Hacker News Reader

Comment by andrecarini | Hacker News Reader

My understanding is you're equating `failing a test` to `lacking the relevant skills and knowledge to do a certain task competently`.

The reality is sometimes tests in academia are just not very well made and don't really test what they are supposed to be testing, and that's usually due to multiple reasons like misaligned incentives, staffing shortages and maybe lack of resources / funding.

I don't think the comparison to flight school is relevant enough in this context because it's a too different of a world to traditional academia.

View on HN

I don't buy the notion that tests do not test relevant skills.

In my long career I've noticed a strong correlation between SAT scores and academic performance as well as job performance.

> I don't think the comparison to flight school is relevant enough in this context because it's a too different of a world to traditional academia.

My dad kept his flight school tests for flying all sorts of airplanes. They bear a lot of similarities with the SATs. There's a lot of math in there for things like fuel consumption, wind, maximum landing weight, glide distance, and so on.

For example, one day he was cruising along in his F-86 when the engine failed. he radioed the tower, and they told him to bail out. But he calculated his speed, altitude, distance, wind, sink rate, air templeratur, etc., and figured he could make the field after configuring the airplane for maximum glide. He made a perfect landing, but still got reprimanded for risking his life bringing the airplane back. But he had worked the math and disagreed that it was more risky to bring it in than bail out.

> I don't buy the notion that tests do not test relevant skills. In my long career I've noticed a strong correlation between SAT scores and academic performance as well as job performance.

SAT tests intelligence (aptitude), not skills. Which is why it correlates with job performance, where intelligence can (over some time) matter as much or more than a starting point of relevant skills.

Nothing's perfect, but the SAT tests do an adequate job.

I just checked, and the SAT math section covers algebra, trigonometry and statistics.

Look at this list:

  Quadratic equations and functions (vertex form, roots, discriminant)
  Polynomial operations and factoring
  Exponential functions and growth/decay
  Radical and rational expressions
  Function notation, composite and inverse functions
  Nonlinear graphs and their transformations

A genius student who had never been taught those subjects wouldn't even know what the symbols meant. A mediocre student who had studied SAT-style questions for weeks leading up to the test would likely outperform a high IQ student who last solved those types of problems over a year prior.

Standardized tests can be a great resource for assessing students, but they're not just testing for intelligence. Test-prep courses average increasing SAT scores by about 200 points. That's not because they're increasing the intelligence of the people taking them.

Somebody who goes to take a test on something they know that they know nothing about could be called many things, but genius is not one, even moreso when they're paying for the privilege of taking that test. What is on the SAT is no secret, so people are free to prepare as little or as much as they might like. If somebody can't be assed to prepare for such a critical test, then they're probably going to be the sort of person who can't be assed to do much of anything in life. And the internet has also largely relegated the inequities in access to training quite obsolete. You can get free high quality training materials on everything for free.

You've never met a smart high school student that didn't study? You've also never met someone who became more disciplined later in life?

There's a difference between not studying and just hand-waving away the SAT, which plays a huge role in things like college admittance. And the SAT isn't particularly hard, even back in the days when it was apparently significantly more challenging, for a clever kid. My prep was picking up some old book on it at Half Price Books and leaving it in the bathroom. Worked excellently!

Or they're a student who has to work a part-time job after school. Or they have a long commute. Or they have hours of practice for a competitive sport or extracurricular. Or they have to take care of their siblings, grandparent or sick parent. Or they were told by their parents and teachers that because they're smart and do well in school that they don't need to worry about studying. Or they know their parents can't afford an elite school anyway, so they know they just have to score well enough to get into a school with in-state tuition and noncompetitive admissions.

Two kids who both went to the same school, were told the same things about the importance of the SAT and dedicated equal time to study and preparation can probably say that the difference in their scores is indicative of differences in intelligence.

But any two random students? There are so many factors that could cause students to underperform their hypothetical max score if they had perfect conditions in the months leading up to the test and day of.

