Comments by "Christian Baune" (@programaths) on "What makes an IQ test question hard? | François Chollet and Lex Fridman" video.
-
10
-
That's because Mensa issues grants. So, you barely hear about them as it's of no interest to MSM. Neither it is new or contricersial.
Member benefits like SIGHT service, SIG groups and various company giving reductions is not that interesting to speak about.
So, quite frankly, what Mensa does is very good but boring.
The loudest Mensa members never saw a membership and I have doubts they even saw an IQ test.
You may have seen one or two real member boasting about it. Just recall that 2% of the population is huge and a fraction of it still stay quite big. So the potential to have 100 assholes who are very loud is there!
Mensa membership was never about solving important world problems thoug.
Identify, foster and promote intelligence. That's what Mensa stands for. The membership goes toward that.
I think Mensa is wrongly represented the rare times it gets exposed on MSM. Too much focus on the people and they forget to remind the goals.
But it's more entertaining that way for the audience.
So, the analogy does not hold. Paying Mensa is like paying WWF because you believe in their cause. They will do their thing and help you work toward their goals.
Very same for Mensa, you pay them, they offer you services and your funds get into grants. And as in WWF a part of the membership goes for running the company.
Now, I acknowledge than "save the animals" is more relatable than "identify, foster and promote intelligence".
3
-
2
-
2
-
2
-
2
-
2
-
2
-
1
-
1
-
1
-
1
-
1
-
1
-
@chunchunmaru3644 No, because those YES/NO questions are challenging something very precise and so it's more controlled.
If there was multiple possible answers depending on how you think, these could "average out" and give more distributed results.
Now, when I create items (for fun), I do have propositions that permit me to know what wen wrong. So, on some of them, I can see if you are using shortcuts wrongly, trying to reuse prior logic or if you are failing to see a specific kind of parameter.
As an example, this question worked well because you got primed by the two first question. So you handled digits as distinct entity in each bracket because that was the case before. It's not a conscious thing, but it's there.
On a test with multiple items, this contamination does happens and is an actual part.
One of my test (that I keep offline) even goes as far as using look alike items with a small chance to prime and prompt for a wrong answer (yes, that is a bit sadistic, but it weed of people who are relying too much on their memory and unable to exert impulse control).
Here is an example of such test: http://www.iquizz.be/question.php
50% of people get the very first question wrong because they miss one parameter and rely to much on their perception. It's one of the easiest item, but it requires one to control his impulse and look at each proposition. Which is (as I can see) seldom done.
1
-
1
-
1
-
@untonsured Neither ^^ I trial the tests online on individual I don't even know. For the offline test, I am also accepting people, but they can only donit once. I would say that I lean on IRT because items are progressive and skewed in a way to be really harder and harder. So, everyone should be able to solve the first item (in fact, 9 of the 10) and very few people should be able to solve the last one. So, the offline test produces a skewed belle curve (like a poisson distribution if not mistaken). That allows to pick 10 items and have a rough estimate.
Also, the offline test is just not raw score to IQ, but also how well you manage depending on level of abstraction, number of parameters, wether it's "local" or not, if some data is incomplete or not, if you stick to previous reasoning, are influenced by the item presentation, or can work with different kind of exercises. Still trying to norm it, but it's quite hard. I don't have the money to hire an agency, so I have to meet people one to one. Also, finding people is hard, because they have to be interested and sit for 2h in the worse case. (was 4, but after 2h, if you are halfway, you're drained. So there is no point to finish the test.)
Amd when I was proctoring as a volunteer, I didn't do the analysis; I was psychotechnician. Which means I can handle the material, but can't evalute.
For my own tests, though, I can. That's also why I avoid giving an IQ figure and stick to a ball park appreciation.
In the end, I would like that my offline test is properly normed and I would give it for free. But there is the danger that people memorize it. Which means I would have to limit its distribution somehow! Pretty sure there would be psychologists interrested in a free tool.
1
-
@AltumNovo Hardest part in making an item is not finding a good "logic". It's to not have confound.
When it's well designed, the answer is unique because one can eliminates all other possibilities as being more complex.
It touch to another subject that is information theory.
As an example, you will easily concede that: "1,1,1,1,1,1,1,1" is a simpler sequence than "1,2,3,4,5,6,7". Though, they are very close in simplicity. You may even not see the gap and says they are both equally simple.
Compared to "2,4,8,16,32,64", they are really simpler.
Now, what about: "1,2,3,4,1,2,4,3,2,1,4,3,2,4,1,3,2,4,3,1". Yes, much more complex than all the previous series.
To evaluate the complexity you use an heuristic that is: "How hard it is to describe it ?" and it means: "What is the minimal amount of information I need to describe it ?".
The first one is "repeat 1". The second is "add 1", the next is "consecutive power of two" (which requires an abstraction: "power of").
The last one is very verbose:
"each group of 4 numbers represent a step. The rightmost number goes left and every other times, the leftmost number goes right. When a number has reached the other end, the left/right-most begins its journey". Even with that, it will still be though for some people to get it.
And so, if faced with a numeric sequence question you have doubt about two possible solutions, measure the quantity of information required and pick the easiest. It's very possible that you are not seeing an easier pattern too.
1
-
1
-
1
-
1
-
1
-
1
-
@AltumNovo What a poor understanding of mathematics!
Mathematics is not arithmetic, geometry, algebra, calculus, topology, percolation ...
Problem solving requires one to analyze the situation properly, identify the steps and the tools he need, get/create tools, apply his solution.
That's what math teaches in a controlled environment.
Now, you had a poor education or you did follow it poorly. A good teacher don't show how to solve a problem, he put you on track when you stray away too much by giving your the proper hint. Happens that the byproduct of practicing problem solving using mathematics as an excuse is to learn some mathematical concepts in the process. But that's still a byproduct.
Now, I'll end the discussion here as you are basically asking me to provide what you can't and just babel assertions without foundations. Even asking studies you refute because they are studies made by people that have to appear to have an high IQ, so it's a circular argument.
I was answering to you for other potential readers who may fall for your boldness.
https://www.scientificamerican.com/article/research-confirms-a-link-between-intelligence-and-life-expectancy/
https://www.businessinsider.com/facts-you-dont-want-to-know-iq-2011-11?r=US&IR=T#breastfeeding-can-increase-a-childs-iq-by-three-to-eight-points-4
1
-
1