Evolution works just fine on polygenic traits. But yeah, mutations happen a lot. Most mutations are bad. Some are neutral. Very few are good. Old fathers introduce the most mutations into their offspring, and look at what happens. The rates of autism and schizophrenia are higher. Mutations are generally bad for brain function, since that is a complex organ. Look up mutational load.

Yes, culture influences selection. If a small group of people right now refuse to use birth control, what will happen over time? As Tyler says, solve for the equilibrium. Selection is happening right now relatively rapidly right now. The concept popularized by Taleb–Lindy–applies to a lot of arrangements at the moment. They do not seem Lindy. Will female careerism will eventually die? If they keep having few kids, yes.

The main differences between the wild forms and domesticated wheat are that domesticated forms have larger seeds with hulls and a non-shattering rachis. When wild wheat is ripe, the rachis—the stem that keeps the wheat shafts together—shatters so that the seeds can disperse themselves. Without hulls, they germinate rapidly. But that naturally useful brittleness doesn’t suit humans, who prefer to harvest wheat from the plant rather than off the surrounding earth.

From a spiritual perspective, that defect is by design and a sign of divine providence. That and cattle that are docile and don’t have horns. That combination which led to what anthropologists call the “wheat-beef people”, from which we are all descended.

When four arches hold up a dome, they have to be fitted together. One way creates a triangular surface, a “spandrel” at the four corners where the arches meet (a nice picture is here). It was soon realized that this “spandrel” could be filled with religious imagery. But, say, Gould and Lewontin, that was not the “purpose” of the spandrels of San Marcos. They came into being for quite a different reason, simply to help hold the dome up. Similarly, they argue, lots of what living things have didn’t happen for the reasons they exist today. In fact, they may not even have any direct purpose–today or when they began.

The paper was part of an anti-adaptationist campaign that Gould and Lewontin and a number of left-leaning biologists were conducting at the time (the two ran a seminar at Harvard called “Marxian Methods in the Biological Sciences”). They were worried that “adaptationist” arguments sounded too much like arguments for existing capitalism. “Adaptationist” arguments say that living things are the way they are because their characteristics are good for them; they help them survive. There is even a hint that “they have to be that way”. Far from it, said G & L. (Gould is famous for arguing that if you rewound the tape of life and then sent if forward again, you wouldn’t get the humans of today; in fact, you might not get anything even close.) Similarly, capitalism was hardly the way things have to be, nor in any sense an optimum.

“Adaptationist” arguments also tend toward what Gould attacked as “biological determinism”, that we are limited by our genes, that various schemes to improve social arrangements or to fix people are not possible, may even be harmful. As lefties, this was anathema to them.

My slogan might be “Evolution selects for traits, and genes only code for proteins”

Exactly, but with one caveat: the genome also codes for emergent properties associated with gene expression but none of the current statistics use these, only proteins/enzymes and their alleles. I am fond of the title of Bret Weinstein’s (Eric’s brother) doctoral thesis “Evolutionary Trade-Offs: Emergent Constraints and Their Adaptive Consequences” (PDF). “Emergent Constraints” is a good description of the problems that occur during gestation and Eric’s focus on learning disorders (i.e. neurodivergent).

Suppose that I have a combination of genes that is far from ideal for survival. But a lot of those genes overlap with genes that are ideal for survival, so evolution cannot be sure what to keep and what to discard.

Gould and Lewontin call these Spandrels:

In evolutionary biology, a spandrel is a phenotypic characteristic that is a byproduct of the evolution of some other characteristic, rather than a direct product of adaptive selection. That is, it is a trait that is not particularly advantageous to have, though it is retained because it is not particularly harmful to have.

Selection acts at many levels, but in different ways per level. The one thing on which selection cannot act is ‘traits’ per se. Selection can only act on things that reproduce and die/disappear. It can act on stretches of DNA, segments of DNA, certain kinds of protein folding patterns, organisms, populations that become more than one population, etc.

The identification of the relevant ‘traits’ of a reproducing ‘thing’ on which selection can act is about causation – simplifying the selection process to ‘major factors’ we can identify, discuss, engineer rationally, etc. This discussion becomes complicated rapidly by issues of identification and naming (spandrels, chins, etc); also about trade-spaces between the traits you choose to name (including the value of extremes, at times), about system dynamics – the fact that the process of selection is incomplete most of the time, and is facing patchy, changing environments, which a single population might span – so that some combination of diversity and maladaptation is preserved.

Getting the foundations right doesn’t necessarily make the subsequent discussion easy; population thinking, causation, etc are hard.

But I don’t think this is right: “My sense is that the relationship between the inputs and the outputs is so complex that not only can we not figure out that relationship, but evolution cannot figure it out, either”

“Pressure And Time.” With enough pressure and time, a river can carve canyons, a tectonic plate can raise mountain ranges, inches of slow becomes glaciers, and glaciers can grind jagged mountains smooth.

In statistics, there’s a world of difference between “cannot” and “better than nothing”, even if it’s only an imperceptible, tiny bit better than nothing.

Let’s say you have a population that is 99.3% A and only 0.7% B, and you have a weak statistical filter such that if you apply it to a population, the proportion of B only improves by 0.1%. That’s a terrible filter, especially it’s costly every time you do it.

But, so long as this filter is better than nothing, if you set up thousands of them in a cascade, you can get to arbitrarily high fractions of B, all the way up to 99.9% if you need to.

And that’s how we enrich Uranium for reactors and bombs, and that’s how evolution changes allele frequencies for complex traits, even when the benefit of any particular variant are weak, and even when there are bad side-effects and costly trade-offs to it (as with sickle cell resistance to malaria).

If evolution couldn’t figure it out, we wouldn’t observe relatively quick changes in group averages for traits or allele frequencies in response to particular selection pressures. But we do. It all depends on how severe the selection is, and how many iterations of generations it has been operating.

For example, consider that there are at least several dozen genetic variations that are much more frequent in Inuit peoples than in other groups, which in aggregate have substantial effects on polygenetically-influenced aspects of metabolism which in turn enhance one’s ability to thrive on diets where most calories come from certain fats.

Or consider artificial selection in animal husbandry or horticulture, and the relatively rapid and incredible increase in the expression of desirable traits we can get just by picking the offspring that have above-average expression of those traits and breeding them with each other. If evolution applies only 1% of that selective pressure, it still only needs more time to get similar results.