3 ms·

Equating SATs with IQ is a bad (and incorrect) conflation. An SAT is (ostensibly) a measure of learned skills. I'm originally from NZ so have never done SAT,

by olliej 5y ago

Equating SATs with IQ is a bad (and incorrect) conflation. An SAT is (ostensibly) a measure of learned skills. I'm originally from NZ so have never done SAT, and I was in the last year that had what were called "University Bursary" exams which serve the same university entrance purpose. The exams seem much more complicated than SATs but then NZ has far fewer people so manual grading is achievable. My experience with standardized testing is through my partner, and what I learned while they were preparing for MCAT and the general postgrad one (acronym???) was: Achieving a high grade in the US standardized testing system seems to benefit hugely from tutoring/training on the actual tasks of taking the tests, including a wide variety of "tricks" that serve no practical purpose outside of the testing system. There is immense time pressure, far more than I believe that there should be (and again private training helps with that as well). Because that training costs money it has the functional effect of creating a system the has a bias in favor of people who can afford it. Given having money already provides considerable benefits do educational proficiency (through many factors), this just reinforces that. So I can understand why there is concern about bias, but I think dropping SAT, etc entirely is far too extreme a response. An IQ test is (again, ostensibly) a test of innate intelligence. IQ testing as a concept has been fairly comprehensively debunked numerous times at this point due to a variety of effects that impact test results (availability of reading material while young, nutrition, cultural background, and so...). Starting from an assumption that IQ is real, then a test that accurately measures IQ should not vary in such a way, if it does, then the test isn't measuring IQ. That's where the IQ concept gets into trouble: no one has been able to develop a test that isn't influenced by factors that are not "innate".

clove 5y ago
You have bought into a logical fallacy - a common one pushed by the many people who have agendas for whatever reason to dismiss IQ (and thereby all psychometrics, in practice). If you are really interested in understanding IQ, which is not a question of being "real" but a metric created to measure the g-factor, you can educate yourself via unbiased sources in the field of Intelligence. Just to contribute a bit to your education, though, let me point out that the examples of "debunking" you have listed confuse within-individual differences over time with between-individual differences, the latter having been shown to be extremely stable.
- Apocryphon 5y ago
  On that note, there was an interesting comment on psychometrics here: https://news.ycombinator.com/item?id=29798887 https://news.ycombinator.com/item?id=29798887
- olliej 5y ago
  I’m unsure how to respond to this - I explicitly did not dismiss iq, but the testing. If the testing has any bias then the results of the testing are not comparable between people. The net effect of this is that any system that has attempted to make “unbiased” choices based on IQ tests as an objective measure have resulted in systems that favour those biases. I don’t deny that people have an “IQ”, “g-factor”, “ISDN”, or whatever. It is very clear that some people are smarter than others[1]. But the biased testing renders is useless for any outside of an individual. [1] to be very clear, I’ve known plenty of otherwise “smart” people who have performed poorly, and vice versa. Generally results in life aren’t purely a result of anything innate.
strikelaserclaw 5y ago
its like the leetcode tests we get get to take for interviews, i can tell you that practice has made me much much much faster at solving these problems than when i first started. I know a ton of people who delibrately practice for the SATs/ACTs, and they improve their score a lot over consecutive attempts.
- olliej 5y ago
  Whiteboard coding interviews should not be requiring l33tcode (you spelled it wrong :D :D :D), and if they are it reflects poorly on the interviewer/company. I’ve only had one interview where I would consider the goal to be in the realms of leetcode and that was imo a terrible google interview question[1]. But even then the code I wrote on the board was not a thing that would compile, it was pseudo code that described the implementation. The point of those interview questions is such that the language should not matter (Junior/uni-hire devs should probably have a couple of real code interviews), but to see how you think, your approach to solving a problem, and your ability to find logic bugs in your solution. Starting from a point of assuming you generally know the language you’re working in (depending on experience), in interviews where I’m the interviewer I try to make it very clear that I don’t care about perfect C/Python/whatever as long as you’re reasonably consistent and I can understand what you’re doing. I also ask about test cases (I want the general behavior and edge cases), and then we work through testcases and see if we find any issues (which you likely will because I’m not going to ask something trivial), and then go through fixing it. I do not consider having a bug in your code in a whiteboard interview to be a problem, and generally expect it. I just want to see you find, diagnose, and fix them if you do have any (see above re testcases) [1] I disagree with google on many of their decisions about whether certain problems are/aren’t “leetcode” problems

by olliej 5y ago

clove 5y ago

You have bought into a logical fallacy - a common one pushed by the many people who have agendas for whatever reason to dismiss IQ (and thereby all psychometrics, in practice). If you are really interested in understanding IQ, which is not a question of being "real" but a metric created to measure the g-factor, you can educate yourself via unbiased sources in the field of Intelligence. Just to contribute a bit to your education, though, let me point out that the examples of "debunking" you have listed confuse within-individual differences over time with between-individual differences, the latter having been shown to be extremely stable.

Apocryphon 5y ago
On that note, there was an interesting comment on psychometrics here: https://news.ycombinator.com/item?id=29798887 https://news.ycombinator.com/item?id=29798887
olliej 5y ago
I’m unsure how to respond to this - I explicitly did not dismiss iq, but the testing. If the testing has any bias then the results of the testing are not comparable between people. The net effect of this is that any system that has attempted to make “unbiased” choices based on IQ tests as an objective measure have resulted in systems that favour those biases. I don’t deny that people have an “IQ”, “g-factor”, “ISDN”, or whatever. It is very clear that some people are smarter than others[1]. But the biased testing renders is useless for any outside of an individual. [1] to be very clear, I’ve known plenty of otherwise “smart” people who have performed poorly, and vice versa. Generally results in life aren’t purely a result of anything innate.

strikelaserclaw 5y ago

its like the leetcode tests we get get to take for interviews, i can tell you that practice has made me much much much faster at solving these problems than when i first started. I know a ton of people who delibrately practice for the SATs/ACTs, and they improve their score a lot over consecutive attempts.

olliej 5y ago
Whiteboard coding interviews should not be requiring l33tcode (you spelled it wrong :D :D :D), and if they are it reflects poorly on the interviewer/company. I’ve only had one interview where I would consider the goal to be in the realms of leetcode and that was imo a terrible google interview question[1]. But even then the code I wrote on the board was not a thing that would compile, it was pseudo code that described the implementation. The point of those interview questions is such that the language should not matter (Junior/uni-hire devs should probably have a couple of real code interviews), but to see how you think, your approach to solving a problem, and your ability to find logic bugs in your solution. Starting from a point of assuming you generally know the language you’re working in (depending on experience), in interviews where I’m the interviewer I try to make it very clear that I don’t care about perfect C/Python/whatever as long as you’re reasonably consistent and I can understand what you’re doing. I also ask about test cases (I want the general behavior and edge cases), and then we work through testcases and see if we find any issues (which you likely will because I’m not going to ask something trivial), and then go through fixing it. I do not consider having a bug in your code in a whiteboard interview to be a problem, and generally expect it. I just want to see you find, diagnose, and fix them if you do have any (see above re testcases) [1] I disagree with google on many of their decisions about whether certain problems are/aren’t “leetcode” problems