Live data from Hacker News

IQ is largely a pseudoscientific swindle

medium.com

321–330 of 400 posts

Re: IQ is largely a pseudoscientific swindle

#321
post #102

Earlier quoted context omitted.

The problem is not whether the answer is binary. The problem is that the author of the test needs to have at least the same level of intelligence (and probably even higher) than what the test is measuring. Otherwise how can you tell if the answer is correct? So to get a good IQ score one needs to be able to figure out what a test designer (with average IQ) thought the correct answer should be.

> Otherwise how can you tell if the answer is correct? For example: By taking much more time and access to all literature while designing the test, which differs from the test protocol. IQ tests aren't quizzes: you don't need to be "clever" to design them, you need to know which aspect of thought process you want to measure. The value comes from their stability: there are tons of tests based on different methodologie…

I gave examples elsewhere in this thread that illustrate why more time won't help. There is a large difference between solving the same type of problem, only larger, fast, and using abstractions at a higher level.

Do you really think that speed of doing long division predicts ability to solve differential equations?

Re: IQ is largely a pseudoscientific swindle

#322
post #66

Earlier quoted context omitted.

The point Taleb is making is that IQ works well in the lower half of the distribution, but not in the upper half. Which makes sense, considering that questions for IQ tests are created by people with close to average intelligence. Imagine third-graders writing college-level math tests, or beginner chess players coming up with chess problems to separate masters from grandmasters.

> Which makes sense, considering that questions for IQ tests are created by people with close to average intelligence. This isn’t true. Designing, testing questions and norming a new IQ test or updating an existing one is very expensive. At minimum you’ll need a Master’s in Psychology, Psychology or Psychometrics to get those jobs. Look at the Educational Testing Service. https://www.ets.org/research/internship-fello…

Here's somebody with intelligence clearly above average making the same point: https://blog.tanyakhovanova.com/2008/12/my-iq/

Re: IQ is largely a pseudoscientific swindle

#323
post #66

Earlier quoted context omitted.

The point Taleb is making is that IQ works well in the lower half of the distribution, but not in the upper half. Which makes sense, considering that questions for IQ tests are created by people with close to average intelligence. Imagine third-graders writing college-level math tests, or beginner chess players coming up with chess problems to separate masters from grandmasters.

>questions for IQ tests are created by people with close to average intelligence Is this really the case? I would expect the opposite to be true, that people with high IQs and specifically organizations like MENSA that select based on IQ would have a vested interest in being involved in that process.

Tanya Khovanova on IQ and MENSA: https://blog.tanyakhovanova.com/2008/12/my-iq/

Re: IQ is largely a pseudoscientific swindle

#324
post #58

Earlier quoted context omitted.

Yet IQ test does not measure ability to acquire and apply knowledge and skills.

Thats pattern identification and solving, which is the bulk of what most IQ tests are. So, it kinda does.

IQ tests measure ability to correctly guess what test creators meant. Many of those questions have multiple correct answers, especially the more "complex" ones.

Re: IQ is largely a pseudoscientific swindle

#326
post #239

Earlier quoted context omitted.

The jung/MB tests are probably the most unscientific drivel that is in wide use. It is insulting to human intelligence that is given any significance at all.

I've often wondered about this, because I can't see how simple categorisation across 4 axis can be either scientific or unscientific. It's just categorisation, it doesn't make predictions. I enjoy reading the different type descriptions, unlike horoscopes (which are basically barnum phrases) they do seem to describe me with a high degree of accuracy. Then again, I stopped at this point, and never dived deeper into al…

>It's just categorisation, it doesn't make predictions.

If it doesn't make predictions, it isn't science.

The Big Five model of personality is reasonably scientific, because it has some predictive power. If you give someone a standardised test to measure their Big Five traits, you can make better-than-chance predictions about their behaviour in the real world. If they're high on openness to experience, they're more likely to vote Democrat. If they're low on agreeableness and conscientiousness, they're more likely to have a criminal record. The MBTI has a tiny little bit of predictive power, but it's grossly inferior to most psychometric instruments with similar aims.

IMO the biggest problem with the MBTI is that it presupposes false dichotomies. Most psychological traits are approximately normally distributed, with most people being somewhere near the average. It makes no sense to categorise people as either "Thinking" or "Feeling" while excluding the middle. Knowing that someone is ESTJ or ISFP tells me almost nothing about them; they might be almost completely average on every domain, or they might be extreme on one of them. This false dichotomy also gives the MBTI very poor reliability - someone who is very slightly above average on extraversion might be slightly below average next week or next month.

Re: IQ is largely a pseudoscientific swindle

#327
post #247

Earlier quoted context omitted.

I scored very well on the GMAT and it’s really testing how diligently you prepare for it, and that is the predictor of success, not the content of the test itself, IMHO.

