You're looking at Less Wrong's discussion board. This includes all posts, including those that haven't been promoted to the front page yet. For more information, see About Less Wrong.

AlexMennen comments on Median utility rather than mean? - Less Wrong Discussion

6 Post author: Stuart_Armstrong 08 September 2015 04:35PM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Comments (86)

You are viewing a single comment's thread. Show more comments above.

Comment author: V_V 09 September 2015 12:47:23PM 0 points [-]

The independence axiom derives most of it intuitive strength from the fact that if you violate it, you can be money pumped when presented with a sequence of decisions.

If your preferences aren't transitive, then even your one-shot decision making system is completely broken, since it can't even yield an action that is "preferred" in a meaningful sense. Vulnerability to money pumping would be the last of your concerns in this case.

Money pumping is an issue in sequential decision making with time-discounting and/or time horizons: any method to aggregate future utilities other than exponential discounting ( * ) over an infinite time horizon yields dynamic inconsistency which could, in principle, be exploited for money pumping.

The intuitive justification for the independence axiom is the following:

  • What would you like for dessert, sir? Ice cream or cake?
  • Ice cream.
  • Oh sorry, I forgot! We also have fruit.
  • Then cake.

This decision making example looks intuitively irrational. If you prefer ice cream to cake when they are the only two alternatives, then why would you prefer cake to ice cream when a third, inferior, alternative is included? The independence axiom formalizes this intuition about rational behavior.

( * with no discounting being a special case of exponential discounting)

Comment author: AlexMennen 11 September 2015 01:16:00AM *  3 points [-]

If you prefer ice cream to cake when they are the only two alternatives, then why would you prefer cake to ice cream when a third, inferior, alternative is included?

You're thinking of a different meaning of "independence". A violation of the independence axiom of VNM would look more like this:

  • What would you like for dessert, sir? Ice cream or cake?
  • Ice cream.
  • Oh sorry, I forgot! There is a 50% chance that we are out of both ice cream and cake (I know we have either both or neither). But I'll go check, and if we're not out of dessert, I'll get you your ice cream.
  • Oh, in that case I'll have cake instead.
Comment author: V_V 11 September 2015 10:39:05AM -1 points [-]

Yes, I believe that this is a stronger version. Median utility satisfies the weaker version of the axiom but not the stronger one.