Richard Haier is a psychologist specializing in the science of human intelligence. Please support this podcast by checking out our sponsors:
- Calm: https://calm.com/lex to get 40% off
- Linode: https://linode.com/lex to get $100 free credit
- BiOptimizers: http://www.magbreakthrough.com/lex to get 10% off
- SimpliSafe: https://simplisafe.com/lex and use code LEX
- MasterClass: https://masterclass.com/lex to get 15% off

EPISODE LINKS:
Richard's Twitter: https://twitter.com/rjhaier
Richard's Website: https://richardhaier.com/ 
Documents & Articles:
1. Child IQ and survival to 79: https://ncbi.nlm.nih.gov/pmc/articles/PMC5491698/
2. Study of Mathematically Precocious Youth: https://my.vanderbilt.edu/smpy/files/2013/02/DoingPsychScience2006.pdf
Books:
1. The Neuroscience of Intelligence: https://amzn.to/3n50DcC
2. The Book of Five Rings: https://amzn.to/3y4Xcc6
3. The Rise and Fall of the Third Reich: https://amzn.to/3zPAW7q
4. Flowers for Algernon: https://amzn.to/3OfRKZS
5. The Bell Curve: https://amzn.to/3Ng4RJe
6. The Mismeasure of Man: https://amzn.to/3N9IkxB
7. Human Diversity: https://amzn.to/3O7Trsc
8. Facing Reality: https://amzn.to/3bfzqkX

PODCAST INFO:
Podcast website: https://lexfridman.com/podcast
Apple Podcasts: https://apple.co/2lwqZIr
Spotify: https://spoti.fi/2nEwCF8
RSS: https://lexfridman.com/feed/podcast/
Full episodes playlist: https://www.youtube.com/playlist?list=PLrAXtmErZgOdP_8GztsuKi9nrraNbKKp4
Clips playlist: https://www.youtube.com/playlist?list=PLrAXtmErZgOeciFP3CBCIEElOJeitOr41

OUTLINE:
0:00 - Introduction
0:43 - Measuring human intelligence
15:11 - IQ tests
37:59 - College entrance exams
46:36 - Genetics
52:35 - Enhancing intelligence
1:00:04 - The Bell Curve
1:12:35 - Race differences
1:31:48 - Bell curve criticisms
1:40:57 - Intelligence and life success
1:50:34 - Flynn effect
1:55:26 - Nature vs nuture
2:22:19 - Testing artificial intelligence
2:34:23 - Advice
2:38:30 - Mortality

SOCIAL:
- Twitter: https://twitter.com/lexfridman
- LinkedIn: https://www.linkedin.com/in/lexfridman
- Facebook: https://www.facebook.com/lexfridman
- Instagram: https://www.instagram.com/lexfridman
- Medium: https://medium.com/@lexfridman
- Reddit: https://reddit.com/r/lexfridman
- Support on Patreon: https://www.patreon.com/lexfridman

Let me ask you to this question, whether it's bell curve or any research on race differences, can that be used to increase the amount of racism in the world, can that be used to increase the amount of hate in the world?

My sense is there is such enormous reservoirs of hate and racism that have nothing to do with scientific knowledge of the data that speak against that.

That no, I don't want to give racist groups a veto power over what scientists study.

The following is a conversation with Richard Heyer on the science of human intelligence.

This is a highly controversial topic, but a critically important 1 for understanding the human mind.

I hope you will join me in not shying away from difficult topics like this, and instead, let us try to navigate it with empathy, rigor, and grace.

If you're watching this on video now, I should mention that I'm recording this introduction in an undisclosed location somewhere in the world.

I'm safe and happy, and life is beautiful.

To support it, please check out our sponsors in the description.

And now, dear friends, here's Richard Heyer.

What are the measures of human intelligence and How do we measure it?

Everybody has an idea of what they mean by intelligence.

In the vernacular, what I mean by intelligence is just being smart, how well you reason, how well you figure things out, What you do when you don't know what to do.

Those are just kind of everyday common sense definitions of how people use the word intelligence.

If you want to do research on intelligence, measuring something that you can study scientifically is a little trickier.

And what almost all researchers who study intelligence use is the concept called the G factor, general intelligence.

And that is what is common, that is a mental ability that is common to virtually all tests of mental abilities.

What's the origin of the term G-factor, by the way?

Such a funny word for such a fundamental human thing.

The general factor, I really started with Charles Spearman.

And he noticed, this is like, boy, more than 100 years ago.

He noticed that when you tested people with different tests, all the tests were correlated positively.

And so he was looking at student exams and things.

And he invented the correlation coefficient, essentially.

And he, when he used it to look at student performance on various topics, he found they all the scores were correlated with each other and they were all positive correlations.

