Ask A Genius 1039: High-Range Tests, potentials and promises

Author(s): Rick Rosner and Scott Douglas Jacobsen

Publication (Outlet/Website): Ask A Genius

Publication Date (yyyy/mm/dd): 2024/07/28

Rick Rosner, American Comedy Writer, www.rickrosner.org

Scott Douglas Jacobsen, Independent Journalist, www.in-sightpublishing.com

Scott Douglas Jacobsen: What is the difficulty of constructing a high-range test? For the sake of conversation, people have different definitions in these communities. I will keep it simple: anything above and not including a 160 IQ on a standard deviation of 15 because that is where the mainstream tests end. 

Rick Rosner: Landscape-wise, I’ve seen claims for IQ. If you go online and poke around, somebody claims that Jesus had an IQ of 300. The greatest man who ever lived must be the smartest man who ever lived, so he must have an IQ of 300. A test with a standard deviation of 16 would give him about 12 and a half standard deviations above the mean score, making him the smartest person in history, as 1 in a sextillion. If you take all the people who’ve ever lived, approximately 120 billion people, and multiply that by another quadrillion, Jesus would be smarter than a quadrillion times all the people who have ever lived.

Many things could be improved with that claim. Adult IQ is calculated based on rarity among humans if IQ is distributed in a normal Gaussian bell curve—the bell curve you’ve seen if you’ve ever taken a statistics class. You’re smarter than half of everybody if you’re 0 standard deviations above the mean.

If you’re one standard deviation, you’re smarter than 84% of everybody. At two standard deviations, you’re one out of 44 in smartness; at three standard deviations, you’re one out of 750; at four standard deviations, you’re one out of 30,000; at five standard deviations, you’re one out of 3,000,000; and at six standard deviations, you’re one out of 750,000,000. Am I doing it right?

So, yes, seven standard deviations, one out of 500,000,000,000, four times the number of humans who’ve ever lived. So, anybody claiming a seven-standard deviation score, scoring a 205 on a test with a standard deviation of 15, would be the smartest person on four planets of Earth combined. It’s not likely that anybody is that smart.

You could argue that intelligence isn’t real or doesn’t fall in a bell curve. Or you could argue that the whole thing is ridiculous, which is the POV I’m closest to. But the deal is that, as we’ve talked about before but not for a while, most IQ tests, including the oldest and most venerable ones like the Wechsler and Stanford-Binet, have close to a 100-year history and have been re-normed on thousands and thousands of people. We know, up to a 150 or 160 IQ, what the degree and type of intelligence measured by IQ tests look like. That’s all you need.

Tests measure 50 points on either side of 100, up to 150 and down to 50. That’s all you need. If somebody maxes the test at 150, they will be bored in school. It would help if you considered paying them a grade, giving them extra study materials, or letting them take college courses in high school.

Advertisements

blob:https://insightjournalonline.wordpress.com/5bb1096c-afd9-47e7-9049-fcf614f534eb

REPORT THIS AD

If they score 50, they’ve probably demonstrated that they might need to be institutionalized depending on their other spectrum of behaviours. I used to do some volunteer work with people who lived in a group home for people with mental handicaps. Some of those people ended up there not for mental handicaps, but who knows what was going on in the late seventies? Some of those people were not as impaired as they should have been to be in a home for people with mental handicaps. On the other hand, one guy, Keith, had an estimated IQ of 25. He needed help with many things but managed within a structured environment where everybody knew his skill levels. He needed help going to the bathroom but could still go roller skating. He could feed himself and had a normal set of emotions, mostly happy.

Anyway, you don’t need to tell the difference between someone with an IQ of 40 and 25. It would help if you worked with them one-on-one to understand their capabilities rather than relying on some hypothetical IQ score. Ditto for telling the difference between someone with an IQ of 150 and 165. It would help if you worked with those people one-on-one to see what they’re into and what they can do, which shouldn’t overly tax an educational system. It does because school systems are broken, but only about one kid in a thousand will score above 150. Only one kid in 30,000 will score above 165.

The average kid in a school system scores around 105, 106, or 110 because kids like Keith, with an IQ of 25, aren’t in the school system. They’re somewhere else. So you could get ten kids, one kid in 150 with a three standard deviation IQ, and one kid in 5,000 with a four standard deviation IQ. You should still be able to give individualized attention to those kids, though not in places like Oklahoma, where they have starved the school system of so much money that they can’t even have classes five days a week.

They had to go to 4 days a week. Regarding schooling, you don’t need to know where people are above 150. The test manufacturers, like those behind the Wechsler Adult Intelligence Scale, claim it can measure up to 175 or 180 using special methods. Some made this claim for the 3rd and 4th versions and possibly the 5th version. However, there’s much skepticism because these tests generally measure between 40 and 160 with a standard deviation of 15. So, people claim it, and people do it. Is it legitimate? Yes.

The Stanford-Binet 4, normed around 1960, is notorious for giving high scores. If you want to get your kid into a highly gifted program, you might take them to a private psychologist who uses this test. Our kid took it and scored highly because we practiced and chose the right test. This is the same test that Marilyn Savant, the world’s smartest person, took as a child, scoring an IQ of around 212-232. However, these claims are often met with skepticism.

