![]() When people smarter than the test designers take an IQ test, they often have to guess what the designers were thinking, but with video games, evaluation can be completely objective. If practice effects are inevitable, it's better for everyone to get practice to a point of diminishing returns instead of trying to prevent people from practicing (IQ tests) or charging money for it (SATs). It might not be as objective, but people could compete on aesthetics too. Roguelikes, many strategy games, and many simulation games are designed to be played many times. Yes, many video games have design goals opposite those of IQ tests, being designed to be played only once and to give the player a strong feeling of progression, but there are many video games, and some are more appropriate. Here's a theory of Alzheimer's I developed - what test score does that correspond to? As for IQ tests, I had a couple proper ones as a kid, and my scores were probably as high as was very meaningful, but probably less impressive than reading Feynman in 3rd grade. I left college after a semester (while that was a failure from the perspective of society, school was holding me back intellectually) but I still took the GRE got a top 1% score without studying, but that's not something I consider particularly meaningful either. My ACT score was in the top 0.1%, but I don't feel particularly proud of that, because it wasn't evaluating any of my actual strengths. Plus, no matter how well you did on standardized tests, there's usually somebody around who did better and wants to brag about it. I'm caught by a catch-22 here: if I don't talk about standardized test scores I got, people will think I'm bitter about doing poorly, but if I do, then I'm bragging about scores on standardized tests which is super lame. The term "video game" is very general (video games are a generalization of animation is a generalization of film is a generalization of photos are a generalization of text) so let me clarify: I'm talking about existing video games which were developed as entertainment. ![]() My view is that some combinations of video games are better "IQ tests" than actual IQ tests are, and better general standardized tests than the SAT. Leetcode, Math Olympiad, and Physics Olympiad problems have been proposed as better alternatives, but there's a lot of memorization of a specialized "toolbox" with those. The same is true of, say, questions on IIT entrance exams. For example, chess is suboptimal because it's both too small and narrow to effectively apply the full power of the sort of general-purpose systems that are most important to evaluate when testing for "general intelligence". ![]() If you're trying to evaluate intelligence in a broad sense, you should use tests with problems that are big enough to use most of a system and broad enough to use many systems. Then there's correlation between performance of various systems on various tasks for obvious reasons. Humans have multiple mental systems, those systems have performance on various tasks which vary depending on the amount of specialization, amount of training, quality of training data, and multiple types of management by other (meta-level) systems. Anyway, given modern understanding of AI and biology, I consider the entire "g factor" framework an archaic and crude way of understanding intelligence. IQ tests, in general, don't seem to be any better for that than the SAT. Is it "unusually high correlation with general intelligence", as represented by a "g factor"? I don't think so. IQ tests are notorious for working poorly above ~135, and I'd say they only really work well for -20 to +0 relative to the designers, with a somewhat wider range for teams. Is it "working for an unusually wide range of intelligence"? Not really. This just mostly doesn't come up, because institutions don't generally use IQ test results. ![]() A few practice runs can often be worth +8 points, and that's for kids that already to some extent do IQ-test-like stuff in school. ![]() Is it "practice being mostly irrelevant"? Not really. Is it "consistency of results across different IQ test types"? Not really that's obviously worse than the above, and many "non-IQ" tests have comparable consistency. That's a best-case scenario for tests designed with that criteria as a priority, and the range is still significant. Reported standard error may be an underestimate, as it does not account for all sources of error. Wikipedia says:įor modern tests, the confidence interval can be approximately 10 points and reported standard error of measurement can be as low as about three points. Is it "consistency of results on a single IQ test type"? Not really. What makes them different from other tests given to people? IQ tests are a type of test given to people. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |