Tuesday, August 16, 2016

Problems of Utilitarianism #3: Hypothetical Happiness Matters, Unless It Doesn't

I spent the previous post about problems of utilitarianism critiquing the repugnant conclusion, by saying that what counts is utility as experienced by individuals, not some big abstract pile of hypothetical total utility points separate from the people that experience them. This explains why it's better to have some happiness-experiencing people than none at all, but not why and not why we shouldn't feed the utility monster.

If we define "bad" as decreasing utility - then quickly and humanely killing someone in their sleep would be acceptable, because they don't suffer. But in fact this isn't what our moral intuitions lead us to do. We endorse ending the lives of living things only when there is no possibility of future positive experience, future experience that would make life worth living. Hence a patient with ALS looking to the future may choose assisted suicide, but we don't sneak up to the hospital bed of someone suffering from pneumonia and give them a quick overdose of fentanyl.

The problem is that not just the pile of utility points, but also all future positive experience is also an abstraction. In fact almost all utility is abstract, i.e. not being currently experienced. We don't just react to immediate pleasure or pain we're experiencing this millisecond, we're also setting up to avoid future pain and gain future pleasure. However, abstract though it is, again it will be experienced non-abstractly by an individual.

One might argue that to make others' suffering less abstract the answer is to increase empathy so you're motivated to help them and everyone's utility increases; or, to decrease empathy so you can't be haunted by others' suffering. It is possible that from a population perspective one of these strategies is more stable and self-perpetuating than another, and this could be modeled experimentally. Whether this will helpfully apply to a natural evolved nervous system (even a social one) is another question, as these nervous systems evolved as machines subservient to the mission of spreading their genes, and if suffering achieves that, then that's what will be employed. When the nervous systems rebel and start thinking the universe is about them is when the story gets confused, which may be why these inconsistencies in reasoning about morality and happiness continue to appear.

Problems of Utilitarianism #2: Parfit and Rawls Are Incompatible

Previous problem of utilitarianism here. Next problem of utilitarianism here.


The Rawlsian conception of a just society is incompatible with Parfit's extension of utilitarianism.

Rawls claimed that a just society was necessarily a very egalitarian one. His argument was that if you were going to be placed into a society without knowing ahead of time what your role would be, if you're smart, you would want a society where there's not much difference between the guy at the top, and the guy at the bottom. That is to say: sure it would be a blast to be a plantation-owner in the antebellum American South, but if you fell out of the sky at random into a role in that society, chances are much greater you'd end up as a slave or tenant farmer breaking your back for one of the plantation owners.(1)

Parfit extended utilitarianism by saying that if we want the greatest good for the greatest number, we should want not just more happiness, but more people. The equation is average happiness of each person * # people = total amount of happiness. In this view, having more people to be experiencing some happiness can even counterbalance the amount of happiness that each person is experiencing. Another way of saying this: if utilitarianism is the greatest good for the greatest number, don't neglect the "number" part.

The full elaboration of this claim runs counter to most people's moral intuitions and lead to what's known as the repugnant conclusion (summarized below).

Imagine two societies: a society of a million people who have the best lives possible, whose lives are 99% worth living. (I don't know, sometimes it's cloudy when they go to the beach, otherwise life is perfect.) Compare that to a society of a hundred million whose lives are only 1% better than death: they groan each day under the oppressive weight of a dictatorship, but sometimes see a nice flower, which keeps them from wanting to kill themselves.(2) Because 99% * a million is less overall happiness than 1% * a hundred million, the repugnant conclusion according to Parfit's interpretation of utilitarianism, is that it's better to have the much bigger, much less happy society.

The obvious rejection is that an individual experiences individual happiness - total happiness is not something that is experienced - and the individual experience of objective happiness is what matters. Of course, if you make that claim, you're arguing against utilitarianism.

To illustrate Parfit's repugnant conclusion concretely in contrast to Rawls, let's apply it to a real historical situation, the concrete example of black slavery in the United States. Of course the QALY (quality-adjusted life years) measurements for utility will necessarily be a little fudged. On the eve of the American Civil War in 1860, the census showed 3,953,761 slaves in the United States. Let's round that up to four million and assume these people had lives 1% worth living(3) (after all they're literally in the horrible dictatorship that I described above.) [Added later: the very next day after I wrote this post, I ran across Robin Hanson's blog post "Power Corrupts, Slavery Edition" which contains the statement "US south slave plantations were quite literally small totalitarian governments".] Now let's compare that to Avalon on the California island of Catalina. Ever been there? It's really nice, as you might expect, and has a population of just under 4,000, and while it's not completely egalitarian, you can't be bought or killed with impunity. It's a really nice place, so let's assume there's 99% average happiness. Parfit concludes that it's better to have that slave society than modern Avalon.

By Parfit's interpretation of utilitarianism, the problem is not the institution of slavery's impact on quality of life, as long as we can overcome this by having enough slaves. Rawls could never recommend choosing a slave society over a non-slave society ("well how big a slave society is it?" the repugnant conclusion says you should ask.) By Rawls (and most of our intuitions) the answer of which you would rather be randomly thrown into is obvious, and wholly contingent on whether moral value comes from some abstract total register of utility points, or the experienced utility of an individual human being. Since policy makers do these calculations to make decisions, this absurd conclusion could conceivably make a difference, and some respected thinkers (Bryan Caplan and Michael Huemer among them) have argued that our intuitions are wrong.

