Found in 3 comments on Hacker News
marketforlemmas · 2014-11-18
· Original
thread
For people who enjoyed this article, I highly recommend Duncan Watts' book "Everything is Obvious" [1]. Its basically the thesis of this article expanded out over the course of a book with a lot of interesting examples and lessons drawn from scholarship (both Watts' own work and many others).
[1]: http://www.amazon.com/Everything-Obvious-Common-Sense-Fails/...
sigkill · 2012-11-26
· Original
thread
>We like to project an unsurprised, unperturbed, and confident attitude. But thinking "I knew it all along" can...
For anyone who's interested in these topic and how people behave when facts are presented in one way or another, take a look at this book. In my opinion, it's completely worth reading this book.
http://www.amazon.com/Everything-Is-Obvious-Common-Sense/dp/...
It isn't a great book, but it does clearly identify the difference between explanatory and predictive accounts, and makes a strong case that explanatory, after-the-fact, accounts are simply noise. The exercise the author engages in here, of "making sense" of two contradictory datasets can be done with almost any phenomenon.
The more systematic and inclusive data gathering is, the more strongly any explanation is going to be tied to the phenomenon it is going to explain, and historians--like any science--should always be asking themselves, "How can I test this explanatory idea? If it is true, then what else ought to be true?"
Otherwise, when you "explain things a lot" you are doing nothing but generating noise. The financial press provides strong evidence for this: tens of thousands of words of "explanation" every single day, and not a single person made rich by any of the predictions the same people make. When you can explain everything and predict nothing, you had better be able to test your explanations by indirect means if you want to be taken more seriously than a fabulist.
The latter part is important: historians and others like them expect us to take their pontifications seriously. But we know the untested explanation is almost certainly nothing but a confabulation. So why should we take anything historians say seriously?
The history of science--which I define inclusively as "publicly testing ideas by systematic observation, controlled experiment and Bayesian inference"--is one of realizing we've been doing it wrong, that ideas we believed and took seriously were actually rather silly. Blood circulates. Geese don't come from barnacles. Bad air or moral turpitude don't cause plague. And so on.
If historians are starting to ask, "Are we doing this wrong?" that's a good thing, and based on my understanding of what science is and how it works I predict that over the next few decades the "digital humanities" will make a lot of traditional beliefs look untenably foolish, and there will be a general shift of the field toward Bayesian methods that will alienate and annoy a great many people whose freedom to confabulate will be greatly curtailed. It will change the face of the humanities to the extent that they will have as little to do with their historical roots as modern psychology has to do with the story-tellers of the early 20th century who gave it its start.