[content warning: dieting, alcohol]

Scott Alexander has a blog post where he writes in criticism of Caplan’s position that many forms of mental illness are merely very strange preferences.

Caplan, bizarrely, refers to the-thing-that-is-not-preferences as “budgetary constraints”. This is strange partially because referring to people who can’t walk without a wheelchair as having “budgetary constraints” is strange, and partially because some instances of actual budgetary constraints are the result of preferences (e.g. the desire to work fewer hours or a job that helps people). I will instead refer to “impairments”.

Caplan’s position is obviously far too strong. A reasonable human being cannot argue that psychosis, executive functioning issues, and fatigue (among others) are not impairments. However, I think that Alexander goes too far in the other direction by saying that no neurodivergences can be understood as preferences.

Several of the examples Alexander discusses seem to be classic cases of shit everyone concerned with preferences already knows about: terminal vs. instrumental values, meta-preferences, and preference conflict. For instance, Alexander writes:

But in order to fully explain alcoholic behavior, we have to take this theory exceptionally far. Consider a typical alcoholic drinks for several years, then “hits bottom”, goes sober, and joins Alcoholics Anonymous. He attends AA meetings three times a week for three years, then has a really bad day and binges on alcohol. Afterwards he is so embarrassed that he attempts suicide, but is rushed to the hospital and resuscitated successfully. After that he goes back to his AA meetings.

Does this man have a preference for going to AA meetings three times a week for several years then getting really drunk then attempting suicide? That’s a weird preference to have. Does he have a preference to drink, and in order to be socially acceptable he ‘covers up’ his one episode of binge drinking by years of AA meetings and a serious suicide attempt which he secretly knows will fail? That is a pretty disproportionately big web of lies, especially when probably no one would blame him for binge drinking one night one time.

But this is an absurd example. Let us imagine a neurotypical woman who eats an unhealthy diet, then decides she’s going to start eating more healthily and goes to Weight Watchers meetings once a week for several years. In one holiday season, because of the stress of the holidays, family conflict, and the ready availability of tasty snacks, she breaks her diet and eats perhaps twice the calories she had previously been allotting herself for a few weeks.

We do not say that we can’t understand what this woman’s preferences are. It is perfectly obvious that she wants lots of fatty desserts, but that she also wants to eat a healthy diet. In fact, this is a very common kind of preference conflict– the conflict when something has short-term positive consequences and long-term negative consequences– seen in everything from trying to stick with an exercise routine to studying for tests. Similarly, the alcoholic’s drinking has short-term positive consequences and long-term negative consequences.

Alexander writes:

And the others? The alcoholic who says “Yup, I’m drinking myself to death and you can’t stop me?” I agree that it is in some sense rational. It is rational because that person has so many problems that drinking alcohol becomes more pleasant than dealing with them. Often, these problems are related to psychiatric issues – for example, many people with PTSD become alcoholics because alcohol helps them briefly forget their traumatic memories. There are many people who say they don’t want help with their drinking problem because they expect “help” to mean “take away the alcohol but give them nothing in exchange”. If “help” meant “replace the alcohol with some healthier coping mechanism that works just as well”, many of these people would take it in a heartbeat. I realize this doesn’t quite disprove Caplan’s thesis for this relatively small group of alcoholics, but I think it’s important to remember that “preference” is different from “they’re doing what they want and all is well”.

But this is a classic case of terminal vs. instrumental values! The terminal value of alcoholics with PTSD is not to be drunk; the terminal value of alcoholics with PTSD is to forget their traumatic memories, and they instrumentally value alcohol as a strategy towards this goal. Of course it’s a good idea to point out to them strategies with fewer negative consequences. But it isn’t a “checkmate, preference-ists!” point.

Imagine, instead, a college student with Alcohol Use Disorder: they spend much of their time either drunk or hungover, often neglect their studies, get in trouble with their parents because of their poorer grades, have blacked out, have sometimes had unsafe sex while drunk, and sometimes intend to drink less because they have a test in the morning but get carried away. They do this not because of their mental health problems but because partying is really fun and they like it, and once they have a job they won’t get to do it nearly as often. It seems to me that the easiest way to describe this situation is not “this college student has an illness which causes them to drink more”; the easiest way to describe this situation is “this college student likes to drink.”

