Richard Haier on the G Factor, IQ Tests, Human Intelligence, and Its Controversies

This episode features a deep dive with Richard Haier into the science of human intelligence, focusing primarily on the concept of the 'g factor' or general intelligence. Haier defines intelligence in the vernacular as being 'smart'—how well one reasons, figures things out, and adapts to novel situations. Scientifically, however, intelligence research largely coalesces around the g factor, a mental ability common to virtually all tests of mental abilities, first identified by Charles Spearman over a century ago through the positive correlation of diverse test scores. The discussion emphasizes that the g factor is the most replicated finding in all of psychology, universal across cultures, and statistically extractable from any battery of mental tests using factor analysis, accounting for roughly half of the variance in test performance.

A key distinction is made between the g factor and an IQ score. While IQ is a score on a test that estimates the g factor, the g factor itself is the underlying commonality. IQ scores allow for a rank order of individuals, interpreted as percentiles, but do not represent a ratio scale (e.g., an IQ of 140 is not 'twice as smart' as 70). The stability of IQ scores within an individual is highlighted by longitudinal studies, such as the Scottish study that tracked individuals from age 11 to 70, revealing a strong correlation between early IQ and later mortality rates, even with universal healthcare access.

The conversation also explores the design of IQ tests, contrasting items with 'face validity' (which appear to measure mental ability) with 'dust bowl empiricism' (items whose content doesn't seem relevant but empirically predict outcomes, like in the MMPI personality test). Examples of IQ test items include general information, vocabulary, digit span (forward and backward), and reaction time tasks like the Hick paradigm, where complex decision-making reaction times correlate highly with g. The speed of information processing is identified as a significant component related to the g factor.

Broader implications touch upon the discomfort and ethical considerations surrounding intelligence research, particularly concerning group differences, but the host and guest advocate for navigating such topics with rigor and grace. The discussion also connects the general nature of the human mind's intelligence to AI design and the idea that proficiency in one area can generalize to others, as suggested by Miyamoto Musashi. The episode underscores that while factors like test anxiety or illness can affect an individual's score, they do not negate the underlying, stable g factor, which standardized tests like the SAT are also highly correlated with.

"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"

"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"

"it is the most replicated finding in all of psychology"

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

"it doesn't matter which battery of tests you use... those g factors will be highly correlated like over 0.9 with each other"

"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"

"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 that would require a ratio scale with an absolute zero"

"one of the most interesting things they found is a very strong relationship between iq measured at age 11 and mortality"

"the content doesn't matter but there it 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"

"the speed of information processing turn and reaction time is a measure of speed of information processing turns out to be related to the g factor"

"standardized tests like the sat they're essentially intelligence tests they are highly g-loaded"