IQ test

[kram1032]

Please describe which scale you are using - the US appears to use a system where 170 represents "genius" level, but the system in the UK uses 130 (or maybe 135?) - of course I think they are equivalent but with different difficulty/scoring.

As said, those tests I (basically my entire class) did back when I was 13-14. I have no clue what ranking system was used. I'm not even sure wether that was even mentioned on the website.
In either case, there simply is no way that not a single person in my class was below average. And I'm not saying they were absolute awesome dudes or something. It's simply that this is statistically pretty much impossible.

IIRC, I once did a Mensa test too and that turned out to be ~140. It's been a while though.
Either way, I don't really consider any of those tests valuable.
They are some sort of educational proxy with maybe a hint of actual underlying intelligence. Nothing more.

[eiffie]

I wish more people would actually look at the test rather than complain about what they percieve the test is. It is NOT culturally biased - it is based on the idea that all intelligence arises from the ability to spot rules/patterns. Yes the number is basically meaningless but compared to others who have taken the same test under similar conditions it does give you an idea of what side of stupid you are on

[cKleinhuis]

i made 113 points in 15 minutes, but clicking the last 10 questions just randomly

[kram1032]

It is very much culturally biased, for language is part of culture and if you do, say, an IQ test in chinese and don't know a chinese word, you'll obviously fail.

Though there are even less extreme examples: Many IQ tests will include wordplays or sayings and either ask you for a meaning or a completion of those.
Even if you're generally proficient in a language, you're very likely to not know all those rarer sayings if you're not born into a culture that uses them.

Also, if you're grown up in an environment with a lot of pattern-matching-style puzzles, you'll likely outperform others just from experience, especially considering that many such puzzles either stem from IQ tests or are used in them as well. That doesn't technically mean you're better in generic pattern matching. It just means that you're trained on those particular examples of pattern matching.

Same goes with all those tests asking you to work through a simple math problem like finding a hypotenuse of a right-angled triangle. Those are barely pattern matching anymore. For many they are routines.

Of course you could design a test that is fair in all regards but even so. A basic IQ test just doesn't adress true intelligence. It's a weak proxy for it, nothing more.

[eiffie]

The test kali linked to has only limited explanation before you are shown a series of tiles and asked to select one that fits the pattern - that is as culturally unbiased as it gets. A child can figure out the first questions with NO instruction. This is the kind of test that can be used to compare artificial intelligence to nature's.
And guess what - If you grow up in an environment with a lot of pattern-matching style puzzles - you will be more intelligent! Why do you think we have nursery rhymes and legos?

Are IQ tests weak - yes. So is our understanding of the universe - do we stop trying to quantify it too?

[taurus]

Are IQ tests weak - yes. So is our understanding of the universe - do we stop trying to quantify it too?

the first "intelligent" answer since quite a while in this thread.

[Tglad]

The criticism of IQ tests is not a criticism of science just that the IQ test is not an intelligence quota, it is a fast pattern recognition test. It is about as close to judging intelligence as a sudoku puzzle.

Should we stop trying to quantify intelligence? As a single value I would say yes, perceived intelligence can come from things such as diversity of skills in a population, communication between people, motivation, picking the right problems to solve, 'work'-life balance, ability to explain... etc. It doesn't map to a single value per person, even in principle. 

[Kali]

Just to mention that I did another test, and I got 128 (it took me 35 minutes). But it was culturally biased in this case, because it had a couple of wordplays in english (my first language is spanish) and as I don't have a vast vocabulary knowledge I probably didn't know the words I supposed to form from the letters (it seems I always have an excuse )

Despite the validity of this tests, at least it's fun for me to make them, and I think they are kind of a good exercise for the brain.

Also, maybe doing several tests makes the brain get better for this kind of mental challenges, and someday I'll fool MENSA to make me one of their members...
Just because I want to meet some very smart beautiful blondes... there must be some of them, although they are very rare specimens

[cKleinhuis]

usually i have no problems with the easy patterns, but at the end with all those kinds of test (after 66% )  i really get stuck, and one time i even heard the explanations for them, and i really could not get the logic behind it after hearing the explanation

[taurus]

The criticism of IQ tests is not a criticism of science...

But it shows a significant weakness of our treatment of scientific evidence. We tend to mistake them for truth!

[kram1032]

We don't "mistake" scientific evidence for truth.
What we do is, that we accept it for truth until it's no longer sufficient.
Basically, we take it as "approximate truth" which is "good enough for now."
And in some cases, like Newtonian mechanics, it's even "good enough" for a wide range of use-cases, namely in scenarios with rather slow speeds, neglectible gravity subtleties and of macroscopic extend.

And we know for a fact that our currently best approximate truth is no longer good enough once we go microscopic and want to consider gravity subtelties. Still, even that limited approximation has brought us a far way. The approximations are well established and very good in a wide range of use-cases.

They are not some random statistical weak proxy that measures something entirely different from what it's supposed ot measure.
(For that matter, they aren't even tools of measuring. Rather they are tools for predicting what's going to be measured)

And don't get me wrong, there are some really good statistical proxies out there, used a lot with great results, despite being mere proxies, giving rise to quite elaborate, yet accurate predictions.
IQ tests simply fail to be useful for predicting quite anything.
At best they predict how successful you have been at school before the tester sees your school certificate.
Not rarely, they don't even do a decent job in explaining that.

This is my biggest problem with IQ tests: Besides maybe being helpful to brag in some circles, they simply don't have a true scientific relevance. They can be abused as status symbol. Nothing more.

Btw, Kali, doing them for fun is fine. I liked doing them as well. - I generally like doing logics puzzles of various kinds.
Just don't expect the value to mean a thing beyond being essentially a grade for how well you did on those specific problems.

[taurus]

We don't "mistake" scientific evidence for truth.
What we do is, that we accept it for truth until it's no longer sufficient.
Basically, we take it as "approximate truth" which is "good enough for now."

That's a matter of attitude. I shared your opinion more than half of my life, so don't think, I don't understand your arguments. In opposite - I'm only sick and tired of them.
The best science ever achieved are partial truthes - not more not less. And science allways fails, when it comes to the bigger correlation. (like IQ Tests demonstrate)
I know I am only a screaming minority - especially amongst mathematicians - and I don't want to kick off that old, boring discussion.
The basics of my point of view are refered in the books of Fritjof Capra, Ilya Prigogine or Hans-Peter Duerr.

[kram1032]

I'm not quite sure what you're refering to when saying, "science fails [for] bigger correlations".
I'm especially unsure what "bigger correlations" would even be.
The only thing that comes to mind is works on huge data-sets but that doesn't really describe "big correlations". Working on such data-sets has only recently become reasonably feasible with the rise of cheap fast computers. And even there, a lot of progress is being made all the time.

I don't see how an IQ test would fit in there though.

Could you please elaborate?

[taurus]

just one point: intelligence is what i would call a multifactorial quality. We a) can't describe it precisely and b) can't even guess how many factors have influence on it.
So in first order approximation it fits my definition of a bigger correlation (correlation of higher order is a better term - sorry, I'm not natively english)
The attempt to describe such a complex combination of attributes with a reductionistic approach like the IQ - teared down to one value - is nothing more than ridiculous. I hope even science believers can understand, why I see no truth at all in this.
For the rest I honestly recommend the lecture mentioned in my last post.

[kram1032]

Uh, I don't see why we are having this argument at all then.
I agree: Tearing it down to a single value is ridiculous.
There have been some more or less successful attempts in which qualitative words got a quantitative meassure attatched to them. IQ tests clearly are not one of them.