« Migration to the UK from East Europe | Gene Expression Front Page | Chicks who fight! Addendum »
September 16, 2004

The land of grey generalities

Many of you have probably heard about an article in The Washington Post with the subheadling Geneticists and Historians Grapple With the Gray Areas of Race. I don't want to get into it at this point on the racial issue, we've got loads of stuff on it on this site. I do want to emphasize the idea of greyness or fuzziness. Speaking in generalities of means and variances where two distributions overlap, how exactly do you claim something is "significant" (why is "95% likelihood" significant and 94% not?).

When we got into it over at Tacitus there were two major classes of individuals who argued that race wasn't significant, class one used the Lewontin 15% intergroup variation point to argue that they were as likely to match someone of another race on any given locus as they were someone of their own race. I won't address that point. On the other hand, there are those who would argue that just because certain frequencies of any given allele are nearly disjoint (that is, A variant is found in group X while B variant in found in group Y at 99.9% frequencies), they are still only nearly disjoint (I'm exaggerating for effect). This line of argument often has an implicit assumtion that we are promoting "ideal types."

But at least with "race" we have allelic frequencies & F statistics. How do you "prove" an assertion like, "Islam has spread more by the sword than other religions?" I will be frank and say that I know enough about history that I could "fake" it and purposely frame the data points to make it seem like Hinduism has been just as aggressive as Islam (at least when pitted against someone who isn't able to draw on the same amount of data). I could look for evidence from the Sena period in Bengal or medieval conflicts between Jains & Hindus in southern India. How many data points would I need? What if I found 20 discrete incidents of Hindu persecution of religious minorities. Inserting them into strategic locations in essay form, I could "prove" my thesis pretty easily to those caught unawares.

Now, jihad can be spiritual struggle. I happen to think throughout most of Muslim history it has more often mean holy war, and today, most Muslims interpret it that way. How do I prove it? If I dig up a survey that shows that the majority of Muslism think jihad is holy war someone could rebut that only the elite matters, the masses do not reflect "true Islam." They could point to the many non-coerced conversions to Islam, or the fact that Dalits in India convert escape the social restrictions of the Hindu caste system. How many examples could they line up? 20, 30, 40? Of course you say: but those aren't representative! How do we determine which facts are representative or not?

I guess what I'm trying to get at is that generalizations can sometimes be hard to communicate, let alone persuade with. Your opponent can always emphasize in their mind all the exceptions, and enough exceptions means that they can flip the generalization in their head. There are all the issues of confounding factors and determining just what the critical mass of facts are to swing from one position to the other.

Convincing others of topics outside science is more about attempting to stimulate internal debate in the other with the presentation of facts. You'll never convince them because you don't know all the facts and processes at work in their brain. You can only hope to allow their own wheels to turn enough so that they stumble on to your position independently because their facts overlap enough with your facts to drive them toward the same conclusion.

Posted by razib at 09:50 PM