1) A selective university might want a system of affirmative action to educate the future leaders of discrete, established communities. "African-Americans" are a large, established, and fairly discrete community. Here the question is whether immigrant blacks have become or are likely to become plausible future leaders of that community—are they similar to native-born blacks for purposes of community leadership?
2) A university might believe that racial discrimination is an important social issue that merits academic study and want to be sure that some people who have experienced that discrimination are represented in its student body. (Sandra Day O' Connor's opinion in Grutter v. Bollinger—the recent Michigan affirmative-action case—endorses such a rationale: "By virtue of our Nation's Struggle with racial inequality,[minority] students are both likely to have experiences of particular importance to the Law School's mission and are less likely to be admitted in meaningful numbers on criteria that ignore those experiences." [Emphasis added.]) Here we'd want to ask whether immigrant blacks are likely to have experienced the same or similar types of discrimination as native blacks.
3) A university might believe that racial discrimination has depressed the grades and standardized test scores of black applicants and wish to correct for this bias in the available data. (O'Connor's Grutter opinion also suggests this rationale: The quotation above indicates that minority students "are less likely to be admitted in meaningful numbers" because of "our Nation's Struggle with racial inequality.") Here, we'd want to know whether discrimination has had a similar adverse effect on the performance of the immigrants as on native-born blacks.
None of these questions are answered or even clarified by any abstract definition of "African-American." And conversely, the answer to the affirmative-action question may not be relevant in other contexts in which race is at issue. For instance, suppose you decided that immigrant blacks don't count as African-American (of course I'd discourage you from putting it this way) for affirmative-action purposes. You could arrive at this conclusion—under rationale No. 3 above—based on a belief that whereas high-school grades and standardized-test scores typically underpredict the potential of native-born blacks, they are an accurate predictor of the potential of immigrant blacks. That position would still not commit you to agreeing with Alan Keyes that Barack Obama—the child of an African immigrant—is not African-American. It would be perfectly consistent to say that although Obama doesn't qualify for affirmative-action preferences, you consider him "African-American" in the context of his candidacy for the U.S. Senate.
Here, the real question is whether Obama's experience as someone whose appearance will have exposed him to his share of antiblack racism (regardless of his cultural or ancestral heritage) is likely to make him more sympathetic to the political concerns of others victims of racism. Frequently, the words "African-American" are just shorthand for such a presumption.
Arguing about who qualifies for a racial label is not only unnecessary, it also flirts with perils best left in the past. For instance, while Harvard's controversy focused primarily on immigrant blacks, to a lesser extent it called into question the identity of the children of interracial couples. Such an inquiry threatens to revive—under a new rationale—the odious 19th-century ideas of "blood quantum" and racial purity.
Perhaps what's most unfortunate about these pointless debates about the meaning of racial identity is their sloppy and volatile mix of the personal and the political. For instance, Keyes (channeling the spirit of liberal multiculturalism to perfection) emotes: "My ancestors toiled in slavery in this country. ... My consciousness, who I am as a person, has been shaped by my struggle, deeply emotional and deeply painful, with the reality of that heritage." This might be a moving first sentence in an autobiography, or an important revelation to one's psychotherapist, but in the context of national politics, who cares? Subjective accounts of personal identity are at best distractions from more tangible and objective issues of racial injustice—such as employment discrimination, residential segregation, and an often racially biased criminal justice system. At its worst, the narcissism of identity politics threatens to mire the struggle for racial justice in intractable conceptual debates and ineffectual emotionalism.
The correct definition of racial identities seems important to a lot of smart people. But that doesn't make it at all important in fact. Maybe if we argued about it long enough, we could all agree on an omnibus definition of racial identities. And maybe there are angels, they do dance on the heads on pins, and if we thought about it long enough, we could figure out how many could crowd on before one would topple off the edge. Because nothing whatsoever depends on the answer to the latter inquiry, we've all quite reasonably stopped caring. I expect the angels will forgive us. And soon enough the nation's blacks, whites, Latinos, Chicanos, Hispanics, Asian-Americans, Orientals, Negroes, colored people, and African-Americans will thank us, if we stop caring about the terminology and definition of races and get on with the important work of fighting racism.