Why not diversify the finish line?

and, who shouldn't fight for equality?

I’m trying to get back to a regular publishing schedule; my idea is to treat this as more of a blog, more willing to scribble down what I’m thinking and see where it takes me. In that spirit… I’ve had some thoughts kicking around on what I think of as niches in political ecologies. Which is just sort of a pretty way of saying, “what if we thought about society like a machine?”

Fig.1—Contemplating The World as an object apart from the self.

These thoughts have latched onto a couple of pieces in The Argument: the first, a conversation on the failures of the Biden administration between Matthew Yglesias and Jerusalem Demsas which examined his tortured approach to diversity in government; the second, a piece by Kelsey Piper on the bizarre politics of standardized testing, entitled “You can’t diversify the finish line.”

In both of the cases examined, there’s a maybe cynical but probably well-intentioned effort to address the fact that a sample of the population selected for merit is less diverse than the population as a whole, indicating something in the system causing inequality—and that attempt backfires. What can we learn from these failures?

The Whole and the Sum of its Parts

President Biden was unusually explicit about his intention to have his administration reflect the makeup of America, or to correct for the theretofore over-representation of white men. He’d do things like “vow” to only consider women for Vice President, or black women for Supreme Court nominees. This ended up causing a bunch of problems. Matt Yglesias argues Biden seemed cornered into picking Kamala Harris as a running mate, a politician he ended up working really poorly with, and that this relationship set the stage for her loss in 2024. You also have a pattern of secretarial nominations apparently based on the ideal of diversity in government:

Matthew Yglesias: And Biden would also do things, you know — it was Marcia Fudge, said she wanted to be agriculture secretary… because, actually, the Agriculture Department runs the food stamp program… And that’s a very important issue for, you know, the urban poor, a constituency that she cares about a lot, and that she was tired of, you know, HUD being just, like, the Black job in the cabinet. And then Biden offered her HUD.

Xavier Becerra, who has had a remarkable career resurgence and is gonna be governor of California, but he wanted to be attorney general, which is a job that he was qualified for. But Biden was so invested in the cutesiness of making Merrick Garland attorney general that he didn’t want to give Becerra the job that Becerra wanted. But he did want to have a Hispanic person in the cabinet. So he offered Michelle Lujan Grisham HHS Secretary, which is a job she was well-qualified for, but she didn’t want to be in the cabinet. So then Biden offers HHS to Becerra, which is not a job — he has no professional background in that… Then the White House got really mad that Becerra wasn’t good at the job.

It felt a lot like the Biden administration repeatedly made hires based on the optics of inclusion and representation (because Biden kept saying that’s what he wanted to do), and then treated those people poorly because they didn’t perform to the standard expected of them, usually because they were put in positions they were ill-suited for and not adequately supported by their boss.¹ The obvious takeaway is that you shouldn’t make really ham-fisted appeals to equality if you’re going to be resentful about it and take it out on your colleagues or get catty with the press.

But I think there’s a second, more generalizable lesson in something Jerusalem says midway through the conversation:

Jerusalem Demsas: …[W]hat’s so crazy about this is, like, there’s obviously an internal push by some, you know, high-ranking minority political bodies, whether it’s, like, CBC [Congressional Black Caucus] or CHC [Congressional Hispanic Caucus] or various members of, you know, the political establishment that’s tried to promote diversity in the ranks, that, like, you want to put forward your people in order to make sure that there’s still an impetus to diversify.

Like, you have to have, like, political groups that … care about this in order to, like, create pressure for people to actually consider identity in some form or fashion.

In other words, it’s not so much that we shouldn’t care about diversity among principals of the administration—it’s that the White House shouldn’t care about that. The president is selecting candidates for high office from a pre-existing pool—basically, congresspeople of good reputation. He needs to make selections which will lead to these offices being executed skillfully, and which will work together as a cabinet. Adding a third condition to skill and interpersonal chemistry definitionally means giving at least a bit less consideration to those first two conditions, which is a problem because those are vital for administration success and the third is not; it’s possible to deliver good government for members of minorities you don’t belong to.

