Ontology Of Psychiatric Conditions: Tradeoffs And Failures
To what degree are psychiatric conditions more like diseases (always bad) vs. diverse neurotypes (potentially good)?
[previously in sequence: Taxometrics, Dynamical Systems. Epistemic status: speculative. This should go without saying, but when I talk about “failures” in this post, I mean failures of biological processes, as in the term “congestive heart failure”; I don’t mean to accuse people with psychiatric conditions of being failures.]
Most psychiatric disorders are at least partly genetic. Some, like schizophrenia and ADHD, are very genetic, probably 80% plus. This is strange, because having psychiatric disorders seems bad, so you would expect evolution to have eliminated those genes. Researchers looking into this question argue between two hypotheses.
First, a failure. Evolution is imperfect, so some bad genes manage to slip through. This sounds dismissive, but it's definitely true to some degree. Thousands of different genes contribute to risk for conditions like ADHD and schizophrenia, with each adding only a tiny amount of risk. When a gene is only very slightly bad, it takes evolution millennia to get rid of it, and during those millennia people are getting new very-slightly-bad mutations, so it all balances out at a certain level of bad genes per generation. Those bad genes are sufficient to explain the existing amount of ADHD and schizophrenia; they're just evolution not working as well as we'd hope.
Second, a tradeoff between two goods. The genes for psychiatric disorders are good in some way. Maybe having some schizophrenia genes (maybe not enough to give you schizophrenia) makes you more creative and raises your inclusive fitness. This keeps schizophrenia risk genes in the population, and sometimes two people with very high level of these genes will mate and their child will have schizophrenia. "Higher creativity" vs. "lower schizophrenia risk" is a tradeoff, and different people are at different points on the tradeoff, and some people will be so far to one end that they will get schizophrenia.
Recent research has pretty heavily favored the failure hypothesis. If you have enough people's genomes, you can use some complicated math to infer how evolution is affecting different genes. And on most of the schizophrenia risk genes we know about, evolution has been gradually eliminating them in a way that looks like they're on net harmful - not keeping them around in a way that looks like they have counterbalancing advantages. In the modern day, people with genes for psychiatric disorders tend to have fewer, rather than more children than people without those genes - except in the case of ADHD, which I'm tempted to cynically attribute to them being less likely to remember to use contraception.
Also, a lot of the theories about how psychiatric disorder genes are good suggest that different disorders are good in opposite ways. For example, schizophrenia genes are supposed to give you more artistic creativity, whereas autism genes are supposed to make you more cool-headed and rational. This makes a kind of intuitive sense looking at the symptoms of the disorders. But it turns out that many, many of the genes that cause autism also cause schizophrenia, and vice versa. They seem to be general genes for having mental disorders, with a wide variety of negative effects - which seems like a better match for the first theory where they're just plain bad news and evolution hasn't gotten around to eliminating them yet.
Also, things that we know are failures are known risk factors for psychiatric disorders. For example, if your mother got a serious infection when she was pregnant with you, your risk of autism goes up 70%. In fact, just being conceived in March raises your autism risk a bit - it means you were in an especially vulnerable developmental stage duriing flu season! Prenatal or perinatal pesticide exposure probably increases autism risk. If you couldn't breathe for a while after birth - either something went wrong during the birth process or your lungs just took a while to come online - that doubles your risk (see exact numbers for that and sixty other prenatal and perinatal factors here). None of these things are good or have any counterbalancing advantages. They're all just failures of the normal developmental process that can be expected to make you worse off.
So the failure hypothesis is definitely the way that the evidential winds are blowing.
Still, I can't help feeling like some tradeoffs have to be involved. Some of the research findings that make me think this:
Many large studies (1, 2, 3, etc) show a strong correlation at the genetic level between autism risk and intelligence (especially mathematical intelligence), suggesting that the more autism genes you have, the smarter you are (although actually autistic people tend to have lower measured IQs than neurotypicals). Other studies (1, 2, 3, etc) find an association between schizophrenia and creativity (especially artistic creativity) - see eg the story about James Joyce swimming vs. drowning. Still other studies (1, 2, etc) show that anorexia is associated with higher achievement in school. The effect remains even when controlled for IQ, suggesting it probably involves perfectionism and being willing to go to extremes in order to achieve socially valued outcomes - in the same way as, well, anorexia.
Also, a very few disorders do seem to be genetically opposite each other.
Overall genes that increase risk for ADHD decrease risk for OCD, and vice versa, suggesting that at least one advantage of ADHD is lowering OCD risk. It also supports a common-sense idea that there's a spectrum from much-too-careful to much-too-carefree whose extremes are ADHD and OCD, but which might have other advantages - being either careful or carefree can be helpful in certain situations.
