I want to respond to James G's critique of this post. First because it was pretty intense, second because I usually enjoy reading his blog, and third because maybe other people have the same objection. I'm doing it here because his blog is closed to comments.

There is no basis to allege that everyone who says, “affirmative action is racist” is trying to position “affirmative action” in the very heart of the “racism” cluster. Clusters-in-thingspace, especially nebulous ones like “racism”, are huge volumes. That affirmative action belongs somewhere in this volume, rather than well outside, is a claim worth making even if affirmative action isn’t a central member...Affirmative action is racist!” draws attention to a cartographic error. “Affirmative action” shouldn’t be remote from “racism”; it is a marginal member of the racism cluster.

I would ask James why exactly we're trying to create a "racism" cluster to begin with. Are we ontologists who place things in categories for fun in our spare time? If so, his cartographic metaphor is apt; we're just trying to draw a map of conceptspace and we should be politely reminded that "affirmative action" is in the wrong part.

But in fact, our real reason for drawing a "racism" cluster is to make hidden inferences (see the section titled "Hidden Inferences" here). Most people see Hitler, the KKK, and South Africa, and decide racism is bad. Therefore, anything in the "racism" cluster is bad. Therefore, most people who want to draw maps of racism are not disinterested cartographers but people trying to convince others that something is bad because it is in that cluster.

To give an example, suppose a mapinguary is trying to lure you into the Amazon jungle to eat you. "Come to the city of Cayenne", it says. "It's in France, so you'll be perfectly safe." You reflect that France is in fact way up in Europe and not covered in jungle at all. So you go to Cayenne and it turns out you're in French Guyana and the mapinguary eats you.

The advice that you should place Cayenne in France is useful for cartographers but dangerous for people who don't want to be eaten by mapinguaries. While I agree with the philosophical point that not all French territories must agree with our archetypal example of France and that Cayenne is a perfectly acceptable marginal example of Frenchness, the practical point is that on the only criterion we're interested in, safety from mapinguaries, Cayenne defies all the inferences we could be expected to make from its Frenchness. Furthermore, the whole reason it was brought up was the hope you would make these bad inferences. It is not always bad and irrational to note that Cayenne is in France, but this particular example was, and there are some things that are such marginal category members that one can in practice be pretty sure the reference is motivated.

I think I agree with you that sometimes "affirmative action is racism" can be useful as a slogan if you believe many of the same features that make the KKK bad also make affirmative action bad; I guess it would be sort of an opener for the discussion "List the reasons you don't like the KKK; now look and you'll see that many of those same things are true of affirmative action." This seems legitimate to me if it's true.

On the other hand, I think there are a lot of people who hate the KKK for reasons that don't apply at all to affirmative action, yet who might still feel they have to dislike affirmative action merely because it's in the category "racism". Trying to trick them into this is the Worst Argument in the World.

This is kind of tricky since in a lot of cases maybe there are ten reasons someone dislikes the KKK, and affirmative action shares only one of these reasons. So it's not completely dishonest, as it may be an honest attempt to point out the one feature both genuinely share. But it's hardly advisable either, especially if we don't expect the person to be able to keep in working memory that nine of the ten reasons they dislike racism don't apply to affirmative action.

I want to eventually retitle this "Guilt by Association Fallacy" (or something) and rewrite it a bit (with that eighth anti-liberal example I promised sewing-machine) but I'll wait for all the criticism to come in, especially Konkvistador's.