Ugh, visual shape processing. You grow up with that sorts of shapes (and patterns, and consecutive patterns that are regular, and so on), Africans don’t. You grow up with everything in left to right order or right to left order, they don’t.
What do you think goes on formally (mathematically) with the correct answer, anyway?
The correct answer is the one where the whole thing with the square filled in can be least complexly represented with most culturally common operations (mirrorings, rotations, superpositions, etc) done on orderings of the squares. You have a penalty for each operation (more for less common operations), you add those scores for the whole set of relations, you pick the smallest. That’s roughly what a programming contest solution for that sort of thing would look like (leaving aside the question of hardcoding or inferring the patterns and their penalties themselves).
Yes, the operations are in some sense fundamental, but you haven’t reinvented them, you learned them, from when you were categorizing visual input as a child.
As an IQ test, it has two parts: the visual input you are exposed to as a child, and the matrices themselves. Since we’re all acquiring a sufficient training dataset, it works just fine as an IQ test for us.
edit: Also, try replacing the square to fill in with a circle, and see how many people will get that wrong. Empty box to fill in is a cultural concept. A child unused to this will think they need to use that box as part of the answer.
Ugh, visual shape processing. You grow up with that sorts of shapes (and patterns, and consecutive patterns that are regular, and so on), Africans don’t. You grow up with everything in left to right order or right to left order, they don’t.
What do you think goes on formally (mathematically) with the correct answer, anyway?
The correct answer is the one where the whole thing with the square filled in can be least complexly represented with most culturally common operations (mirrorings, rotations, superpositions, etc) done on orderings of the squares. You have a penalty for each operation (more for less common operations), you add those scores for the whole set of relations, you pick the smallest. That’s roughly what a programming contest solution for that sort of thing would look like (leaving aside the question of hardcoding or inferring the patterns and their penalties themselves).
Yes, the operations are in some sense fundamental, but you haven’t reinvented them, you learned them, from when you were categorizing visual input as a child.
As an IQ test, it has two parts: the visual input you are exposed to as a child, and the matrices themselves. Since we’re all acquiring a sufficient training dataset, it works just fine as an IQ test for us.
edit: Also, try replacing the square to fill in with a circle, and see how many people will get that wrong. Empty box to fill in is a cultural concept. A child unused to this will think they need to use that box as part of the answer.