And even beyond the studies - I used to work in the business district of San Francisco, meaning I got to see a lot of very high-functioning people with mental disorders. I was constantly surprised how many people - while genuinely suffering from their conditions - also seem to be succeeding partly because of them. The bipolar programmer who is nonfunctional half the year, but his company keeps him on anyway because he codes at an absolutely superhuman level of brilliance while manic. The obsessive-compulsive cybersecurity expert who finds weaknesses everybody else missed. The endless line of autistic people succeeding in math-heavy jobs, exactly the way the stereotypes would predict.
When I was in medical school, the joke (not really a joke) was that everyone with ADHD went into emergency medicine. I've since treated a couple of doctors with ADHD, and sure enough they are all in emergency medicine. But now I'm in the Bay Area, and the joke is that everyone with ADHD founds a startup. This has made its way into lots of business advice columns (1, 2, 3, etc), and is also confirmed by my personal experience - though I should also mention just how disproportionately people with ADHD become salespeople. My impression - and I could give you a couple of studies that vaguely back it up, but they'd all be sort of hokey - is that people with ADHD are worse at dealing with low levels of stimulation and situations requiring careful focus, but sometimes better at dealing with high levels of stimulation and situations requiring frantic task switching.
(if it weren't for medical confidentiality laws, I think I could make a fortune running a combination psychiatric clinic / employment agency. As soon as I diagnosed someone with ADHD, I would sell their contact information to recruiters from sales companies, the local ER, and Y Combinator. Autistics, big tech companies. Obsessives, anyone in security. Schizotypals, the nearest art school. Psychopaths, local government. It'd be great!)
This is all fun to think about, but again, the research is pretty discouraging. It shows that most genes for psychiatric disorders are clearly negative. It also shows that most psychiatric disorders decrease functioning, including the form of functioning one might expect them to help. People with autism have low IQ, people with schizophrenia have low creativity (at least on some ways of operationalizing creativity), people with ADHD have low IQ and generally worse outcomes in every domain. So the current scientific consensus is that no, it's definitely failures.
I think the best way to reconcile this is to say it's sometimes failures and sometimes tradeoffs, with the failures sufficiently more common that studies that naively just sum everything up (eg check whether the overall genetic correlation is positive or negative, or whether average IQ is higher or lower) mostly find a failure. Let's discuss in more detail what this might look like.
Suppose you hear of a nearby country where guilty criminals often walk free. There are just as many police officers as in your own country, they make just as many arrests, but when the trials are over, you hear a lot of stories about guilty people getting acquitted. You might consider two different hypotheses about why this happens.
One, the justice system in that country is terrible and cannot distinguish innocent from guilty people. Detectives constantly screw up, juries make bad decisions, judges are idiots.
Two, that country is deeply committed to the idea that no innocent person should ever go to jail. They carefully instruct juries not to convict unless the evidence really is "beyond a reasonable doubt", and are willing to accept that this means a lot of people who realistically probably broke the law won't get punished. Maybe they demand multiple witnesses before conviction, or lots of material evidence.
In first hypothesis, the problem comes from a failure. The justice system just sucks; if they could make it better they would, but they can't.
In the second hypothesis, the problem comes from a tradeoff. You can trade off "number of guilty people who correctly get punished" vs. "number of innocent people who correctly go free" by varying your evidential threshold for guilt higher or lower. This country chose to make it higher They get some good things out of this (fewer innocent people wrongly get punished) and some bad things out of this (fewer guilty people correctly get punished). Overall there is less wrongful imprisonment, but it's harder to catch criminals. This happens to be their choice on the tradeoff space; it would be a value judgment to call it better or worse than your own.
But in reality, failures and tradeoffs will interact. In a country biased towards finding people innocent, it only takes a tiny failure to let a murderer go free. In a country biased towards finding people guilty, it would take a huge failure, or many different failures in parallel. If a particular murderer goes free, we can always blame both the country's position both on the tradeoff (because if it had been less biased towards finding innocence than it was, the problem might not have happened) and the country's level of failure (because if detectives, juries, etc were more competent, they might have been able to meet even a very high burden of proof).
Or, consider the Stanislav Petrov scenario: apocalypse via nuclear false alarm. If the US falsely believes that Russia has launched nuclear missiles, they will "retaliate" with real nuclear missiles, starting a nuclear war. There are two sets of risk factors for such a false alarm. First, the US radar has to fail catastrophically, reporting a missile launch when no such launch has happened. Second, the US has to decide to respond to an apparent missile launch by immediate retaliation, rather than a "wait and see" approach.
