This is something that probably feels obvious, but has a big impact on my thought process.

Calculating a utility function

It's often helpful to calculate whether something has more benefits that costs. Call it a utility function, cost-benefit analysis, or whatever you want. A guy offers me ten dollars for my umbrella that I'm currently using to stay dry. I ask myself, how valuable is ten dollars, and how valuable is it to be dry right now? If my utility function comes out positive; if the benefits outweigh the costs... I take that deal, right?

The flaw in that logic

The issue is that "net positive utility" is rarely anyone's goal. Game theory is constantly trying to figure out the right way to make decisions. FDT and UDT seem like the frontrunners right now. One thing they both agree on is that an agents goal should be utility maximization. Not positive utility, maximizing utility. This means we must consider all the alternatives and decide which is maximal. In our umbrella example from before: A guy offers me ten dollars for my umbrella I'm using. It's not enough to think that's a good deal, you need to be asking if there is a better deal. It's there anyone offering 15 dollars for the umbrella? Could we share for 5$?

Works for negative utilities too

Kevin lost his job and has to pay rent today. He has a watch worth twice his rent money, he takes it to a jewelry store that says they will pay 1x rent for cash today. It's not a good deal. In fact, it is negative utility. But if out of all the choices Kevin has, all of them have negative utility, and the watch is his best plan, it is actually a good choice to take the deal.

Even if a choice has negative utility, if it is better than all the alternatives it is the right choice. It is still the maximal utility.

Judging others decisions

Now we get to my point. Have you ever heard of a senator making a bad choice? Have you looked at a decision your boss made and said, this is a move in the wrong direction! Your neighbor or friend seems to be making bad a negative utility decision. It's very possible they are acting perfectly rationally and making the best decision available to them.

People might be more rational than it seems.

It is not enough to be able to calculate the net utility function. Even using probabilities and lots of effort; at the end of the day a utility calculation means nothing without the context of all the alternative utility calculations.

Not many people have the full context of decisions being made by others. Therefore not many people can figure out whether a person is acting rationally.

In an older post of mine, I try to sort out whether it is ever right to pass judgment on others. This is the biggest argument I have against judgement. I try to remember this any time my thoughts seem to get judgemental.

New Comment
4 comments, sorted by Click to highlight new comments since:

Utility, as constructed by such things as the VNM theorem, is only fixed up to an affine transformation. Thus utility scores are not absolute, and neither are differences in utility, but only ratios of differences of utility.

Some conceptions of utility are purely ordinal, not cardinal - consistent preferences expressed in choices (do X instead of Y) can be modeled as an ordering function, without numeric values.  Assignment of numeric values to utility is purely a modeling choice, and can lead to incorrect assumptions, like addititivity (which is only valid if choices are fully independent).

The idea of "negative utility" is part of this extension, which is less-well-supported and may or may not make sense in a formal rational analysis.  It certainly makes sense in a humanist psychological sense, but combining the two is ... difficult.

I think this is a valid point, although some people might be taking issue with the title. There's a question about how one should choose actions, and in this case, the utility is relative to that of other actions as you point out (a action sounds good until you see that another action has ). And then there's a philosophical question about whether or not utility corresponds to an absolute measure, which runs orthogonal to the post.