But interest groups know that having a seat a the table is a pretty sure bet of having their interests kept in mind, so they naturally work to elevate members of their group to prominence, and pressure the White House to consider them. In a sense they are correcting for the disproportionate chance of not being represented caused by the skew in Congressional demographics compared to the general population. As a result (as Yglesias recounts with prior instances of inclusion of Jewish and black Americans in government) you do see an increasing diversity among candidates for high office—but none of the weird bind that seems so unique to Biden.

The generalized lesson here seems to be that different parts of a machine do different things, so in thinking about how we can improve the machine, we should avoid committing fallacies of composition (a quality of one part must therefore be a quality of all parts of the whole) or division (a quality of the whole must therefore be a quality of each part). The best way to get a machine to produce the outcomes you want is not to bend every piece of it around that goal. Even if the president himself earnestly believes that goal to be laudable, unintended consequences can come about if he takes it as his job to fix it.

But… isn’t it kind of weird to conclude, effectively, that sometimes you need to ignore inequality to reduce it? And if you’re in the system, how are you supposed to know when you’re best off pretending you don’t care about diversity?

Part of the Problem, and the Solution

I think Kelsey Piper’s piece, which follows many others doing important service in covering the politics around education, shows us this conundrum in action.

The problem is that we see unequal rates of performance between race cohorts; something must be causing that inequality, and we know it isn’t like, the inherent superiority of certain races. Many have jumped to the conclusion that the test itself is unfair. The solution which backfires, in this case, is removing standard testing as a consideration for admission into advanced high schools or colleges. This doesn’t do much to help if it’s the case that the tests do not cause, but simply measure, the inequality (in fact, most of the arguments that the test are unfair boil down to agreement with this: the tests “reify” and thus “perpetuate” educational inequalities).

Black kids in America are much less likely to be reading at grade level. They are much less likely to take advanced math classes and less likely to take and pass AP tests. But taking away the AP tests wouldn’t mean they knew the math. And taking away the SAT wouldn’t teach them how to read.

If this bothers you, you are — I hope and assume — bothered by the actual reality of kids’ preparedness, not by the fact that somebody measured it.

This is the same the perverse logic progressives readily criticize when it comes to the French refusing to gather statistics on race in the name of universal human values. Looking away from a problem doesn’t make it disappear. How did we get here?

The war against gifted programs has not led to any educational improvements… And it’s politically a disaster… But it keeps popping up, I think because the discomfort over these disparities is so profound and the easiest path to relieving that discomfort is simply hiding the evidence.

I think there’s definitely a lot of this—there’s a reflexive feeling that if a test shows unequal outcomes, the test itself must be flawed and ignored, that any unequal outcome is inherently in itself wrong; that’s the sort of discomfort Piper is describing.

But I think there’s a well-intended, psychologically simpler explanation for a lot of this stuff: “It’s bad that there are racial inequalities in who gets to college → so we should stop making there be racial inequalities in who gets to college → so if I am making there be racial inequalities, I should stop!”

Fig. 2—Asking what one can do for one’s country.

Again, the problem is that a sample of candidates needs to be selected for merit, in this case so that they are put on a path to academic success. Again, the selection process produces a sample unreflective of the makeup of the population—but because the test is fair, we know that the problem is upstream of it; “students with good odds of success on the test” is not a pool that includes many of the ones who face various material disadvantages.² So when the person who worries about racial inequality says to themselves, “How can I be part of the solution? What parts within my reach perpetuate the problem?” and they thus reach for test-based admissions, they fail to solve the actual problem—instead, they either send possibly ill-prepared kids to schools which aren’t right for them, or they cause other actors to compensate, which in this case often means falling back on less fair selection standards (other measures of application quality more sensitive to economic inequality).

So: Know your place…?

In both these cases, the problem is that we’re seeing a skew in who finishes the race, because something is holding some of the runners back—the sample selected does not reflect the population selected from. In both cases, the attempted fix is to “diversify the finish line,” to change the metrics or manner by which selection is made. But that leaves unsolved the initial problem, and degrades the outcomes of the process.

