Update (from sometime in 2015, when I first wrote this as part of a school essay): You know what’s depressing? Stumbling across something that says in two sentences what you already said in a whole blog post. Whereas I need 1700 words to get my point across, the brilliant Paul Graham puts the same idea in a mere footnote.
Spoiler alert, here’s the conclusion of this whole post:
The intellectually honest argument for not discriminating between various types of people… is not that everyone’s the same, but that it’s bad to do wrong and hard to do right.
Such elegance. -sigh-
What categories do
In and of themselves, categories are useless. It’s only when you place two or more categories side-by-side that they become useful, because then we can think about the contents of each category in new ways.
This rest of this post will explain these two sentences in detail.
(How thrilling! But it is interesting, I promise, at least after this next paragraph).
When I say categories are useless “in and of themselves,” I mean we should stop thinking of categories in terms of how well they describe their contents.
Categories can never provide absolutely accurate descriptions, because they’re a way of assigning relative descriptions (specifically because they describe groups of things relative to other groups of things). In doing so, categories make it easier to think because they are a way of assigning names to groups of things.
Since names always abstractly represent a broader group of concrete items, it is pointless to debate whether or not they are absolutely accurate. Instead, we need to evaluate their usefulness based on what they say about similarities and differences between groups.
So categories are only useful when you need to recognize and describe relative differences, not absolute ones. (Everyone who made it through this paragraph is now thinking “obviously.” Just had to make sure we were on the same page about what a category is.)
Abstract generalizations vs. concrete descriptions
To see why categories describe relative differences, it’s important to recognize that things are neither completely abstract nor completely distinct: they are different, but only in shades of relative difference.
At some level of complete abstract generalization everything is one; but at the opposite extreme of pure concrete description, all things are completely discrete. This is why any method of abstraction is only useful when it retains some concreteness. Without concreteness, there would just be one name for everything.
The inverse is also true: any method of concrete description is only useful when it retains some generalizability. Without generalizability, our thinking power would be constrained by the constant need to describe everything precisely. For instance, you couldn’t just say “could you please move that chair” and assume the other person knew which item to move. You would have to say something like “that device for sitting upon, consisting of four legs and a straight back, made out of polished mahogany.” Of course even that might not be specific enough—let alone all the specificities that would have to go into describing what you meant by “could” and “you” and “move.”
Intelligence means making accurate generalizations
In order to think intelligently, you have to avoid the two extremes of either completely abstract generalization or completely precise description. I evaluate intelligent thinking—coming up with intelligent thoughts—according to the criteria of accuracy and usefulness. I define “intelligence” as being able to comprehend the truth of things. (I know that’s vague, but I read it in Amusing Ourselves to Death and liked it.)
So to think intelligently, which means to come up with thoughts that are accurate and useful, you must combine the concrete with the abstract. In other words, you must combine the narrowly descriptive with the broadly generalizable—in a way that describes reality accurately (because it describes the truth of things), but not so accurately as to be useless (and unwieldy).
However, it’s also true that all descriptions are generalizations (to paraphrase Wittgenstein, all descriptions involve words, and all words are abstractions). The question is whether the generalizations are accurate enough to be useful.
Categories make it easier to think intelligently
Categories simplify the complex relationships between things. They provide a level of abstraction that goes beyond the individual items that each category contains, while remaining below the level of extreme abstraction that would put everything in the same category.
In this way, categories are a middle ground between complete separateness and complete oneness. In the same way, intelligent thinking represents a middle ground between completely precise descriptions and completely abstract generalizations.
When is simplification bad?
Simplifying complex relationships between groups is not inherently bad, but it can have bad consequences when categories devolve into human stereotypes. When you define something according to the category in which you place it, you often end up assigning all the characteristics of that category to the thing, even if they’re not a complete match. This is what most left-wingers are (correctly) against—having your characteristics, and hence the opportunities you are offered, determined on the basis of your membership in a certain category. At a certain point, the fact that things are put into a category will cause those things to become more like their category.
I think the root of the arguments for and against stereotyping is the tradeoff between accuracy and generalizability. Some people defend their use of stereotypes by saying stereotypes wouldn’t exist if they weren’t at least somewhat accurate. The more sophisticated defenders may also claim that stereotypes make it easier to predict people’s behaviour, so one can make decisions faster—perhaps going so far as to say that stereotypes are a default human heuristic.
Others oppose stereotypes on the grounds that people are all distinct and should not be pre-emptively judged according to what the people in their stereotypical “category” might tend to do. Implied here is that categories/stereotypes cannot capture all the nuances of distinct items (in this case, people), so you’ll be more accurate if you evaluate things on an individual basis. According to this argument, accuracy trumps convenience. You should judge people as accurately, and therefore as individualistically, as possible, even if it’s more convenient to judge people as a group.
Accuracy vs. convenience
I don’t think the “accuracy” objection is a good enough reason to oppose stereotypes. What if sacrificing some accuracy in order to gain predictive powers is a valuable tradeoff for some people?
For instance, let’s say scientists prove that men are, in general, better at math than women. Math departments could streamline their admissions process by only accepting men. They would make it easier on themselves by not having to even consider the applications of women, since on average those women will be worse than the pool of male applicants. (Assuming admissions are based on merit alone. Of course the pool of women applicants might be stronger than the men since they pre-filtered themselves through the high-school math stream, but let’s also assume the male and female pools of applicants are representative of the national distribution of mathematical ability).
Sure, you can argue that it’s stupid for the school to do this, because maybe it will miss out on some outstanding female mathematicians. But it’s not stupid if the school understands the tradeoff and rationally chooses to continue. (The tradeoff being that it’s sacrificing the potential to admit a few great women in exchange for the time the admissions committee saves by not having to evaluate all the women’s applications, almost all of which will be rejected anyway.) Is this the end of the argument for why people should be considered on their individual merits, instead of the characteristics of the group they fall into?? The debate boils down to “you should do it because it’s more accurate” “well I don’t care about accuracy, I care about saving time and mental effort” “ok then, never mind”? I think not.
Using categories wisely
The argument against stereotypes needs to rest on more than just whether or not they’re useful. Regardless of the relative accuracy of the stereotype, we shouldn’t stereotype people because it’s unfair. For that reason, my perspective on categories is different when it comes to people (as opposed to things). Categories always involve tradeoffs between accuracy and generalizability, so there are always consequences to the categories you choose. But when you’re categorizing things, those consequences are only between either accuracy or generalizability. When it comes to people, the consequences of losing accuracy are worse because there’s a dimension of fairness involved.
I originally defended the use of categorization from an easiness perspective: categories make it easier to think better thoughts, because they simplify the process of comparing things. By “better,” I of course mean thoughts that are more useful and more accurate than what they would be if you had to describe everything by its particular unique characteristics, all the time. But the convenience argument doesn’t trump everything—when categories are applied to people, they are often used unfairly, so the unfair practices cannot be defended simply because they make some people’s lives easier. In short, fairness is more important than simplicity.
So the way to go about categorization is to recognize that things are neither completely the same nor completely distinct. “All is one” results in too much abstraction and not enough accuracy; “everything is distinct” results in too much concreteness and not enough usefulness. Either of these two extremes can be used to describe anything—but the truly intelligent (and therefore truly accurate and useful) way of thinking requires finding the right balance, and method of categorization, for the task one faces.
It’s particularly problematic for people because its the privileged group that’s applying the categories (with consequences).
 Actually, categories describe relative differences between groups of groups of things, since they describe groups of names, and names already describe groups of things.