The first risk factor is a failure. The missile system was poorly built, stupid, buggy, defective. There is no advantage to having a missile system like this. It just means you were too dumb to make a better one.
The second risk factor is a tradeoff. The slower you are to retaliate, the less likely you are to respond to false alarms - but also, the less likely you are to respond to real alarms in time. This is why the military has their DEFCON system. DEFCON5 (total safety) means the military doesn't see any immediate threat and apparent anomalies can be investigated at leisure. DEFCON1 means nuclear war is imminent and everyone should be on hair trigger alert. Being on DEFCON 1 vs. DEFCON 5 is a tradeoff. Both have their advantages and disadvantages.
If a false alarm starts a nuclear war, it will probably be because of a failure plus a tradeoff: a failure to have decent radar, plus a potentially reasonable decision to be at high alert. The stronger the tradeoff, the less of a failure you need to produce disaster, and vice versa. If the military was on DEFCON 5 (low alert), then it might have taken a phenomenal level of failure to make the commander launch the missile; maybe every radar system in the country would have had to fail simultaneously over a long period. On the other hand, if the military was on DEFCON 1 where they were expecting an enemy launch any second and the commander's finger was hovering over the launch button, maybe just someone sneezing would have set off the hair-trigger and caused the launch. You can have have a disaster from failure-with-no-tradeoff or from tradeoff-with-no-failure - but it's a lot easier when you have both at once.
I think this is how autoimmune diseases work, more or less. The immune system can be either more or less active. This is a tradeoff: the more active it is, the better it is at dealing with infection, but the more likely it is to accidentally attack the body and cause autoimmune symptoms. And then the immune system can either fail or not fail. Some people have very active immune systems, but the systems are very good at what they do and only attack genuine pathogens. Other people have very weak immune systems, but the immune system is so confused that the one thing it can muster up the energy to attack is your own body. Like the accidental nuclear war, the autoimmune disease is the combined result of a tradeoff plus a failure.
And I think this is how anxiety disorders work too. People can be more or less anxious. Being anxious all the time is no fun. But having no anxiety at all isn't fun either; maybe you poke a lion with a stick to see what happens, and then your genes don't make it into the next generation. Having higher or lower baseline anxiety is a tradeoff, just like having your country's military on high alert vs. low alert. But if you also have a failure - your anxiety flares up at things that aren't really that bad, and doesn't go away - then you have an anxiety disorder.
Either end of the tradeoff is defensible. But either end of the tradeoff is vulnerable to failure.
If you have high anxiety (but no failures), then you're a very careful person and it probably works out well for you, unless you're absurdly careful.
If you have low anxiety (but no failures), then you're a very brave person and that probably works out well too, unless you're suicidally brave.
If you have high anxiety (and many failures), then you're probably naturally protected against failures of too-low anxiety, but especially vulnerable to failures of too-high anxiety, and you panic constantly over little things that don't matter.
If you have low anxiety (and many failures), then you're probably naturally protected against failures of too-high anxiety, but especially vulnerable to failures of too-low anxiety, and you poke lions with sticks for no reason.
So problems with anxiety - constant disproportionate flares of anxiety that serve no legitimate goal - are a combination of tradeoff and failure.
Let's look at a complicated more real-life example: autism.
As we discussed before, there are plenty of well-known environmental risk factors for autism. None of these things - having to fight off infections in the womb, exposure to toxic chemicals, brain damage from suffocation - have any upside. They are just plain bad. There is no reason they should make your brain any better in any way, and as far as we know they don't make your brain better in any way. They just damage delicate developmental processes. They are part of the failure bin, not part of the tradeoff bin.
A second cause of autism is "de novo mutations" - new mutations that begin with the autistic person themselves, rather than being transmitted to them from their parents. For example, a certain especially bad type of de novo mutation was found in 20% of autistic patients and only 10% of their siblings, suggesting it significantly contributes to autism. Almost all de novo mutations are bad, for the same reason that changing a random letter in a book almost never causes the book to be better, and almost always makes it worse. In case you don't believe this for some reason, scientists have checked and found that the mutations are negatively selected against, meaning evolution agrees they are bad. It's unlikely that any of these mutations have positive tradeoffs. This is another failure.
Even normal autism risk genes that people get from their parents are mostly straight-out bad. At least this is how I interpret Tsur, Friger, and Menashe who find evidence of negative selection for most autism-related variants - based on their place in the genome, it looks like evolution has been trying hard to get rid of them. Also, some risk genes are shared with other psychiatric disorders, like schizophrenia, bipolar, ADHD, and Tourette's. These disorders all have very different symptoms, and very different associated personalities, but they all have one thing in common: they are abnormal and dysfunctional. Most likely these genes are involved in basic brain functions - making sure cells get enough oxygen, keeping the blood-brain barrier strong, preventing inflammation - and less functional alleles generally make everything worse. These, too, are in our failure column.