Of course the counterargument is: if individually experienced utility is all that matters, isn't it better to have one really happy person then two ho-hum people? Shouldn't we feed the utility monster then? I don't know, other than to say fatalistically, that possibly moral reasoning either is not a real process, and that we are unable to make decisions like this about groups of people that we do not know. Which would be terrible, considering that modern societies are forced to do so all the time. But it would be consistent with Adam Smith's thought experiment about losing a single joint of a finger versus an earthquake in China that kills a hundred thousand. Humans cannot reason about abstract people as moral agents, because we did not evolve with a need to do so - other than as threats or trade partners.

NOTES

1. Rawls also suffers from the problem of differing agents: assume that someone doesn't care about relative status, only absolute comforts. If such a person gets his head frozen and wakes up in a future where there are absolute un-displaceable overlords but who give them amazing experiences and material comfort, that person might not care, even though you would chafe under such an uber nanny-state regime. I also wonder how meaningful the question of a choice can be, because there is no neutral position to choose from and all are habituated to the specifics of times and places. I.e., to me England appears a nightmarish dystopia but the people I've met from there seem to be reasonable people who enjoy their lives and even return there voluntarily, so who knows.

2. If you think assigning numbers to such situations is spurious and academic, I'm afraid I must inform you that they are very concrete and very real-world, as health systems use units of DALYs and QALYs all the time to make decisions. And some systems do assign negative values, meaning that some conditions are considered to make life not worth living, i.e. they are literally worse than death.

3. I tried to look up the suicide rate for slaves, as this would give an idea of how many slaves thought their life was not worth living. Although I couldn't find numbers, apparently suicide was unexpectedly rare, and the threat of execution by owners would not have been an effective deterrent for slaves who thought continued slavery was worse than death. In several places (e.g. here) I saw an article referenced: David Lester, Center for the Study of Suicide, "Suicidal Behavior in African-American Slaves," Omega: Journal of Death and Dying, 37:1 (1998), 1-13.

Problems of Utilitarianism #1: Real Utility Monsters

Next problem of utilitarianism here.


Utilitarianism is often formulated as the claim: "The best society is the one with the greatest good for the greatest number."

There are many problems with this, first and foremost is that such an abstract formulation submerges the question of how to achieve and maintain this. To make this concrete, it doesn't even distinguish between radical capitalism and radical communism.

But another problem troubles it, one which crops up in multiple places in reasoning about moral societies: the problem of differing agency. Many of us understand on some level how troubling this is to the Enlightenment project of organizing societies, and this is evident in our discomfort discussing (for instance) behavioral genetics.

Here is an innocuous case of differing agency - one might call it trivially differing agency - that is not problematic for utilitarianism: I kind of like chicken. But my wife really likes chicken. If we get to the end of the meal and there's one piece of chicken left, the obvious best choice is to give it to her, because there will be more happiness in the world if she eats it. In the same way, I once refused a free ticket to a PGA tour event because I can't stand golf, and it is almost certainly true that whoever got that ticket instead of me, they enjoyed the event more. My taking up a spot at such an event would be an anti-utilitarian travesty.

Differing agency remains innocuous only so long as agents differ somewhat randomly in their specific tastes but not on average in the intensity of their pleasure and suffering. To illustrate this problem, Robert Nozick imagined a utility monster, that would always derive more enjoyment from everything. It doesn't habituate, it has no hedonic treadmill. You could imagine the utility monster as some kind of hedonistic superintelligent alien that had come to Earth to experience chocolate ice cream and massages, and experience them it does, on wondrous levels of ecstasy we can't begin to imagine. To it, we are as dim beasts, barely able to register pleasure compared to the raptures that the monster can attain. If we are true utilitarians, we always have to give our chicken and golf tickets (and chocolate ice cream and massages) to the utility monster. (Let's assume it's a nice utility monster that doesn't destroy things like the one below, it just likes the things we like, more than we like them, which is still a big problem.)



Not exactly how Nozick imagined it, but hey it's funny.
From Existential Comics


And as it turns out, this is not a thought experiment, because humans actually do differ, both in their capacity to experience pleasure, and the damage done by negative stimuli. People in the throes of a manic episode take great hedonic value from a great many things, including money and sex, which is partly why such episodes are psychiatric emergencies. Do we feel obligated to help them continue spending sprees or accept their propositions? People with borderline personality disorder are much more badly harmed by social rejection than the rest of us; do we feel obligated to constantly reassure them that we are their best friends, to the exclusion of other people who are healthier in this regard? You might argue that over time the greatest good is not to help them make worse decisions that will surely harm them in the long run. But there are certainly people whose happiness set points are constitutionally lower or more fragile (anti-utility monsters), and outside of mental health professionals there are very few of us who see a moral obligation in continually propping up their current hedonic state.

My gut reaction is that we don't have such an obligation, but I can't see why we shouldn't, if utilitarianism is correct.

Of note, Nozick also critiqued the Rawlsian conception of a just society, but there is a further critique of Nozick in the instantiation of societies of humans, which again relies on the actual differing capacities of humans that affect the quality of their agency. And despite heroic efforts to create equal agents, humans continue to stratify themselves based on these differing qualities.