Bits of Books - Books by Title


Uncontrolled

The Surprising Payoff of Trial and Error

Jim Manzi



When you try to predict how people will respond to some intervention (offer, incentive, change of environment etc), you can come up with an analysis that 'proves' anything you want. And often all it takes is a tiny change in initial assumptions to radically alter the expected outcome. There is no alternative to controlled experiments.

This is a problem because there are too many self-proclaimed experts who want us to follow some pre-determined policy that will lead to (insert yr goal here) happiness/prosperity/success.

Two important concepts. First is evolution, which shows how a system can capture and exploit implicit insight without creating explicit knowledge; a model of trial and error. Second is randomized clinical trials, which test competing claims and provide data over anecdote as a way of deciding the best way forward.

vast majority of reasonable-sounding interventions will work under some circumstances, and not under others. For example, school literacy programs - one may work in middle-class white district but fail in lower-class or multi-cultural setting.

"Life is a perpetual instruction in cause and effect." Ralph Waldo Emerson

A scientific theory can never be proven beyond doubt, bc there is always some hypothetical situation or experiment which may nullify it.

Karl Popper's concept of Falsification. A scientist starts by developing a hypothesis, tests t and tries to disprove it. If he can't disprove, he puts it out into the world, and other scientists try to disprove. As more and more scientists fail to disprove it, it becomes accepted as a useful explanation and practical guide.

If you can't design and conduct a test that will disprove the idea then it's not worth debating because it doesn't help in any way.

Science can never deliver absolute Truth. When we say something is scientifically proven, this is shorthand for "a group of competent scientists believe this theory has passed many falsification tests and can therefor be treated as reliable in practice."

Thomas Kuhn put forward idea of a paradigm, a concept which has been widely misused ever since. He suggested that in order to make practical progress, scientists accept an underlying set of assumptions about the physical world and how it works. This paradigm creates a coherent discipline, giving scientists the structure to perform their day-to-day experiments. Anomalies - factual observations which contradict the paradigm - are either held outside the paradigm as problems to be solved later, or the paradigm is modified slightly to accommodate them.

This model works well for a while until enough anomalies accumulate unresolved or the paradigm is twisted so far out of shape that it becomes unworkable. When that happens, someone comes along, often from outside the core speciality, and proposes a new paradigm that works better. This is Kuhn's paradigm shift.

Classic example of this was overthrow of paradigm that everything revolved around the Earth. Originally the idea worked but as observational accuracy improved astronomers had to keep adding epicycles (little circles within circular orbits) to make it cohere. Became unwieldy to point of being crazy, and still didn't fit data well. Eventually Copernicus suggested that Earth moved around the Sun. But because he, like everyone else, assumed circular orbits, that was rejected at first because it was even crazier. Took Kepler to realize that elliptical orbits solved all the problems.



































Books by Title

Books by Author

Books by Topic

Bits of Books To Impress