Hacker News Reader

Comment by darkmighty | Hacker News Reader

I don't get what people have against utility theory (if that's what you mean by utilitarianism). For me it's one of the few models that must be obeyed: there must be a way to create a partial order among possible choices.

Whatever the reasoning behind your actions is, you are going to choose a single action (at each step) -- this forces you to choose the best. That's all utilitarianism is for me -- it tells you nothing about how to construct the utility function. Now the utility this view provides is just that: you can see clearly what sorts of thing you give greater priority, so you'll make the best possible judgement.

View on HN

Utilitarianism <: utility theory (that is, a subtyping relationship). Utilitarianism of the Peter Singer kind is specifically choosing a utility function that consists of a linear, nonweighted sum of one-dimensional happiness/pleasure measurements across the entire animal population of the world, minus a linear, nonweighted sum of one-dimensional misery/pain measurements across the entire animal population of the world.

The inherent problems are the usual objections to hedonic total utilitarianism: the Repugnant Conclusion, wireheading and experience machines, the fact that emotions and sensations just aren't linear real-valued variables, the difficulty in measuring them as such even if they were, and of course the Repugnant Conclusion.

Oh, and the fact that Singer has long-since given up on trying to prove hedonic total utilitarianism normatively correct, and thus simply begins his books by saying, "Well, I can't prove it to you, but let me appeal to your emotions and intuitions to hammer it into you, and then proceed on with my arguments about what I demand from you."

Which is why it's really quite disappointing to see people fawning over Singer's utilitarianism: it's unproven, it's unreasonable on its face, it engages in rather flagrant neoliberal apologia (ie: charity rather than social reform or revolution), and it thus adds up to a series of bad ideas wearing nice suits.

> Whatever the reasoning behind your actions is, you are going to choose a single action (at each step) -- this forces you to choose the best.

This works at the level of individual choice, but it doesn't add up over multiple individuals, because you can't compare different people's utilities on the same scale. But the key claims of utilitarianism under discussion (the ones that are supposed to motivate people to donate to charity instead of purchasing luxuries for themselves) require adding up utilities over multiple individuals. Your argument doesn't justify utilitarianism in that sense.

I agree that you can't necessarily add up utilities over multiple individuals, but you can naturally claim that a "good" individual utility takes into account the well being of others, and if the utility of each action is unique you may say there's a composing function g such that your utility u(a=action,t1=actor 1) is also a function of the utility for other people u(a,t1)=f(a,g(u(a,t2),...,u(a,tN))).

I also think utility can makes sense when there's a set of actors looking forward to make a set of decisions. Again, this set of decisions is going to be unique. So if those actors are able to 1) achieve consensus on a utility 2) coordinate their actions according to said utility, then it also makes sense to talk in terms of a joint utility theory.

> you can naturally claim that a "good" individual utility takes into account the well being of others

That's not the same as utilities of multiple individuals adding, or otherwise being composed into a single utility function. The "good" individual's utility function can certainly depend on that individual's beliefs about how his actions will affect the well-being of others; but it can't depend on the actual utility of those others, because the "good" individual doesn't know the actual utility to others of various states of the world.

> if those actors are able to 1) achieve consensus on a utility 2) coordinate their actions according to said utility, then it also makes sense to talk in terms of a joint utility theory.

I agree with this, but I think the necessary conditions for it are very rare. In most cases of collective action, the actors do not have consensus on a utility function and do not coordinate their actions according to a single utility function.

In fact, in most cases the actors probably do not even share a common or compatible set of beliefs about factual things. In US politics, for example, you have members of Congress who are young earth creationists alongside other members who believe standard evolutionary theory.

The problem is there is no utility function and choices are path dependent.

-Path dependence has no relevance. You are still going to make a single choice at each event no matter if the choices depend on previous history or not.

-"there is no utility function" is an odd claim. A function can obviously be defined, so what's the objection? Note that in any case there is an implicit utility function for which each action you do take has the maximum utility. Of course, there's the recursive problem of choosing the "best" utility function, which just highlights that at some point we stumble upon certain fundamental principles of physics and ethics which are ultimately arbitrary in which we would guide our actions consistently.

The recursive problem is the problem. If the utility function is arbitrary, then utilitarianism is just a tautology.

Of course it's not a tautology. There's no boundless recursivity -- you stop at a point wherein you define your utility function.

What I mean by the utility function being arbitrary is there is no way to derive it from within utility theory, or within mathematics for that matter. In this sense the laws of physics are arbitrary; what you define as good principles of utility are also arbitrary, but that doesn't make them useless -- in fact the stochastic version of the theory is widely applied in decision making and artificial intelligence. It let's you scrutinize the basic principles and act consistently based on that. Not unlike other mathematical tools, it's a way to find the best course of action given a small set of assumptions, and not more than that (i.e. doesn't provide a way to find the assumptions, because it's impossible).

Ultimately we're just certain animals of a species, and things like happiness, ethics, morality, are the outcome of our traits and can't be defined in a vacuum.

As noted, there's also the equality interesting reverse utility theory -- that is, finding utility functions that approximately match our choices. It's just as useful, it shows clearly how we instinctively prioritize things like money, time, feelings; I think it's all quite valuable analysis.

>A function can obviously be defined, so what's the objection?

Ok, so list its definition in a programming language of your choice.

Utility theory applies when we define "utility" as a linear function of some number of real-valued variables inside a model of the world. For instance, if we say "utility world = sumOfMoney world", then we've defined a VNM utility function.

This doesn't mean that total hedonic utilitarianism possesses any such function, or that we've neatly uncovered any such function which accurately describes the desires of real people. It just means someone wrote down a bit of math that works well for trivial cases.

Just because the trivial functions are obviously wrong that doesn't mean good non-trivial functions don't exist. I also don't see why you're stressing on linearity, there's nothing in general utility theory that demands linearity (in particular even in VNM utility).

So here's one utility function (for a company, for example): "Utility = Capital ^ (2/3)". It's one valid utility function (even in accordance to VNM theorem), which is why I don't understand the saying "there is no utility function".

If you read what I wrote in response to gress below, I of course have no hope that utility theory would unveil a good utility function in itself; indeed you cannot do it within mathematics exclusively. I just think it's a useful tool that exposes whatever fundamental motivations you have with clarity. Which again is why I don't see why criticizing utilitarianism is a thing at all.

>Which again is why I don't see why criticizing utilitarianism is a thing at all.

Because utilitarianism is doing utility-theory consequentialism with the utility function strictly defined to be summed-up quantities of pleasure across the whole population. Utilitarianism doesn't mean "consequentialist, VNM-rational decision making", it means, "the maximally effective way of creating and enforcing pleasure for everyone". It's not the utility function anyone really wants, which is why utility theory is such crap. It doesn't matter to have a coherent mathematical formalism if you get given not the slightest idea how to come up with a good value for the infinite-dimensional free parameter.

Reply on news.ycombinator.com