So he inferred from this that there must be some common factor that was irrespective of the content of the test.

And positive correlation means if you do well on the first test, you're likely to do well on the second test, and presumably that holds for tests across even disciplines, so not within subject but across subjects, so that's where the general comes in.

Something about general intelligence, So when you were talking about measuring intelligence and trying to figure out something difficult about this world and how to solve the puzzles of this world, that means generally speaking, not some specific test, but across all tests.

Absolutely right, and people get hung up on this because they said, well, what about the ability to do X?

And they said, I know somebody who's very good at this, but not so good at this, this other thing.

And so there are a lot of examples like that, but it's a general tendency.

So exceptions really don't disprove, you know, your everyday experience is not the same as what the data actually show.

And your everyday experience, when you say, oh, I know someone who's good at X, but not so good at Y, that doesn't contradict the statement of about, he's not so good, but he's not the opposite.

Okay, so we're not, our anecdotal data, I know a guy who's really good at solving some kind of visual thing.

That's not sufficient for us to understand actually the depths of that person's intelligence.

So how this idea of g-factor, how much evidence is there, how strong, you know, given across the decades that this idea has been around, how much has it been held up that there's a universal sort of horsepower of intelligence that's underneath all of it, all the different tests we do to try to get to this thing in the depths of the human mind that's a universal stable measure of a person's intelligence.

You used a couple of words in there, stable and.

Are we gonna have to be precise with words?

I was hoping we can get away with being poetic.

We can, there's a lot about research in general, not just intelligence research, that is poetic.

Science has a phonetic aspect to it, and good scientists are very intuitive.

They're not just, hey, hey, these are the numbers.

You have to kind of step back and see the big picture.

When it comes to intelligence research, you asked how well has this general concept held up?

And I think I can say without fear of being empirically contradicted, that it is the most replicated finding in all of psychology.

Now some cynics may say, well big deal, psychology, we all know there's a replication crisis in psychology and a lot of this stuff doesn't replicate.

There is no replication crisis when it comes to studying the existence of this general factor.

It looks like it's universal, that you find it in all cultures.

The way you find it, step back 1 step, the way you find it is to give a battery of mental tests.

Take a battery of any mental test you want, give it to a large number of diverse people, and you will be able to extract statistically the commonality among all those tests.

It's done by a technique called factor analysis.

People think that this may be a statistical artifact of some kind, it is not a statistical artifact.

Factor analysis is a way of looking at a big set of data and look at the correlation among the different test scores and then find empirically the clusters of scores that go together.

So if you have a bunch of mental tests, there may be a verbal factor.

But those factors have variants in common with each other.

And that is the common, that's what's common among all the tests, and that's what gets labeled the G factor.

So if you give a diverse battery of mental tests and you extract a g-factor from it, that factor usually accounts for around half of the variance.

It's the single biggest factor, but it's not the only factor.

But it is the most reliable, It is the most stable, and it seems to be very much influenced by genetics.

It's very hard to change the G factor with training or drugs or anything else.

We don't know how to increase the G factor.

Okay, you said a lot of really interesting things there.

So first, just to get people used to it in case they're not familiar with this idea, G factor is what we mean.

So often there's this term used, IQ, which is the way IQ is used, they really mean G factor in regular conversation.

Because what we mean by IQ, we mean intelligence, and what we mean by intelligence, we mean general intelligence, and general intelligence in the human mind from a psychology, from a serious rigorous scientific perspective actually means G factor.

So G factor equals intelligence, just in this conversation to define terms.

Okay, so there's this stable thing called g-factor.

now factor, you said factor many times means a measure that potentially could be reduced to a single number across the different factors you mentioned.

And what you said, it accounts for half, half-ish.

Of variance across the different set of tests.

So if you do for some reason well on some set of tests, What does that mean?

So that means there's some unique capabilities outside of the G factor that might account for that.

What else is there besides the raw horsepower, the engine inside your mind that generates intelligence?

Someone might be particularly good at mathematical things, mathematical concepts, even simple arithmetic people, some people are much better than others.

You might know people who can memorize, and short-term memory is another component of this.

Short-term memory is 1 of the cognitive processes that's most highly correlated with the G factor.

So all those things like memory, test-taking skills account for variability across the test performances.

But you, so you can run, but you can't hide from the thing that God gave you, the genetics.

So that G factor, science says that G factor's there.

Well IQ is a score and IQ score is a very good estimate of the g-factor.

You can't measure g directly, there's no direct measure.

You estimate it from these statistical techniques.

Because a standard IQ test is a battery of different mental abilities.

You combine it into 1 score, and that score is highly correlated with the G factor, even if you get better scores on some subtests than others.