And yet there are clear signs that other autism risk genes are associated with intelligence, and some of these are even positively selected. My guess is that these involve a tradeoff. I don't know exactly what this tradeoff is, but I tend to go with Lawson, Rees, and Friston’s speculation that autistic brains think with higher precision. This lets them think more precisely when that would be useful, but it also means they're constantly getting false negatives (ie not recognizing a person's face if the shadows on it are slightly different), constantly having their attention hijacked by minor stimuli that the rest of us would ignore (sensitivities to stimuli very slightly different from expected stimuli, like the dripping of water or the feeling of the tag on their shirt), and constantly getting confused by minimal deviations from routine. And although it's less obvious, some people have speculated that this makes it harder to do the sort of intuitive categorization work that lets you draw conclusions from social situations.
My guess is that somebody who's chosen the far end of this tradeoff naturally ends up as the stereotypical "aspie engineer", who's very smart, a bit off, but not so far gone he can't hold down his job at Google.
Add some failures to this tradeoff, and things get much worse. Your brain gives you false information, and you refuse to recognize anything as a match unless it's precisely the same. This is when you get the sort of autistic people who can't take care of themselves and need to be institutionalized their whole lives.
If, as Badcock and Crespi claim, schizophrenia is the reverse of autism, it might involve being too imprecise - too willing to declare the identity of unlike data - too quick to pattern-match. If autistics are too quick to mistake signal for noise, schizophrenics are quick to mistake noise for signal. In a well-functioning brain, this makes them creative and socially adept; in a poorly-functioning brain that is constantly getting things wrong, they connect everything to everything else and it's a mess.
I don't know if this is quite right. But it would reconcile two seemingly contradictory sets of findings. First, Badcock and Crespi's claim that in many ways autism seems like the reverse of schizophrenia. Second, the genetics and epidemiology data showing that autism and schizophrenia share most of their genes and risk factors. The shared risk factors are things that cause failure.
Go back to the two ways a justice system can go wrong. First, it sets too many guilty people free. Second, it convicts too many innocent people. If you ask which tradeoffs cause each problem, you'll find they're opposites. But if you ask which failures cause each problem, you'll find they're mostly the same - lazy jurors, stupid detectives, insufficient funding. I think this matches the psychiatric case, where disorders can seem opposite in many ways but also share many of the same risk factors.
I think most psychiatric disorders exist on a spectrum from mostly-tradeoff to mostly-failure (what we might call "high-functioning" and "low-functioning" versions of the same phenotype).
Mostly-tradeoff ADHD looks like someone who is adventurous, likes variety, and thrives in high-stress situations, but is also bad at tolerating boring situations.
Mostly-tradeoff schizophrenia (which might be schizotypal personality disorder, or a subset of it, I'm not sure) looks like someone who is creative, attuned to interesting patterns, and charismatic, but also a bit odd and superstitious.
Mostly-tradeoff OCD (which might be obsessive-compulsive personality disorder) looks like somebody who's responsible and perfectionist, but has trouble letting things go.
I think Cluster B personality disorders are already tradeoffs; adding failures just make them less effective ones. For example, tradeoff-antisocial-personality is a con man who uses unethical means to get ahead (and sometimes succeeds); tradeoff-plus-failure-antisocial usually ends up in prison very quickly and never gets out.
I don't think depression, bipolar, PTSD, or addiction fit this model neatly. Depression and bipolar are probably attractor states in dynamical systems; it's possible that having this attractor state has counterbalancing advantages, but I couldn't begin to speculate on what they are. PTSD and addiction are a combination of a genetic predisposition with an event/substance that initiates the disorder; I don't know whether the genetic predisposition has any advantages.
This is why I'm sometimes friendly towards, but other times skeptical of, the "neurodiversity" movement. Many autistic people say that autism is just another, equally good, way to be human, not a disease that requires a cure. I think this is completely true for some autistic people - the ones at a different point in tradeoff-space, who are still on the genetic Pareto frontier of inclusive fitness. I think it's absolutely false for other autistic people, the ones who got that way through brain injuries or in utero infections, and tend to be found in institutions having seizures or trying to bite their own fingers off. Instead of insisting that psychiatric conditions cannot possibly be bad for anybody and it's always just social prejudice all the way down, they would be better off admitting that this can be true for some people and false for others.