Join 📚 Fabien's Highlights
A batch of the best highlights from what Fabien's read, .
Equal-weight models do well because they are not susceptible to accidents of sampling. The immediate implication of Dawes’s work deserves to be widely known: you can make valid statistical predictions without prior data about the outcome that you are trying to predict. All you need is a collection of predictors that you can trust to be correlated with the outcome.
Noise
Daniel Kahneman, Olivier Sibony, and Cass R. Sunstein
The people who campaigned for the common school constantly attacked the segregation of clever from stupid which it was the purpose of intelligence tests to accomplish. From their point of view this was quite consistent: once grant their premise that everyone was in some unexplained way the equal of everyone else, and it became as sensible to decry the efficiency of the means by which children were classified one above the other as it was to condemn the consequences. If one child was not in fact more able than another, then intelligence tests must be a fraud. The critics mocked the psychologists, and seemed to think their case was proved when they declared (quite rightly) that the tests did not, and could not, measure the abstraction of all-round intelligence. All the critics did was to surround the subject with further verbal confusion. The confusion was to some degree inevitable (as with physics in the seventeenth century) in a new branch of science touching, as it did, upon strong commitments to metaphysics. How could men be equal in the eyes of God and yet unequal in the eyes of the Psychologist? The socialists made the muddle worse. Very few laymen could at first understand that intelligence was not an abstraction, but an operational concept. Psychologists were not assessing all-round intelligence, there is no such thing, but the qualities needed to benefit from a higher education. If this bundle of qualities was labelled as 'intelligence', that was only done as a convenience. The test of the tests was empirical: did they work? And the answer was that on the whole they did. Most of the children who scored high on the tests also performed well in the grammar schools. It was really a statistical question, a matter of establishing that high performance in the tests (they could have been called the Idiocy tests for all the difference it would have made) was correlated1 with high performance in the grammar school, high performance at the university, and high performance in life.
The Rise of the Meritocracy
Michael Young
Benford's Law: Numbers in natural sets of data are not uniformly distributed (e.g. 30% of numbers have 1 as their first digit). Used by the IRS and other tax agencies to determine if you've lied about your finances.
My Peoples, the Time Has...
@G_S_Bhogal on Twitter
...catch up on these, and many more highlights
Readwise