Because again, it's what's common to all these mental abilities.

So a good IQ test, and I'll ask you about that, but a good IQ test tries to compress down that battery of tests, like tries to get a nice battery, the nice selection of variable tests into 1 test.

And so in that way, it sneaks up to this G factor.

And that's another interesting thing about G factor.

Now you give, first of all, You have a great book on the neuroscience of intelligence.

You have a great course, which is when I first learned, you're a great teacher, let me just say.

Your course at the teaching company, I hope I'm saying that correctly.

Is when I first heard about this G factor, this mysterious thing that lurks in the darkness that we cannot quite shine a light on, we're trying to sneak up on.

So the fact that there's this measure, stable measure of intelligence, we can't measure directly.

But we can come up with a battery test or 1 test that includes a battery of variable type of questions that can reliably or attempt to estimate in a stable way that G factor.

So for me as an AI person, it's fascinating.

It's fascinating there's something stable like that about the human mind, especially if it's grounded in genetics.

It's both fascinating that as a researcher of the human mind and all the human psychological, sociological, ethical questions that start arising, it makes me uncomfortable.

I get that a lot about being uncomfortable talking about this.

Let me go back and just say 1 more empirical thing.

It doesn't matter which battery of tests you use.

You can take any 12 of them at random, extract a G factor and another 12 at random and extract a G factor.

And those G factors will be highly correlated, like over 0.9 with each other.

It doesn't depend on the content of the test, is what I'm trying to say.

It is general among all those tests of mental ability and tests of mental, you know, mental abilities include things like, geez, playing poker.

Your skill at poker is not unrelated to G.

Your skill at anything that requires reasoning and thinking, anything, spelling, arithmetic, more complex things, this concept is ubiquitous.

And when you do batteries of tests in different cultures, you get the same thing.

So this says something interesting about the human mind that is a computer is designed to be general.

So that means you can, So it's not easily made specialized, meaning if you're going to be good at 1 thing, Miyamoto Musashi has this quote, he's an ancient warrior famous for the Book of 5 Rings in the martial arts world.

And the quote goes, "'If you know the way broadly, "'you will see it in everything.'" Meaning if you do 1 thing, it's going to generalize to everything.

And that's an interesting thing about the human mind.

Okay, so what's the difference, if you can elaborate a little bit further, between IQ and G factor, just because it's a source of confusion for people.

People use the word IQ to mean intelligence, but IQ has a more technical meaning for people who work in the field.

And it's an IQ score, a score on a test that estimates the G factor.

And the G factor is what's common among all these tests of mental ability.

So if you think about, it's not a Venn diagram, but I guess you could make a Venn diagram out of it, but the G factor would be really at the core, what's common to everything.

And what IQ scores do is they allow a rank order of people on the score.

And this is what makes people uncomfortable.

This is where there's a lot of controversy about whether IQ tests are biased toward any 1 group or another.

And a lot of the answers to these questions are very clear, but they also have a technical aspect of it that's not so easy to explain.

Well, we'll talk about the fascinating and the difficult things about all of this, but...

So by the way, when you say rank order, that means you get a number and that means 1 person, you can now compare.

Like you could say that this other person is more intelligent than me.

you can say is IQ scores are interpreted really as percentiles.

So that if you have an IQ of 140 and somebody else has 70, the metric is such that you cannot say the person with an IQ of 140 is twice as smart as a person with an IQ of 70.

That would require a ratio scale with an absolute 0.

Now you may think you know people with 0 intelligence, but in fact, there is no absolute 0 on an IQ scale.

So relative to other people, somebody with an IQ score of 140 is in the upper less than

whereas somebody with an IQ of 70 is 2 standard deviations below the mean.

So it's similar to like in chess, you have an Elo rating that's designed to rank order people.

If your ELO rating's twice another person, I don't think you're twice as good at chess.

It's not stable in that way, but because it's very difficult to do these kinds of comparisons.

But, so what can we say about the number itself?

Is that stable across tests and so on or no?

a number of statistical properties of any test, they're called psychometric properties.

You have validity, you have reliability, reliability, there are many different kinds of reliability.

And IQ tests are stable within an individual.

There are some longitudinal studies where children were measured at age 11.

And again, when they were 70 years old and the 2 IQ scores are highly correlated with each other.

This comes from a fascinating study from Scotland.

In the 1930s, some researchers decided to get an IQ test on every single child age 11 in the whole country.

And those records were discovered in an old storeroom at the University of Edinburgh by a friend of mine, Ian Deary, who found the records, digitized them, and has done a lot of research on the people who are still alive today from that original study, including brain imaging research, by the way.

Really it's a fascinating group of people who are studied.

