Saturday, November 22, 2008

On generalizations

"Generalizations are always evil"

"Generally speaking, anyways"

Are generalizations intrinsically good or evil? Neither, obviously.

As an amateur critic, I talk all the time about people and their beliefs. I could either use an anecdote, implicitly generalizing my experiences to a larger whole, or I could skip that step and talk about groups of people as if they were abstract entities. Or I could stop being a critic altogether. Obviously, I have no business unconditionally condemning generalizations. They're all I have.

However, generalizations can obviously be abused. On the extreme end is racism and other bigotry. One could argue that the evil is in the falsity of such generalizations. Women and ethnic minorities are not actually inferior to white males. Homosexuals are not actually perversions created by the Devil. But perhaps a bit of the problem is in the generalization itself? How might we correct for this?

I think a little intuitive statistics goes a long way. Simply put, generalizations apply to groups, and not to individuals.

Let's say you've scientifically proven that a certain group of people usually exhibits a certain trait X. If you take any subset of the group, and count how many people have trait X, you're going to have a certain amount of theoretical uncertainty, just due to random sampling. That uncertainty is roughly sqrt(N) people, where N is the size of the sample. So if you have only a single individual, the uncertainty is roughly one person.

I'm sure somewhere in there is the message, "You can choose", just like they told us in Minority Report.

And that only goes for scientifically proven generalizations. I can assure you that any generalizations I make are not scientific at all. If I were to estimate, I'd say 0% of the statistics and generalizations on this website are backed up by scientific evidence. They should be treated with the appropriate skepticism.

This is why I take personality tests with a grain of salt. Even if they are indeed scientifically tested (which is more than we can say of many personality tests), they're still based on the idea that I should take statistical results and apply them to myself. The uncertainty is one person!

On the flip side, if generalizations can't be applied to individuals, this leaves a rationale for people hold onto their preconceived prejudices despite the clear existence of counterexamples. Hrrrm...