Beyond these mainstream tests, hobbyists and serious researchers study exceptionally high human intelligence. Ron Hoeflin, for example, is a serious philosopher interested in this field. Paul Cooijmans is another practitioner who writes super-high IQ tests, often challenging those with high self-importance.

Building a test that can measure up to 5 standard deviations (180 IQ on a standard deviation of 16 or 175 on a standard deviation of 15) is very difficult. The difference between SD 15 and SD 16 tests is minor, but SD 16 can give a few extra points for the highest IQ claims. Hoeflin’s Mega and Titan Test have been argued to measure up to 5 standard deviations. Over 4,000 people took the Mega Test, and their scores were compared to other tests, providing a convincing argument for its legitimacy.

Jacobsen: David Redvaldsen, a professional statistician, published a review in a peer-reviewed journal, supporting Hoeflin’s claims to some extent. While it may not meet all psychometric standards, it advances the field and gives some legitimacy to these high-range tests.

Rosner: Sadly, the Mega and the Titan are obsolete because they’ve been around for so long. The Mega has been out since 1985. And the Titan since 1990. They’ve been spoiled because they only have 48 problems each. Half the problems are verbal, and they were written before Google. Almost all verbal problems are analogies, and 80% of the analogies can easily be solved via search engines, especially Google. Who’s going to use Bing?

You can’t give a test in which you don’t have to do any thinking to solve 20 of the problems right off the bat. The other problems on the test have been circulated on the Internet for decades, so you can probably search out the answers to at least another 15 of those problems just by clicking around. Without thinking, you could get at least a 35 out of 48 on the Mega test. So it can’t be used anymore, which is a small tragedy. I still respect people who are trying to build these tests.

As we’ve talked about before, it’s hard. Hoeflin worked hard to find inherently difficult problems while being fairly simple to state. One pitfall of many ultra-high tests is that the hardest items pile esoteric detective work on top of esoteric detective work for three layers deep. You do a bunch of research, figure out one part or step in the analogy, then do some more detective work to see what some other manipulation is, and you may still need to be done. It’s not that it’s an inherently tough problem. For example, one of Hoeflin’s problems was: what if you took a bagel and sliced it so that the slice through it was a Mobius strip? How many pieces would you end up with if you did that three times?

Advertisements

blob:https://insightjournalonline.wordpress.com/5bb1096c-afd9-47e7-9049-fcf614f534eb

REPORT THIS AD

That’s easily stated, but building the mental picture that’ll lead you to the answer is hard. That item was written in the seventies or eighties before we had CADCAM and computer graphics programs. You can solve it without thinking much by drawing it on a computer. But it’s inherently hard because it asks you to develop a mental picture of what a Mobius strip slice through a bagel would look like and how multiple slices would interact. Problems like these are rare.

Another issue is that Hoeflin is a very smart guy, but can somebody who doesn’t have an IQ of 190 create IQ problems that legitimately measure up to 190? Yes, if that person were sufficiently conscientious. It’s similar to how late-night and award shows come up with good jokes.

Shitty late-night shows may not achieve this. Quality late-night shows have teams of a dozen writers who write hundreds of jokes a day. Out of those hundreds of jokes, the host writes some themselves and then picks a dozen or two dozen jokes out of a few hundred. The best ones fit best into what the host wants to do that night. If the host gets 200 jokes and still doesn’t find jokes they think are good enough, especially for an award show where you’re trying to make a big impact, the host can say, “All right, we need another batch.” The writers go back and pump out another 60, which can happen multiple times. You have to do the same thing for an ultra-high IQ test. You can’t just come up with a problem and think, “This is pretty hard, and I can figure out another couple of add-ons to make it even harder.” No, you have to come up with dozens and hundreds of potential test items, and you want to pick the ones that seem inherently hard and elegant.

Elegance is a nice thing to have because often, with an elegant item, you’ve figured out the whole thing once you’re done, and the correct answer snaps into place. Hoeflin had many problems like that. Cooijmans also has problems that snap into place satisfyingly when you’ve finally arrived. Other test writers don’t have that level of excellence as often.

A third problem to consider is whether intelligence can go to 190 or 200 and how would it work. There’s this concept of ‘g,’ which is generalized intelligence, one of the foundations of IQ, suggesting a general problem-solving skill level that can be applied to any problem. This concept relies on many assumptions. The world may have problems of limited complexity, or super hard problems may take forms that humans or AI with big data can’t solve. AI is built from big datasets, making it apparent that some problems might require exceptionally rare human intelligence, such as a 200 IQ, or even an engineered intelligence, either on its own or working with humans, to handle the vast amounts of data needed to find the answer within a human lifetime. 

Comments? 

Jacobsen: No comments.

License & Copyright

In-Sight Publishing by Scott Douglas Jacobsen is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License. ©Scott Douglas Jacobsen and In-Sight Publishing 2012-Present. Unauthorized use or duplication of material without express permission from Scott Douglas Jacobsen strictly prohibited, excerpts and links must use full credit to Scott Douglas Jacobsen and In-Sight Publishing with direction to the original content.