimmed the article, but I agree with your interpretation. Claude agrees too.

The completeness here is a property of a ranking reconstructed from choices, and it comes almost entirely from premise 0: because the rule always names something acceptable, every pair gets settled. [This is now clarified in the section of the linkpost with the setup.] But the rule may name both options as acceptable, which the theorem records as the two being equally good — and that is also exactly how an agent who found them incomparable, and picked arbitrarily, would behave. The result therefore cannot distinguish “equally good” from “not comparable”, and so does not show that an agent deliberating about what to do must arrive at a complete ranking. It shows their behaviour is representable as if they had one.