there isn't one intelligence
there are multiple ones
learning and teaching should be adapted to the strengths of each child
empirically, there is a high correlation between these "different" intelligences
validation of educational approaches based on the theory has been limited
differences in preferences and skills
learned differences (rather than innate factors)
physiological differences other than intelligence
Does this even meet our definition of intelligence?
Psychometrically, we define intelligence as the "g factor", so this is everything but the "g factor".
These measures are still very useful, but theyprobably shouldn't be called "intelligence"
test performance on a variety of mental tasks
see whether there is a common factor that can account for correlations among performance for all of them
intelligence is measured as an "IQ"
quotient of what?
original definition, give tests to a variety of individual at different ages, performance increases with age, for a given performance, find the average age at which that performance is achieved, this yields the "mental age", divide the mental age by the physical age, hence the "intelligence quotient"
modern definition, for a given population and age group, the mean performance on a test is scored at 100, one standard deviation within the population is a change in 15 IQ points, intelligence scores are distributed normally, so this lets you easily translate the scores into population frequencies, intelligence scores are distributed normally because they are the aggregate of many individually random decisions (central limit theorem), both at the question level and possibly at the level of many different individual skills
short term memory
long term memory
correlations among school performance
correlations among intelligence test subtests
There does seem to be something like a "g factor" (general intelligence).
What does it mean?
causal models, how does one variable cause another one?, i.e., if we intervene and change one variable, how does the other variable behave?
factor models, how can variation of a quantity be explained as the variation of component quantities?
Spearman's Two Factor Model
Carroll's Three Stratum Theory
note: correlation is highest among people with low intelligence test scores (what does this mean?)
intelligence measures are no good if all they do is predict performance on intelligence tests
how do they relate to real-world performance?
academic performance, correlation is high in elementary school, about 0.6-0.7, correlations drop to about .5 in college, and .4 in graduate school
job prestige, very high correlation (.9) between prestige of an occupation and the average IQ of people who take it, what does this mean?, correlation with average IQ, prestige, prestige, occupational prestige (USA), occupational prestige (Germany), high correlation (.7) with individual IQs, low dispersion of scores in highly prestigious occupation suggests minimum requirements
job performance, how does IQ correlate with job performance within each job?, generally, good correlation (.55), highest for jobs with high complexity, job specific tests are no better at predicting performance than general IQ tests
income, correlation between IQ and income is about .4, somewhat higher if you look at peak income
other correlates, morbidity and mortality, crime
moderate correlation with brain volume (.3-.4)
reasons are unknown, maybe "more hardware" makes you smarter, bigger brains correlate with various other factors, such as childhood nutrition and body size, these other factors may be determining
small correlation with height (.2), may be due to mating preferences rather than causal factors
"g factors" can also be identified in animals (mice, rats, monkeys), correlating problem solving and learning on a variety of tasks, what kinds of problems do these animals solve?, we'll talk about animal intelligence later
classical view of intelligence
deduction and induction
trying to explain all of human intelligence through logic falls flat
common trope in SciFi ("Mr. Spock")
limitations of logic, fails to take into account randomness (but logic is a special case of Bayesian probabilistic reasoning), computations can be very expensive, so the brain may not be able to carry them out, traditional logic doesn't deal well with time, incorrect information, belief, etc.; AI has developed more complex "logics", where does it come from? Is it hard-wired? How?
Generically: given some start state and some constraints, reach some goal state.
When it is a mental process, it is often the same as "planning".
Solve the problem mentally before solving it physically, since physical exploration is more costly than mental exploration.
Long term planning involves decision making about probabilities, expectations, and risks, as well as preferences.
Retain past experiences and apply them to new situations.
There are two fundamentally different kinds of learning, episodic memory, skill learning
You can lose episodic memory while still retaiing skill learning.
Transform complex physical realities into a stream of symbols, and back.
recognize objects in 3D scenes
recognize objects from the sounds they make
identify objects by touch or smell
understand spoken language
produce spoken language (far more complex than playing the piano!)
walk upright (small errors mean death)
handle and manipulate tools