Not to get ahead of the story, but 1 of the most interesting things they found is a very strong relationship between IQ measured at age 11 and mortality so that, you know, in the 70 years later, they looked at the survival rates and they could get death records from everybody and Scotland has universal health care for everybody.

And it turned out if you divide people by their age 11 IQ score into quartiles and then look at how many people are alive 70 years later.

I know this is in the book, I have the graph in the book, but there are essentially twice as many people alive in the highest IQ quartile than in the lowest IQ quartile.

Now why this is the case is not so clear since everyone had access to healthcare.

Well there's a lot, and we'll talk about it, you know, just the sentences you used now could be explained by nature or nurture.

Now, there's a lot of science that starts to then dig in and investigate that question.

How are the test design, IQ test design, how do they work?

Maybe some examples for people who are not aware.

What makes a good IQ test question that sneaks up on this G factor measure?

Well, your question is interesting because you want me to give examples of items that make good items and what makes a good item is not so much its content but its empirical relationship to the total score that turns out to be valid by other means.

So for example, let me give you an odd example from personality testing.

So there's a personality test called the Minnesota Multiphasic Personality Inventory, MMPI.

I've heard about this test recently because of the Johnny Depp and Amber Heard trial.

I don't know if you've been paying attention to that.

They had psychologists on the stand and they were talking, apparently those psychologists did, again, I'm learning so much from this trial.

They did different, a battery of tests to diagnose personality disorders.

Apparently there's that systematic way of doing so and the Minnesota 1 is 1 of the ones that there's the most science on.

There's a lot of great papers which were all continuously cited on the stand, which is fascinating to watch.

It's okay, I mean, this is interesting because you're right, it's been around for decades.

There's a lot of scientific research on the psychometric properties of the test, including what it predicts with respect to different categories of personality disorder But what I want to mention is the content of the items on that test.

All of the items are Essentially true false items true or false.

True or false, I think Lincoln was a better president than Washington.

What have all these, what does that have to do, And the point is the content in these items, nobody knows why these items in aggregate predict anything, but empirically they do.

It's a technique of choosing items for a test that is called dust bowl empiricism.

That the content doesn't matter, but for some reason when you get a criterion group of people with this disorder, and you compare them to people without that disorder, these are the items that distinguish.

Irrespective of content, it's a hard concept to grasp.

But from, because I consider myself part psychologist because I love human-robot interaction, and that's a problem, half of that problem is a psychology problem because there's a human.

So designing these tests to get at the questions is the fascinating part.

Like how do you get to, like what does Dust Bowl empiricism refer to?

Yeah, so it's the test is Dust Bowl empiricism.

But how do you arrive at the battery of questions?

I presume 1 of the things, now again, I'm going to the excellent testimony in that trial, the explanation, because they also, they explain the tests, that a bunch of the questions are kind of make you forget that you're taking a test.

Like it makes it very difficult for you to somehow figure out what you're supposed to answer.

But we're getting a little far afield because I only wanted to give that example of Dust Bowl empiricism.

When we talk about the items on an IQ test, many of those items in the Dust Bowl in Pearson method have no face validity.

In other words, they don't look like they measure anything.

Whereas most intelligence tests, the items actually look like they're measuring some mental ability.

So you were bringing that up as an example as what it is not.

Okay, so I don't wanna go too far afield on it.

Too far afield is actually 1 of the names of this podcast, so I should mention that.

So they feel the questions look like they passed the face validity test.

If I, 1 of the subtests on a standard IQ test is general information.

Let me just think a little bit, because I don't want to give you the actual item.

But if I said, how far is it between Washington DC and Miami, Florida, within

Well, you know, it's not a fact most people memorize, but you, you know something about geography.

But it's also seems like it would be very cultural.

So there's that kind of general information.

And I choose that word because that word was removed from the IQ test because people complained that disadvantaged people would not know that word just from their everyday life.

Okay, here's another example from a different kind of sub-test.

A sailing competition, a competition with boats.

Not necessarily sailing, but a competition with boats.

I'm gonna repeat a string of numbers, and when I'm done, I want you to repeat them back to me.

Now the actual test starts with a smaller number, you know, like 2 numbers, and then as people get it right, you keep going, Adding to the string of numbers until they can't do it anymore Okay, but now try this.

I'm gonna say some numbers and when I'm done I want you to repeat them to me backwards.

Okay Now so I gave you some examples of the kind of items on an IQ test.

General information, I can't even remember all, general information, vocabulary, digit span forward and digit span backward.

Well, you said I can't even remember them.

See all Lex Fridman transcripts on Youtube

Richard Haier: IQ Tests, Human Intelligence, and Group Differences | Lex Fridman Podcast #302