On a more philosophical note, Alexander does point out some problems with “preference” as a concept:

And his assumptions about the causes may be wrong. Bob’s issues are probably caused by what we call “sickness behavior”, a chemical defense in which the immune system notices an infection and releases cytokines telling your brain to avoid action and conserve energy in order to help with recovery. But one of the theories of depression I have found most plausible is that it’s a malfunctioning of sickness behavior – you’re not necessarily really sick, but your immune system releases its “stop acting and lie in bed all day so we can recover” chemicals anyway. If flu and depression have the same proximal cause, and the same effects on your life, where does Bryan draw the budget/preferences line?

For that matter, does Bryan ever get tired? I mean, suppose he is up very late one night and then has to go to work on only an hour of sleep. If he’s like the rest of us, he probably does a terrible job, can’t concentrate, and maybe rushes through things to get home early so he can catch a nap. Is this a budgetary constraint, or different preferences? In one sense it seems budgetary – he is lacking a resource (sleep? mental energy?) that would allow him to do a good job if he had it. In another sense it is clearly preferential – he places much less value in working hard and much more value in rushing home to get a nap.

Either way, this seems like a fruitful way to think about conditions like ADHD. Someone with ADHD, like someone who’s working on an hour of sleep, finds themselves miserable and unable to focus. If we call this a budgetary constraint, Bryan’s whole argument comes tumbling down. But if we call it a preference, then it’s a very strange type of preference, one where the usual method of “oh, great, you’re doing what you prefer!” is entirely the wrong approach.

I am sympathetic to this point. (Ozy Is Not A Preference Utilitarian, friends– although Alexander is.) However, it offers no principled way to say that some things are mental disorders and some things aren’t. Your friendly neighborhood evangelical Christian can say “look, jerking off to porn is a misfiring of sexual behavior– you’re not actually going to have sex with the doe-eyed girls badly faking orgasm, but your brain releases all its ‘OOH YEAH LET’S FUCK’ hormones anyway. Therefore porn addiction is a very serious disease and half the men in the country need to come to our rehabs.” The only thing you can do is say “no, depressed people are unhappy and don’t want to be depressed, but men who jerk off to porn are happy and want to jerk off to porn.” And then you’re back to the squidgy ‘preference’ issue again.

And I don’t think any of that addresses the reason that preferences matter.

Trying to find out what the Good is is really really hard and everyone disagrees about it. However, whatever the Good is, we can all agree that people mostly want the Good for themselves, on account of most people take a strong pro-good-things and anti-bad-things position. Therefore, liberal society agreed that we would all try to respect people’s preferences, as long as they weren’t interfering with other people’s preferences. And while there are a lot of issues here (when do we decide that people’s preferences are sufficiently dumb that we can interfere with them for their own good? how do we make sure people know enough that they can actually follow their preferences? what does “interfering with other people’s preferences” mean exactly?), in general I think this plan has worked out pretty well and I’m a fan of it.

Now, if you imagine yourself behind the proverbial veil of ignorance, trying to design a society so that your preferences will be met whatever they turn out to be, then you might want to put in rules about when Society ™ will interfere to help people get the things they want. You might lay out the following conditions for society-provided help:

  • If people are so unhappy about not having something that large numbers of them attempt suicide.
  • If people have a very strong conflict between what they want and what they want to want.
  • If people are so impaired they can’t get a lot of the things they want.
  • If people’s instrumental preferences are very, very bad at helping them reach their terminal preferences.

And then you get (respectively) depression, borderline personality disorder, and gender dysphoria treatment; exposure therapy and every kind of addiction rehab; ADHD, depression again, and psychosis treatment; PTSD, depression again, borderline personality disorder again, addiction again, anxiety again… actually this one is a lot of mental disorders.

Furthermore, you might say:

  • If people want to die, given that this is a very permanent decision, we will make sure they still want it in a couple weeks.
  • If people want to seriously hurt others, we will try to prevent this.
  • If people are very disconnected from reality, we will treat them so that we can find out what their informed preferences are, although if someone has an informed preference to be disconnected from reality then they should be allowed to do so.

Which gets you three kinds of nonconsensual treatment, although all far more limited than American society currently has.