I’m sure you did improve. Practice effects exist. They’re also limited. Just as I will get faster if I train to run the 100m or run a marathon you can get better at the GMAT, LSAT etc. But there’s a plateau you hit where you can’t get better. If you try to take someone from 50th percentile to 99th on the GMAT in the majority of cases you will fail. An increase of 0.26 standard deviations isn’t nothing but on the GMAT…

Yes and no. You can certainly practice GMAT style questions directly and get better at them that way, but the real trick is to meta up a level and see that they follow a pattern, at which point getting a very good score is easy. I think the test is really testing for that (I guess deliberately or why put the pattern in in the first place?). Or maybe it is really testing for the personality type that likes to prepare in advance rather than winging it, that would obviously correlate well with success in any field.

Re: IQ is largely a pseudoscientific swindle

#328
post #148
post #97

reading the Wikipedia article, it says that scientific consensus among surveyed psychologists and educational researchers is that IQ is a valid and accurate mesaure of the important components of intelligence https://en.wikipedia.org/wiki/Intelligence_quotient#Reliabil... If you're going to challenge that scientific consensus you better have something better to bring to the table than "it's being used by racists" and…

Wikipedia is not a totally reliable reference when researching non-quantitative information, especially in area with active stakeholders that likely edit wikipedia. And I would not trust "scientific consensus" in this instance given psychology's recent well documented reproducibility crisis.

Individual anonymous comments on HN are even less reliable.

Re: IQ is largely a pseudoscientific swindle

#329

Earlier quoted context omitted.

The frustrating thing from the perspective of an individual differences researcher is that Taleb is not at all correct about IQ becoming less predictive of performance at higher levels! Studies pretty consistently show that IQ is a valid predictor of performance all the way up the scale, so that, for example, someone with an IQ of 145 is likely to perform (at school, on the job, etc) better than someone with an IQ of…

> The frustrating thing from the perspective of an individual differences researcher is that Taleb is not at all correct about IQ becoming less predictive of performance at higher levels! Studies pretty consistently show that IQ is a valid predictor of performance all the way up the scale, so that, for example, someone with an IQ of 145 is likely to perform (at school, on the job, etc) better than someone with an IQ…

That's not really addressing the point, is it. IQ correlates well with measures of performance. Taleb's point, if we are to generously interpret his screed, is that some classes of valuable mental performance are very hard to generate metrics for or even define at all, and therefore, you would expect them to correlate with IQ less or not at all.

For instance the chart that claims IQ correlates well with military performance and poorly with "creativity". But how is creativity being measured? It's going to be far easier to measure the performance of a soldier than a creative person, the latter isn't even a clearly defined category, lots of people consider themselves to be creative but generate unpopular dirge. So all he's doing here is dissing people whose performance is more easily measured by nature of their job.

I have to admit the article annoyed me. I actually did believe him when he said IQ doesn't correlate at the upper reaches - I have even studied psychology but it was a long time ago, and wondered if my knowledge was just out of date or forgotten. So I'm glad I read these comments and learned that it's not the case. But it mostly annoyed me because I thought it would be an interesting addition to the ongoing story around poor replication rates and practices in some branches of science. But it seems actually to be some sort of political article targeting conservatives. Even with the wrong parts of his article, I do agree with some of his overall points, but when I got to the end and saw it was tagged "Racism" and "Alt-Right" rather than science related terms, I realised I'd been reading a bog standard rant of the sort that might appear in the Guardian or Vice - just dressed up in a scientific and haughty tone, to make it sound authoritative.

The problem with his overall point is that whilst yes, an observer of history might have wondered how Britain went from primitive tribes conquered by the Romans, at the outskirts of known civilisation, to owner of the largest empire the world has ever seen, that was over a very time long period and the average Roman person would have been quite right to stereotype the population of the north at that time - the changes that invalidated those stereotypes took place over far longer than a human lifetime. People using IQ scores to measure certain types of performance today would be entirely valid to do despite the correlation with race, if they were looking for workers whose performance does correlate with IQ. The fact that not all categories of valuable work may be measurable or correlate doesn't detract from that fact.

Re: IQ is largely a pseudoscientific swindle

#330

Earlier quoted context omitted.

The frustrating thing from the perspective of an individual differences researcher is that Taleb is not at all correct about IQ becoming less predictive of performance at higher levels! Studies pretty consistently show that IQ is a valid predictor of performance all the way up the scale, so that, for example, someone with an IQ of 145 is likely to perform (at school, on the job, etc) better than someone with an IQ of…

He addresses that, the issue is that you aren't identifying "performance" but rather "performance in the sort of test taking environment that exists at school". His entire point is that, while IQ may identify good, obedient candidates for BS jobs, it is not broadly applicable as anything like a general metric for life.

Does anyone use it like that though? Isn't he arguing with a straw man? Who is using IQ to allocate funding to pop singers? IQ scores aren't used that much and only ever for predicting success of jobs where success is easily measurable (which is what Taleb seems to mean by "obediency-based BS jobs").
Post reply on HN