We're talking about the SAT, not the gaokao [1]. It's a relatively short test that you can take it whenever you want, you can even retake it if you want, however many times you want. There's minimal memorization required and it's mostly testing basic aptitude. But the overwhelming majority of people are of average aptitude, so the overwhelming majority of people will do "poorly" on the SAT when comparing themselves to outliers.

I don't understand the effort to try to eliminate the reality that innate ability exists. I think we should accept this and work to cultivate it. Just because somebody of average aptitude doesn't mean he might not excel in other ways. The world needs all sorts of people, and I think the ideal system would work to figure out and cultivate these talents in everybody.

[1] - https://en.wikipedia.org/wiki/Gaokao

I know a person who has a pretty modest ability, shall we say. But he still managed to become a multimillionaire as a young man.

There's plenty of time during school hours to study up on the math. My high school was deficient in math, everyone knew that, and I took math classes at the local community college to make up for it.

I never studied for tests in school, not for the SATs, either. I tried that in college, that did not work. The stress of the tests was pretty bad. I realized I had to study hard, and I did, and it was effective at reducing test anxiety.

"A mediocre student who had studied SAT-style questions for weeks leading up to the test would likely outperform a high IQ student who last solved those types of problems over a year prior."

That's why hard questions exist.

> A genius student who had never been taught those subjects wouldn't even know what the symbols meant.

This is pedantic. Take it to the logical next step. Words and letters are symbols too. How would someone who can’t read do on the SAT?

> A mediocre student who had studied SAT-style questions for weeks leading up to the test would likely outperform a high IQ student who last solved those types of problems over a year prior.

Would make an interesting experiment, if it hasn’t already been studied. I would put my money on the lapsed high IQ student though.

Hal Finney was famous at Caltech for skipping class, flipping through the textbook a couple hours before finals, and acing them.

But he was one in a million. I always felt like a moron around him. But Hal was also charismatic and everyone liked him, including myself.

I suspect Hal was a "skin job" alien.

> Test-prep courses average increasing SAT scores by about 200 points.

Perhaps because they finally bothered to learn the material. I am skeptical that "strategy" makes much of any difference.

I guess the question is: would you rather hire someone with poor SATs and god-tier Leetcoding skills or vice-versa?

I’d for sure take the high SAT scores. I think the SAT has more correlation with/is more predictive of work success than leetcode.

Do you also think LLM leaderboards accurately reflect the capabilities of the models being tested? If you do, then I can easily point you to numerous academic papers pointing out the various flaws in many leaderboards (from poorly designed benchmarks like bABI and the original SQuAD, to data contamination, and more).

In that same way, any test, including the SAT and GRE have flaws. They can be gamed in ways similar to LLM leadeboards: test prep makes you better at them. That's one of the main reasons universities moved away from SAT; they were afraid that it disenfranchised lower socioeconomic status students (and it does to some degree). The issue is that the test is positively correlated with success in an undergraduate program, so they threw out the baby with the bathwster. The real issue is that the SAT is not able to distinguish the capabilities among students to the degree it purports to.

And if you want an anecdote to match all yours, the first time I took a GRE practice test, I got a 3 on the writing. Not because I'm poor at writing, but because I didn't really know what they were looking for. After reading a test prep book, I got a 4.5 on my next practice test and a 5 on my final practice test. When I finally took the actual GRE, I got 6 on the analytical writing. Trust me, nothing changed in my writing ability over that time. In fact, I didn't even practice the skill except through those three practice tests. Clearly the test was not capable of determining my real ability to make an argument; it merely tested my ability to adapt my writing to what was supposedly being tested.

Interestingly, the vast majority of universities that got rid of the GRE requirements for PhD programs are not going back on that. Turns out that the students with the highest GRE scores are the ones most likely to drop out of their STEM PhD. [1]

[1]: https://journals.plos.org/plosone/article?id=10.1371/journal...

I took the GREs, I don't recall a writing section.

