Imagine you come across a society that categorizes people by their finger length ratios. They say if you have a longer 2nd digit your race is 2D, longer 4th digit your race is 4D.

Newspaper descriptions of people frequently say if they are 2D or 4D, especially if crime is involved. Job applications ask if you are 2D or 4D. There have to be laws banning 2D and 4D discrimination because it's so common.

You tell them it's crazy that they spend so much time worrying about 2D vs 4D. That it hardly matters! In return they tell you of course digit-race exists! You can easily measure it, and it's partly heritable. In fact it's scientifically unsound to ignore digit-race because it correlates with so many health issues [1].

Your cries that those health issues would be better understood with other ways of thinking about populations are ignored.

[1] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5296424/

It’s only diagnostically relevant in the context of imprecise population categories, lack of knowledge about disease origins, and the absence of genetic testing for conditions (i.e not in 2024). As others have pointed out in this thread, for example, the conflation of Black with prone to sickle cell disease is erroneous and more correctly determined by ancestry proximate to the region of the disease’s origin. Similarly, “white” in the US includes a wide swathe of genetics with various degrees of predisposition to melanoma (e.g. Northern vs Southern European).

It's not diagnostically determinative, but it's diagnostically relevant. Not all black people have sickle cell disease, nor do all people of sub-Saharan African descent, nor do all people who have one parent known to a carrier of variant HBB genes etc. We're just talking about the degree of statistical likelihood that something from a patient history points to a diagnosis.

> more correctly determined by ancestry proximate to the region of the disease’s origin

Is the better metric just as easy to determine though? In other words, it may be a mere proxy to a "better" metric, but that means it's still correlative, and we can only abandon it if we have no better proxy.

In most modern health care settings, yes.

You're saying we can determine ancestry proximatry genetically?