In an ideal world, what one wants is equality both in procedure and in outcome -- we want people to be treated the same, and we want them to end up getting the same level of goods. But we're not in an ideal world, and the historical record seems clear that if we implement race-blind policies, we don't get race-neutral outcomes. Instead, we get harm done to minorities -- fewer blacks in the top universities, fewer blacks in important positions in companies, fewer blacks owning their own businesses, fewer blacks voting, fewer blacks living in better neighborhoods, and so on.
Those are bad outcomes, and presumably we think they are outcomes which wouldn't occur in an ideal situation. We thus want to make those outcomes stop happening. That requires abandoning race-neutral treatment for a time, in order to achieve race-neutral outcomes. The manner of abandonment, as Adrian has rightly stressed, can be very different in different situations. In some situations, we admit to universities minority applicants whose manifest academic record is less strong than the average admittee, on the theory that the latent intellectual capacities of the minority applicant may not be well-expressed by the academic record, since the applicant has had to deal with an inferior educational background and greater extra-academic obstacles. In employment, we treat as prima facie evidence of discriminatory hiring practices simple numerical disparities in racial makeup of a company's employees. In government practice, we make available business startup funds to minority applicants that aren't available to the general population.
I don't see why it would be surprising for such actions to reduce discrimination in outcome. It seems to me that the reduction of discrimination in outcome is of greater importance than the reduction of discrimination in procedure and attitude, both because people frankly suffer much more from discrimination in outcome than they do from discrimination in procedure and attitude and because discriminatory attitudes typically piggyback on discriminatory outcomes -- so long as, and largely only so long as, blacks are poorer, segregated, and less well educated, they will be treated as socially inferior by many people.
The thought that perfect legal neutrality in treatment will lead to neutrality in outcome seems to me to be a simple historical fantasy. If you take all the cultural goodies and give them to one person, and then tell everyone "OK, now play fair, and don't take more than your share", then the one initial beneficee is just going to keep doing better -- in fact, better and better over time, as he reaps the benefits of his starting capital.
There are, of course, legitimate questions to be asked about the propriety of particular implementations of the general strategy of affirmative action, and there is a genuine need for a careful statement of the underlying rationale of affirmative action (I hope I've at least gestured in that latter direction here). But I admit to finding it distressing that such a simple-minded and misguided critique of affirmative action as the "whites only scholarship" represented could win any support at all from intelligent people (a CNN poll showed that over 65% of people thought it made a good point). I'm left thinking that we leftists have failed terribly in making our case properly to the American people.