Thursday, July 30, 2015

Review: The Signal and the Noise: Why So Many Predictions Fail - But Some Don't

"You can get better, so do better" is a weird message to take away from a book that is ostensibly about how bad humans are at statistics, but its actually the consquence of a simple tool: Bayesian reasoning. Bayes' formula is an agnostic equation. Imagined into words by a Scottish pastor and put into math by a French atheist, it tells us, "Given what you know, you can be more accurate in the future."

We can, as a society, embrace the Bayesian way of thinking and instead of polarizing ourselves, update our worldviews to take in new evidence and reach consensus quickly about how things are happening.... or maybe we can't. For every bit a pop statistics in Signal and the Noise, there is some psychology 101 and Econ 200 that has to be rehashed: humans are irrational creatures designed for an environment we're not actually living in, and, taken as a whole even individually rational behaviors can turn out to be irrational. Its a quagmire, but not one that Bayesian reasoning doesn't allow us to see our ways out of.

Another big thing that Silver insists on is that we can't be a slave to computers or big data. Eh. Silver seems to be bringing out the same luddite arguments of yesteryear, namely, "They can't do what we can!" which is surely true. Computers and Big Data do, currently, need many hands helping them. But the entire point is to make those helping hands computers, too.

Must read.

No comments:

Post a Comment