Hacker News Reader

Comment by lisper | Hacker News Reader

I think you are underestimating the extent to which standardized tests can be (and have been) biased.

https://www.nea.org/advocating-for-change/new-from-nea/racis...

That's soft bigotry of low expecations. All you have to do is look at data that includes Asian Americans, which is always conveniently omitted from these racist narratives. Even non-Americans routinely do better on American standardized tests.

It is not. GP is saying that the "expectations" in question aren't as applicable to potential as they're purported to be.

>Asian Americans, which is always conveniently omitted from these racist narratives.

Ironically, so is the diversity of the Asian American community.

To the contrary, the NEA wildly overestimates it and employs junk question-begging "disparate impact" [0] reasoning throughout. The article is full of stuff like this: "There is a clear correlation, for example, between test scores and property values."

To the extent that society is meritocratic at all and intelligence is heritable (and it is), we should expect test scores to correlate with all manner of measures of success, including property values. Articles like this don't even take that question seriously. They just ignore it. It's proof by repeated assertion. It may be fashionable to insist that this is prima facie evidence of bias, but that is a question of logic and not a difficult one, whatever exceedingly average minds like Ibram X. Kendi think of it.

[0] As a legal concept "disparate impact" is what it is. The law means whatever its authors intend it to mean. But as a matter of logic, it's embarrassing, and beneath this forum.

"Since their inception a century ago, standardized tests have been instruments of racism and a biased system."

Standardized tests were invented in Sui dynasty China, in the early 600s AD, as a way of selecting officials for the imperial bureaucracy. They were invented precisely because they were more objective than the prevailing method of selecting officials - recommendations from the aristocracy.

There is a long history of standardized testing being a means for rewarding merit, instead of more easily corruptible methods of selecting officials/students/etc., such as recommendations. Just to illustrate my point: Do you know why Harvard abandoned standardized testing in 1926 as the sole means of determining admissions? Because "too many" Jews were passing the admissions test. Harvard's "holistic" admissions policy was invented for the sole purpose of restricting Jewish admissions.

I think you're underestimating how biased literally every other possible metric of admissions can be.

The entire concept of an “admissions” department was based on the historical fact that too many Jews were being admitted and too few WASPs were. So they included a “character” criteria and fixed the problem.

Now, too many Asians are being admitted based on test scores. Oh no! To fix this problem, Harvard consulted their history department and included a “personality traits” section. Is it any wonder that Asian students scored low on this?

https://www.nytimes.com/2018/06/15/us/harvard-asian-enrollme...

Oh my. This obsession over race is so bewildering to foreigners like me.

Yeah the SAT is the worst admissions metric, besides all the other ones.

What would you use in its place? You think grades don't have bias?

Interviews, transcripts, (optional) test scores, letter of recommendation, a set of common essays across all schools, and an optional supplemental section or portfolio to showcase any personal achievements not covered by the other standard categories.

Oh wait! That exists — it’s called the Common App, and it’s what most private colleges today use, from the Ivies to elite tiny liberal arts colleges with the largest share of students from the 0.1% that you’ve never heard of, like Pomona College.

They use it because it allows them to ignore standardized test scores and just do admissions based on their own preferences. They used less merit-based metrics because they don't WANT a meritocracy.

Question. Which tells you more about a student:

1. An essay that was probably written by chatGPT then edited by the students parents

2. A test taken in a supervised, controlled, timed environment

1. Would absolutely not get you in anywhere selective

2. Would absolutely not either

I believe the whole concept of “meritocracy” for purpose of admissions is a lie— choosing the criteria to measure against is itself a subjective act.

> I believe the whole concept of “meritocracy” for purpose of admissions is a lie— choosing the criteria to measure against is itself a subjective act.

Let me ask you. Do you also believe that requiring a display of proficiency in mathematics to get into the best schools is inherently discriminatory? What about requiring a demonstration of the capability to understand and complete basic subject matter material in the fields or reading, writing, or scientific literacy?

There’s no shortage of people who meet any of that criteria!

The whole point of a selective college is they have to select from a pool of already qualified applicants. There is no objective measure to measure against when you’re splitting hairs. Even were you to limit it to “objective” requirements like test score and GPA, how do you decide between two students for one spot when both have the exact same scores?

There is no shortage of perfect scores applying to Harvard. And yet a majority, or even a plurality, of any given class of admits didn’t have perfect scores.

There is no shortage of perfect scores applying to Harvard.

1,000 people typically receive a perfect SAT score yearly.

Essays are incredibly biased though! Do you really think that for some reason essays actually have to be written by the person applying, and can't be gamed with money? That an overworked public school teacher is going to write a better letter of recommendation than a private school teacher? That a rich kid is going to have worse extracurriculars, portfolio or achievements?

The Common App is great, but it's not magically less open to bias than standardized test scores.

