Preferences and Constraints
Bryan Caplan has written a few times about why he believes mental illnesses are ‘socially disapproved preferences’. His view is that you can distinguish a preference from a constraint with the gun-to-the-head test:
“If suddenly pointing a gun at alcoholics induces them to stop drinking, then evidently sober behavior was in their choice set all along.”
If you point a gun at someone with ADHD and tell them to write the essay that they’ve been putting off doing, they’ll (probably) do it. If someone pointed a gun at me and told me to bench press a 120kg barbell, I wouldn’t be able to do it (sadly!). So we can conclude that ADHD is preference, whereas you can be genuinely constrained in your actions by your lack of strength. Similarly, depression is a preference - if someone with depression had a gun pointed at their head, they would get up out of bed.
You might object ‘but why would anyone want to have depression?’. The answer is that they don’t want to have depression, but that doesn’t prove that depression isn’t a preference - people who have the gene that makes coriander/cilantro taste like soap would likely rather not have that gene too, but their choice not to eat coriander/cilantro remains a preference rather than a constraint - they could eat it if there were strong incentives to do so. Caplan makes the distinction between preferences and meta-preferences. I might prefer not to have the preferences that I have, but that doesn’t stop them from being preferences - I don’t particularly like exercise (which is a preference), but I wish that I did. There has been a bit of back-and-forth between Scott Alexander and Caplan on the idea of mental illness as a socially disapproved preference, and I must confess that I haven’t read every post in the Caplan-Alexander debate, so I apologise in advance if I say some things that they’ve already been over.
Caplan’s claim is that people who have ADHD have a strong preference to do things other than work or focus on some specific task, and we can test this with the gun-to-the-head test. If it’s the case that people with ADHD could finish some task if faced with the threat of being shot, it’s a preference. I’m dubious about this - I’m not sure that anyone is claiming that the constraint faced by people with ADHD is that they’re literally unable to do some 30 minute task if there is a huge incentive to do the task, but rather that they find it much more difficult to do sustained chunks of work over long periods.
And for the more reasonable claim about what the constraints faced by people with ADHD actually are, I think we can make ADHD ‘pass’ the gun-to-the-head. Suppose that the maximum amount of intensely focused work the average person without ADHD can do is 7 hours per day (I don’t know if this figure sounds right, adjust up or down depending on your view). And now imagine that we put a gun to the head of a person with ADHD, and tell them that they have to match the productivity of someone without ADHD over the course of a year, and they’ll be shot if they don’t. I think that many people with ADHD would simply be unable to match the output of the person without ADHD. If I’m right, does Caplan now accept that for a person with ADHD, working long hours is a constraint rather than a preference?
Similarly, suppose that you tell an alcoholic that they’ll be shot if they drink a single drop of alcohol over the course of the next ten years, and one in five alcoholics ends up drinking and being shot, does Caplan concede that for 20% of alcoholics, sobriety wasn’t in their choice set? Nobody claims that the constraint faced by alcoholics is that they’re unable to go thirty minutes without a drink, the constraint they face is that they are unable to stay sober for significant periods of time.
Scott Alexander points to the fact that some people genuinely do choose death over their mental illness, writing:
If Bryan uses his gunshot analogy one more time, I am going to tell him about all of the mentally ill people I know about who did, in fact, non-metaphorically, non-hypothetically, choose a gunshot to the head over continuing to do the things their illness made it hard for them to do. Are you sure this is the easily-falsified hill you want to die on?
This is a powerful point - does Caplan at least accept that some people with mental illnesses are constrained in their actions rather than people with socially disapproved preferences? There are plenty of people who literally choose the gun, so even with the gun-to-the-test, we should accept that those people are constrained in their behaviour, right?
I’m not sure that Caplan ever responded to Scott’s second post, but I guess from his writing that he thinks that failing the gun-to-the-head test makes something definitely a preference, but passing the test doesn’t necessarily make it a constraint. Imagine that an intruder breaks into a man’s house, and tells the man that he must leave the house or he’ll be shot. If the man refuses to leave and ends up getting shot, this doesn’t make it the case that the man was literally unable to leave the house. For Caplan, the key point about constraints and preferences is that people with preferences respond to incentives - alcoholics drink less if the costs are higher, people with ADHD do more work if their incentives change, etc.
But then we get to a weird point where the examples that Caplan actually uses don’t work. For instance, he cites cancer as an example of a genuine constraint - they won’t live longer even if you pay them to. Or will they? There’s some evidence that if people face significant tax incentives to die at a certain time, they’re able to do so. Kopczuk and Slemrod find that ‘evidence from estate-tax returns suggests that some people will themselves to survive a bit longer if it will enrich their heirs’. I don’t want to say we can be totally confident in this paper, but it doesn’t seem totally implausible. If the response is, ‘okay, some people may be able to postpone their deaths by a few days, but they can’t postpone it for a year’, we should note that this also likely applies to Caplan’s examples of preferences - people with severe depression may be able to get out of bed for a day to avoid death, but they are unlikely to be able to act in a way indistinguishable from a mentally-well person for a year.
I think the most obvious solution to the problem is that mental illnesses or neurodevelopmental disorders can’t really be classified as either constraints or preferences. Rather, most mental illnesses and/or neurodevelopmental disorders impose some constraints on people while also altering their preferences. People with ADHD (sometimes) have a strong preference for relaxation over work, but they may also be genuinely constrained in the amount of work that they’re able to do over a sustained period. I also just think there’s something slightly nasty about using the word preference when it comes to people with depression or other severe mental illnesses - that isn’t an argument about it being wrong, but I think given Caplan’s previous claims that people ought to try and be kind in order to be convincing, it’s worth bearing in mind.
Caplan's arguments have always looked to me like they're confusing the map for the territory. The distinction between budget constraints and preferences comes from our economic models of consumer choice, which we use to try to explain and predict how consumer actions affect various important variables in the aggregate. These models are good, and very often very helpful - but they remain models, and (after all) all models are wrong. And one of the ways they are wrong is that when you break down the aggregate and look at specific cases of consumers making choices, their decision-making isn't easily factored into preferences (on the one hand) and budget constraints (on the other). Mental illness is just the clearest example, but human psychology generally is quite messy in ways that don't fit into simple models of decision-making. That's not to criticise those models: in many circumstances they can be useful both descriptively (they're approximately correct, and this approximation gets more correct as we aggregate examples) and normatively (they tell us how we should be thinking, rather than how we actually are thinking). But Caplan is just refusing to admit that this model isn't the exact truth, demands a clean and sharp distinction between budget constraints and preferences, and then - upon noting that mental illnesses aren't budget constraints in the idealised spherical-cow sense demanded by his model - concludes that they must purely be preferences.
This seems like it's reaching towards the best response to Caplan's position that I've seen, which is that most mental illness works as a budget constraint. People with these mental illnesses are still able to do the things their conditions inhibit, but it's harder and costs them more, and they'll run out of resources sooner than a more healthy person would.
Similarly a lot of people in wheelchairs could, exhaustingly and painfully, walk 50 feet if you put a gun to their head, But if you tried to make them walk 5 miles you'd have to shoot them.