Anyhow, the questions were all about freshman engineering knowledge.

There are three major parts of the modern GRE: Verbal, Quantitative, and Analytical Writing. You could easily look that up, or ask if you didn't know.

Responding off the cuff without any reflection on the comment you're responding to doesn't move the conversation forward in any meaningful way. It just comes across as disrespectful.

Do you think that LLM leaderboards don’t? Do you think a Llama 3 is going to beat an Opus 4.7 on any leaderboard?

The real issue is that standardized tests disenfranchise lower SES students less than any other metric.

Everyone who takes the SAT has to sit in the same room for the same amount of time answering the same questions. You can’t just pay someone else to take it for you (like essays) or select which difficulty level you take (like going to a prep school with grade inflation), or luck out in who your parents know (like recommendation letters).

Some may have better opportunities to learn the material, but, at the end of the day, you have to actually learn the material. There’s no getting around that.

As your own GRE anecdote shows: A little studying with some inexpensive books makes all the difference. Unless things have radically changed, a couple SAT or GRE test prep books are significantly less expensive than just one college textbook.

Bluntly, the reason SATs are better correlated to college performance than other measures are because of the reasons I mentioned. They strip away most of the privilege of coming from a high-SES family.

You appear to be arguing with a strawman; at least you seem to have ignored a central part of my message:

  The issue is that the test is positively correlated with success in an undergraduate program, so they threw out the baby with the bathwster. The real issue is that the SAT is not able to distinguish the capabilities among students to the degree it purports to.

> I don't buy the notion that tests do not test relevant skills.

> In my long career I've noticed a strong correlation between SAT scores and academic performance as well as job performance.

A test doesn't need to test the relevant skills for that, it just needs to test _something_ that correlates with academic performance and job success.

In my country, we have public exams to get into Uni, with the ones for high demand majors being very competitive, but performance in that exam is not a good predictor of academic performance.

The guy who got into my uni class as #2 in that exam dropped out after a few semesters because he couldn't beat calculus. The #4 took several extra semesters to graduate despite not working/not interning. Several others in the top third struggled through. We had _maybe_ 2 or 3 guys who straight-A'd the entire major.

I myself got in as #17 and still failed a few courses. Thankfully no one cared throughout my professional career.

> In my long career I've noticed a strong correlation between SAT scores and academic performance as well as job performance

On average. However, I've also had the experience that some of the most competent people I've known had rather difficult teens and twenties.

Hiring someone who flunked out of highschool, worked odd jobs for 10 years, then got a diploma and a degree is higher risk, higher reward. They are often times harder workers, unusual thinkers and more grateful for what they have.

My wife and I are perfect poster children for demonstrating the danger in making the assumption that tests are an accurate assessment of someone's mastery of material.

My wife has anxiety, and the time-limited, high-stakes, pass-or-fail nature of a test makes it much, much harder for her to perform well. Outside that context, she's blazingly intelligent, performs very well under real-world high-pressure situations, is extremely diligent at getting any other assigned work done well on time, and has repeatedly written the guidelines for processes and procedures in her jobs.

I'm the opposite in many ways. I generally do somewhat poorly in a classic classroom environment, primarily due to ADHD hampering my ability to get homework done regularly and on time. However, I test amazingly well. I consistently finish in 50-70% of the time of other students, with scores in the high 90% range. (In my jobs, I think I do pretty well.)

Naturally the plural of "anecdote" is not "data", but I've known enough other people who fit both of our molds that I think it's fair to say that basing even a plurality of your assessment of someone purely on tests, especially standardised tests, is going to mislead you at least as often as not, in both directions.

Correlation isn’t causation. Students with test aptitude and ambition will score hire regardless of the test content.

Whether or not tests are bad the fact is that nobody has even come close with something objective that can replace it at scale. Teachers have 30 kids in their class- they actually in reality have no idea what their students can or cannot do.

Reply on news.ycombinator.com