Not only that, but the person reading the essay is also biased and will select for students who align with their bias. It is a terrible admission metric.

It’s naive to think schools don’t have a system in place for this: separating piles into buckets of test score, ordering by grade, marking a certain number from each bucket as worth another look, then ordering by essay, marking a certain from each group, and repeat on any other metric.

Many schools, selective or not, actually do this whole process — multiple times, with each admissions agent doing a separate order of criteria, to ensure everyone’s application gets read at least twice. The idea being that those with the most “let’s give them another look” across the board are the most notable. Then from that shortlist the debates comparing each applicant, usually sorted by geographic proximity to each other, begin (at Harvard, if you’re from Texas you’re not really competing against New Yorkers for a spot, you’re competing against other Texans for the XX number of Texan spots they usually admit a year).

I did a short stint as a student worker in the admissions office of a very selective college in California (<5% admission rate, but not one many could name off the top of their head), and this is more or less how it worked

> That an overworked public school teacher is going to write a better letter of recommendation than a private school teacher?

LOL. My understanding is the really good prep school college counsellors golf with one or more high-up folks in elite university admissions offices, and get the inside scoop on exactly what they and their peers in other universities are looking for in any given year, such that they can even tune an essay or letter of reference for a given school based on that non-public information and advise students which schools to focus their application efforts on, based on their background & activities.

> letter of recommendation

We've gone full circle.

Transcripts aren't useful across different high schools. Rich kids have more connections for rec letters. Rich kids get professional essay help, and lying kids make up a good story. Kids and parents with lots of time/money/connections on their hands get a portfolio of community service etc built up. I knew these kids in high school with resumes like veteran philanthropists, and it worked.

I think the only good one out of those is the interview.

Is there literally any test in which:

"Students of Color" receive scores in the same distribution as white/asian/hispanic students (same fraction of 1s, 2s ... 35s, 36s on the ACTs for example)

AND

Top scores are meaningfully distinct from the population average? (because the first condition can be trivially fulfilled by having everyone score the same)

What is the evidence that the tests now are biased other than that different groups score differently?

When your history ensures that some groups are socioeconomically disadvantaged, every possible method will be biased. It's unavoidable.

So why don't we focus on socioeconomically disadvantaged individuals instead of focusing on racial metrics?

With affirmative action gone, that's pretty much exactly what's going to happen.

Everyone understands that standardized tests are biased. They are still the least bad way to identify students from underprivileged backgrounds who have high potential to succeed in college.

I am someone, and I don't understand why they are. Do you have a primer handy on this subject?

SAT scores are correlated with wealth. This CNBC article is a decent introduction to the issue, but be aware that anything you read on the subject is likely to be pushing a particular narrative so it's tough to find a neutral primer anywhere.

https://www.cnbc.com/2019/10/03/rich-students-get-better-sat...

It’s funny how that article doesn’t mention the most obvious reason - it turns out intelligent people generally don’t get jobs as janitors, retail workers, or other minimum-ish wage jobs. Those intelligent people then have intelligent babies.

For some reason a large part of our society doesn’t want to acknowledge that genetics (highly) influence intelligence.

Anyone who has lived in the real world knows that the conditions in which you grow up have a massive effect on where you end up. There are entire communities of people who live in deep poverty. It's not because they're all - uniformly down to the last person - dumb. The parents are poor, don't have a high-quality education, work long hours, can't afford tutoring, can't help their children with their coursework, etc.

There are entire countries trapped in poverty. China was desperately poor until just a generation ago. Did Chinese people suddenly become smart for some reason? Which is the next country that will magically go from being filled with dumb people to being filled with smart people?

As I said, just a bit of experience with the real world and critical thought will show that your explanation is nonsense.

> Did Chinese people suddenly become smart for some reason?

Yes, they stopped following strict communism. They pivoted to authoritarian capitalism and got wealthier.

I wasn't talking about different countries. Obviously there are issues that can hold large geographic regions down.

(Basically) no intelligent teenager in America is held down by poverty. My parents were relatively poor, didn't go to college, worked long hours, and I went to a public school and certainly didn't have any tutoring and didn't receive much help with homework.

I did well enough on my ACT to go to a private college for pretty close to free. Even if that doesn't happen you can go to a state school that doesn't need great grades, take out loans for the entire amount, get a degree for a good paying job, and suddenly you're out of poverty.

thank you. I read that article and the linked NEA one. Lots of info on biased roots in history, not much information on the mechanisms that create the bias or unbiased standardized tests as an alternative.

I would think there would be a lot of demand for an unbiased standardized test. All I saw for suggestions was to use GPA instead, which clearly has a lot of bias potential... is there a recommended replacement for the SAT/ACT?

That’s an awfully big article with no actual examples of how the tests are currently, in any way, biased.

They're not biased at all. People don't like them because they're accurate.

Reply on news.ycombinator.com