Well, that hearing was interesting. Obviously, Sessions says he isn't a racist, even though he was denied a federal judgeship on the grounds that he is a racist.
Note my language. Racist or not racist. Two choices. It ain't that easy, kids.
Sessions denies that he is a racist, sincerely believing that he cannot be characterized as such. In all likelihood, he has never participated in a cross-burning or lynching. (If evidence surfaces of that, well...) If that is the standard for what constitutes a racist, then anyone who doesn't fit that bill must, by definition, be "not racist." Therein lies the problem of the dichotomy.
Buzzword time! "Precision." You think you know what that means. You don't. In social science terms, it refers to how fine-grained your measure is. There are a hell of a lot more than two categories here. In the National Election Studies survey, for example, we ask people to rate their "feelings" towards African-Americans and other groups on a 100-point "feeling thermometer," with higher scores being "warmer," and more positive. In 2012, 22.7% put African-Americans at precisely 50, meaning neutral. 15% put them at the very top-- 100. 1.1% admitted to really hating African-Americans, putting them at 0. Arithmetic mean: 67.39. By contrast, 19.2% put whites at 50, 14.2% put whites at 100, 0.7% put whites at 0, and the arithmetic mean was 72.13. So, people's assessments of African-Americans were somewhat more negative, on average. Shocker, right?
Of course, what do those numbers mean? We don't really know. What we want to know is how people behave. So, think of it this way: you are asked to look over resumes, and call people in for job interviews. Two identical resumes. One is from an applicant with the first name, "Emily," the other is from an applicant with the first name, "Lakisha." Everything else is identical. You should have an equal probability of calling either. Would you?
This has actually been studied experimentally. "Lakisha" is less likely to get a call for an interview. Why? The name is more common among African-Americans. But, the effect is probabilistic. Even at the individual level, there is a lower probability that you will call Lakisha. And that could change from day to day. Based on what? Daily experiences. The fact that it varies over time, across individuals, and does so probabilistically, though, means that we cannot dichotomize racism. It makes no substantive sense, from a social science perspective, to classify people as "racist" or "not racist." Empirically, we want to measure peoples' attitudes towards race, and what makes people more likely to make certain types of decisions, but it isn't a dichotomy. Some people are more likely to ignore Lakisha's resume than others, but it is probabilistic.
The dichotomization of racism makes no social scientific sense.