Inbreeding leads to an increase in genetic problems that exist in the genetic pool that is being in-bred. That's how you get things like "the island of the colorblind" that Oliver Saks wrote about with an island where a rare form of colorblindness where the center of the eye that is usually predominantly made of cones, which are what perceive color, is instead made of rods, which is what most people have primarily for their peripheral vision which is more useful for spotting motion and for dark vision, but has no color vision at all. People with this condition are extremely photophobic, have less visual resolution, and no color vision (as opposed to more common types of color-blindness such as red-green where people still see colors). But the point is, it's common there, because the genes for it existed and concentrated.

But likewise, take an isolated group and if a gene is absent, say Tay-Sachs which is a risk for my ancestry but is fairly uncommon for many people's ancestries then they can inbreed to their heart's content, they can do sibling-sibling and they have a higher risk of Tay-Sachs from out-breeding than they do from inbreeding. Sure, there is the chance of it randomly mutating and re-arising in the gene pool, but that isn't any more likely because of inbreeding.

So, if autism is a purely genetic trait (which isn't clear and seems unlikely but might be... I'm really not sure at this point) then you'd have to see how prevalent the gene was in the community.

Which I know, you made that point. I just... it's baffling how people can throw such bad arguments around. I have found that with inbreeding in particular people tend to have a lot of misunderstandings. I think it's because they tend to combine a small amount of understanding of genetics with the incest taboo and squick. Even though a single generation of incest isn't actually genetically necessarily a risk to the children (unless you have good reason to believe your family carries a fairly dangerous recessive). And less squicky matings but carried out over multiple generations can actually be a problem, if they do involve cumulative narrowing of the gene pool. Although it does also maximize good traits... see purebreds/thoroughbreds for some of the best and worst of both ends of that.

Oh wait, sorry, logic, reason, facts, data... what on Earth am I doing on the internet?