In the education case, things are arguably even worse—it’s an attempt to erase the finish line, which obviously actively makes the problem more difficult to solve. We need to know how students of all backgrounds objectively perform in order to diagnose what’s going on when they see diverging placement down the line.

What you need to do is attack the root of the problem; if you’re not positioned to do that, you risk getting in the way.

Or… do you? Because in the first case, the interest group caucuses sure seem to be able to diversify the finish line without causing much trouble, or to do something like it, by making their successful members more salient to the selector—more obviously a good pick. But they don’t get picked if the selector decides they’re not right for the job or just plain doesn’t like them, which avoids the trouble Biden faced. What would be an analogous intervention in the test-taking case? Dedicated resources or community efforts to support the academic performance of disadvantaged students? Hey, that reminds me of something else in Kelsey’s piece…

For one thing, as one Black Bronx Science graduate pointed out, we had more gifted and talented programs in middle schools. Those nurtured and challenged the brightest students and set them on a path for the competitive schools. When those pathways vanished, the share of Black students in the exam schools started to collapse.

So—no, you don’t need to attack the root of the problem itself, necessarily. Sometimes a kludge is really the best way to keep a machine working; a band-aid is all you can do for some wounds. What is the case is that “fixing” parts of the machine that aren’t generating the problem isn’t gonna stop that problem-generation. What’s more, if that part of the machine has another loadbearing function, trying to compensate for the problem risks having it fail altogether. So if you’re part of the system, concerned that your reification of the problem needs correction, you have to be aware of the possibility that that’ll cause you to do a worse job of what you’re paid for. And that has consequences for the people you’re responsible for. In the case of the constant churn of politics, that’s fine; every administration eventually fails. But in the generational work of shepherding children through education, the fallout of hasty mistakes persists.

The timescale that educational reform works on is an example of what encourages overcorrection—people just won’t wait around to see real change, they want to see action now. So, sound leadership is needed to expedite real change. But that work needs to be attentive to the responsibility for those who rely upon you, and who will eat the cost of your failure—whether that’s the reputation of your caucus, the success of your administration, or a generation of students.

That responsibility, and urgency, shows why the kinds of additive adjustments that can be piloted is often preferable—it can often be accomplished more quickly, and gives one the space to work creatively. It also isolates the damage when experiments backfire, which itself might be part of the solution: note that in the presidential case, the onus is on Congressional groups to make good recommendations so that later recommendations are taken up. Consider whether your fixes incentivize that sort of accountability, or actually work against it.

So, no, you shouldn’t just ignore inequality—far from it, the difficult process of true and lasting reform requires extraordinary commitment. But we all should have the discipline to stop and determine what actually needs fixing, and not let ourselves get pushed into short-circuiting the basic mechanism by which the process works— and whatever else, not to screw around with the tools by which we diagnose success!

---

AI Disclaimer: I made this connection and wrote up this argument. After I had drafted my take, I batted my thoughts around with Claude to stress-test my analysis, and made changes based on what I felt were revealed overreaches in the original, points my argument didn’t really support—I tend to drift into grandiose generalizations given enough time at the keyboard. Neither the words nor the structure of the piece are generated by AI; the analysis is influenced by it inasmuch as hashing it out with a friend would have influenced it.

1. These picks weren’t necessarily bad in themselves—Justice Ketanji Brown Jackson seems to me to be doing as good a job one could on the Roberts court. But even in those cases, this process threatened to unfairly tarnish the reputation of the nominee—who in all these cases did in fact work their ass off to be at a place in their career that justified selection by the White House for key office, and deserved better!

2. Ofc this is an empirical question; it’s possible that test results are skewed by a bias against black and Latino children, in which case we should absolutely change the test! The problem with judging its fairness by outcome, however, is that it’s easy to design a test that produces the metrics you want to see but for the wrong reasons. You have a replica of the problem in miniature: You need to understand what the function of the test is and how to carry it out, not just what you’d like the outcomes to be.