This is confusing, I think. Sexual selection as I understand it is also explicitly concerned with certain traits, and it’s clearly not artificial. The important distinction here is “breeding for” - “intelligent” control over the reproduction of a population. IOW, I think Der Trihs is correct in saying that the term for human artificial selection is eugenics.
Personally, I would define “artificial selection” as selection carried out by humans with the deliberate intention of producing specific traits. Humans can produce changes in animal populations unintentionally. For example, we have caused a reduction of antler size in some deer populations by taking the largest males for trophies, and caused certain fish to mature at younger ages by preferentially harvesting the larger ones. I would regard that as a case of natural selection by a predator.
Humans selecting their mates on the basis of attraction, social status, or other features is also natural selection. It would only be artificial selection if you chose your mate with the specific intention of having blue-eyed children, say.
Given the enormous population size of humans at present, and the fact that the population is becoming increasing mixed at the global level, it will be difficult for selection to have an effect on gene frequencies on our population as a whole. Most evolutionary change is thought to take place in small, isolated populations.
Sexual selection is a form of natural selection. Sexual selection is not selecting for those traits specifically. That is, the female is looking for a fit male individual; she is not looking only for whomever has the brightest colors or best mating dance - those things are representative of the male’s fitness, not an end unto themselves.
Whereas there is no benefit to the chihuahua to look like it does; it looks that way because humans wanted it to look that way.
In the opposite direction, many medical advances also result in such changes.
For example, the development of insulin. Previously, most diabetics died before reaching breeding age, so that diabetic genetic lines tended to die out. But with insulin, diabetics can live near-normal lifespans, and can produce children. More of those children are themself diabetic, thus the percentage of diabetics in the human population is rising. (There are also allegations that modern diet is another contributing factor.)
So while increasing the proportion of diabetics in the human population was not an intended result of developing insulin as a treatment for diabetes, it is an actual resulting artificial selection. This is also true of most such treatments for formerly terminal diseases of the young.
This is an important point that is hard for most pro-eugenicists to grasp. Of course if we prevent those with deleterious alleles from breeding, we’ll slightly reduce the incidence of the allele in the next generation. But the selective pressure from this on rare alleles is tiny.
Imagine a genetic disease that affects 1 person in 10,000. That means in the United States we would expect to see about 30,000 cases. So we just cull those that express the trait, and bing bang boom, the trait disappears. Except, the trait is only expressed in those that have two copies of the allele. And how common is the bad allele in the general population? To be expressed in 1 out of 10,000 people, that means that 1 in 100 people are carriers for the trait. That’s assuming random mating of course.
So what you’ve done by culling those that express the trait is eliminate only 1% of the gene. That’s not very strong selective pressure. And the rarer the trait, the smaller the selective pressure by culling those that express the trait.
And this is why such large number of deleterious alleles persist in most populations, because the selective pressure against them is very low. As the allele gets rarer and rarer, the selection pressure against the allele gets weaker and weaker. And so the alleles persist at an equilibrium state, where the trait is selected against but new mutations keep reintroducing the trait.
And most people are carriers for several of these deleterious genes, it’s just that since they are pretty rare, it’s also pretty rare to mate with another carrier. Unless you’re inbreeding, in which case the chances go way up.
Yes, but this effect is really quite small. Not culling those who express deleterious recessives means the frequency of the allele in the next generation increases slightly, or rather, that it doesn’t decrease slightly. Except, if as in my example 1 in 10,000 people has the trait, that means 1 in 100 has one copy of the allele. So if we don’t cull those with the trait, that means that instead of a frequency of 0.01%, the next generation shows a frequency of 0.0101%. That isn’t a big increase. Sure, over hundreds of generations you’ll see an effect. Except, we haven’t had modern medicine for hundreds of generations.
More precisely, the trait is selected against but new mutations keep reintroducing other distinct deleterious traits. It’d be very unlikely for separate mutations to come up with the same trait.
Or maybe not.
Not necessarily. A variety of different mutations can have the effect of knocking out a particular enzyme, for example. And when you have a population of billions, rare mutations can keep popping up.
It’s not like you’re guaranteed bad results from inbreeding. But if you’re a carrier of a bad allele, your sibling is 50% likely to also be a carrier, which means that one of your offspring will be 25% likely to get two copies of the allele and express the trait. So it’s not that inbreeding causes genetic defects, it’s that if you are are carrier for a genetic defect, your relatives are more likely to also be carriers for the same allele. Of course, if you’re NOT a carrier for the allele, your relatives are also more likely to also not be carriers.
And we see that certain populations tend to have much higher incidences of certain traits than others, Tay-Sachs disease and hemophilia are the canonical examples.
I’m surprised I’m the first one in this thread to mention the Lebensborn program in Nazi Germany.
Previous threads on topic: How can we speak with such authority re dog breeding genetics but not human genetics? - Factual Questions - Straight Dope Message Board & (better) What, scientifically, is wrong with eugenics? - Factual Questions - Straight Dope Message Board
Of course it has happened! Do done of you read the Great Historian Heinlein? The Howard Families are amongst us but the Masquerade is still working… 
Seriously, the idea Heinlein proposes, of a Foundation, a corporate entity with an indefinite life span, operating the selective breeding programme to produce long-lived humans would answer some of the points raised up thread about the problem of long human generations. Not to say it would work but there is clearly a genetic component to human longevity and ageing (though how much is by no means settled).
Nitpick: it’s Homo sapiens. “Sapiens” is not a plural, you can’t knock the “s” off the end.
Everyone is probably carrying a few bad recessive alleles*. Spontaneous mutation in the egg and sperm create a few new bad recessive alleles each generation. The probabilities you quote are absolutely correct, but I’d say it’s probably true that all relatives are carriers for a few shared bad recessive alleles. A single cross between siblings or cousins will increase the risk of (but not guarantee) genetic defects in their offspring. Several generations of inbreeding will almost certainly result in the expression of a number of genetic defects.
Now, some of these mutations may result in subtle genetic defects in affected offspring, like a moderate risk of some disease. Other recessive alleles might be so deleterious that they stop early development, which would just manifest as a lower fertility for two related people.
To circle around back to the OP, this is why animal breeding usually involves inbreeding to fix desired genetic traits and outcrossing to remove bad genetic traits.
*I’ve seen estimates of around 5-10 for any one individual. We’ll probably have a more exact figure soon thanks to projects that are sequencing thousands of human genomes…
Inbreeding also removes the bad traits, since a trait is only selected against when it’s expressed. Consider, for example, a population (presumably close relatives, possibly even clones) who all have one copy of the same recessive gene. So the incidence of that gene in the population is 50%. Take two of those individuals, and breed them together. A quarter of their offspring will have two copies of the recessive gene, half will have one copy, and a quarter will have no copy. The ones that show the recessive trait, you cull (kill them, or sterilize them, or just let the trait itself kill or sterilize them). So what you’re left with is 2/3 of the surviving population having one copy of the bad gene, and 1/3 of the surviving population having no copy of the bad gene, thus reducing the incidence of the gene from 1/2 to 1/3.