There has been a long running debate going on between Bryan Caplan and Scott Alexander on how to understand mental illness. Caplan argues that mental illness doesn’t really exist. Very briefly, Caplan uses the distinction between budget constraints and preferences in consumer choice theory to analyze the behavior of the “mentally ill.” A key component of his view is what Caplan calls the “gun to the head test.” If you put a gun to the head of a diabetic and told them to normalize their insulin levels (without medical intervention), they wouldn’t be able to do anything differently. But if you put a gun to the head of someone with an overeating disorder and ordered them to put down the doughnut, they would be able to do so. This, Caplan says, shows the overeater is capable of doing otherwise while the diabetic is not. Therefore, the diabetic faces a constraint and has a true illness, whereas the overeater just has a really strong preference for eating lots of food, and therefore compulsive overeating is just fulfilling a preference and doesn’t qualify as an illness or disorder.

Scott Alexander replied that consumer choice theory is inadequate for understanding or classifying mental illness for a variety of reasons. Bryan offered a rebuttal, Alexander came back with a rejoinder, and now Caplan has responded yet again. I recommend reading the entire exchange for full context. While I admire both of these thinkers greatly, I think Alexander has the stronger arguments.

At the highest level, I think Caplan gives far too much credence to the consumer choice model he uses. Economic models are useful tools, but like all models in social science, they are useful because they are simplifications. The map is not the territory, and the model is not reality. And any model of human behavior that does not perfectly and completely describe reality (which is to say, all of them) can end up being more confusing than enlightening when misapplied.

Consider the compulsive overeater. Overeating has many potential causes, of course, but one of these causes is leptin deficiency. Leptin is a hormone that regulates hunger and desire to eat. In his book The Hungry Brain, Stephen Guyenet describes it in the following way:

While a normal child may be about 25% fat, and a typical child with obesity may be 40% fat, leptin-deficient children are up to 60% fat. Farooqi explains that the primary reason leptin-deficient children develop obesity is that they have “an incredible drive to eat”…leptin-deficient children are nearly always hungry, and they almost always want to eat, even shortly after meals. Their appetite is so exaggerated that it’s almost impossible to put them on a diet: if their food is restricted, they find some way to eat, including retrieving stale morsels from the trash can and gnawing on fish sticks directly from the freezer. This is the desperation of starvation.

Yes, such a person might very well put down the doughnut (or trashcan scraps and uncooked fish sticks) if you held a gun to their head at any given moment. But so what? Their behavior still seems to me like it’s much better described as a budget constraint caused by low leptin levels, and not as someone merely fulfilling their unusual and socially disapproved preference to eat themselves into oblivion.

Another reason I find the gun-to-the-head test unimpressive is that it contains a hidden premise that I don’t think can be justified. Here’s how Caplan describes this test in his most recent post:

If any incentive in the universe makes you stop, you must have been able to stop all along. Incentives matter implies voluntariness implies preference implies non-disease.* That’s my Gun-to-the-Head Test.

The hidden premise behind this test is the idea that any behavior someone can engage in (or refrain from) while under extreme, life-threatening duress is therefore something they are capable of engaging in (or refraining from) at all times, for their entire life. But I don’t see any reason to believe this is true. Consider, for example, the case of mothers who have lifted cars off the ground to save their trapped children. Suppose a week before that happened, you asked these women to deadlift 500 pounds in the gym and found none of them could do it. Yet, a week later, they lifted considerably more weight than that in order to save their child. I’d say this is just a case of showing that what a person is capable of doing is different in normal circumstances and in extreme circumstances.

As I understand it, Caplan’s argument would commit him to saying that since there was at least one “incentive in the universe” that made them lift such immense weight, that shows they must have been able to lift such immense weight all along, and their inability to pull off a 500-pound deadlift the prior week wasn’t a real constraint, it was just them expressing their preference for not lifting heavy weights. That’s what a straightforward application of consumer choice theory would imply, but that only shows the limits of consumer choice theory. Yes, incentives did matter in their car-lifting feat, but that does not imply the inability to carry out such a feat in normal circumstances is therefore “voluntary” in any meaningful or interesting way, nor does it imply that the genuine inability to deadlift 500 pounds the week prior was actually just a preference.

(As an aside, Alexander is also unimpressed with this test, offering to “tell [Caplan] 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?” But notice the asterisk above in Bryan’s description of his test. That asterisk leads to footnote where Caplan implies that even if someone does take a gunshot to the head over altering their behavior, that still wouldn’t falsify his argument, because “incentives don’t matter does not imply involuntariness, though it leaves the possibility open.” When every possible outcome of one’s hand-picked method of testing their view can still be interpreted as compatible with that view, then it’s not a very impressive test, and holding it up as some sort of ace-in-the-hole for the argument doesn’t inspire confidence.)

People are sometimes temporarily capable of things in extreme duress they couldn’t achieve in normal circumstances. This is both common sense and widely known. This aspect of human behavior doesn’t fit into the simple consumer choice model of constraints and preferences – and that’s okay! Consumer choice theory isn’t and shouldn’t be treated as a theory-of-everything, meant to explain and classify all forms of human behavior. It’s just a useful oversimplification for understanding a small subset of human life.

In the closing paragraphs of Caplan’s most recent post, he cites an argument from someone named Emil Kierkegaard making the case that homosexuality is best understood as a mental disorder – a position Caplan disagrees with, arguing that homosexuality is simply a preference. Caplan closes out by saying:

Emil’s position may fill you with rage, but it’s logically consistent. So is mine. Choose one, because as Emil’s namesake Søren Kierkegaard titled his most famous book, it’s Either/Or.

In response to this, I would quote Ralph Waldo Emerson’s observation that “A foolish consistency is the hobgoblin of little minds, adored by little statesmen and philosophers and divines.” To go back to my observation in the beginning, economic models – including consumer choice theory – are not perfect descriptors of all reality. And when your model doesn’t fully capture reality, forming all your beliefs to be perfectly consistent with that model is not automatically a virtue. The understanding we gain of the world from any of our models will always be limited and partial. In light of this fact, being a little inconsistent will often be more truth-preserving than perfect consistency. And on this topic, I find Scott Alexander’s less-than-perfect consistency far more truth-preserving than the total consistency of either Caplan or Kierkegaard.