Comments on: Books that make you stupid? http://www.hartnup.net/wordpress/archives/2008/01/25/books-that-make-you-stupid/ John's all-purpose blog, with Debbie's name in the title too! Thu, 08 Jan 2009 17:34:02 +0000 http://wordpress.org/?v=2.6.5 By: Brendan http://www.hartnup.net/wordpress/archives/2008/01/25/books-that-make-you-stupid/#comment-38734 Brendan Fri, 25 Jan 2008 20:21:33 +0000 http://www.hartnup.net/wordpress/archives/2008/01/25/books-that-make-you-stupid/#comment-38734 The SAT is used in America as a college entrance examination. It was an acronym for "Scholastic Aptitude Test" for years, but (like many "intelligence" tests) was scientifically shown to have zero correlation to scholastic aptitude. Colleges continue to use it because it's the status quo, and they have no better ideas. A "School" in this context (as far as I can tell) is an American college that has a Facebook network. All of the ones that I saw on the list are 4 year colleges, but it seems feasible that a 2 year college could be included. While it's an interesting data mining exercise, I wholeheartedly agree that it produced a largely meaningless set of results. The SAT tests your ability to take a standardized test, and is only a good predictor insofar as standardized test taking is concerned. Since the data are grouped by school, average SAT scores are used; this introduces an incredible amount of imprecision. Without considering the standard deviation (or a higher granularity) of scores from a given school (for example, my SAT score was nearly 350 points higher than the average at my school), it's silly to attempt correlation. It's also silly to take the top 10 books from a given school, since this method effectively limits us to popular fiction and required high school texts. This worst assumption is that enjoying certain books and being smart have a causal relationship. This doesn't take into account the fact that some books impart knowledge to the reader without the reader's enjoyment. It begs the question of intelligence and artistic taste, rather than investigating it properly. The actual data samples are so opaque (it's like comparing averages to averages) that the end result is slightly humorous, but ultimately meaningless. I'd be loathe to put my name on a study like this. The SAT is used in America as a college entrance examination. It was an acronym for “Scholastic Aptitude Test” for years, but (like many “intelligence” tests) was scientifically shown to have zero correlation to scholastic aptitude. Colleges continue to use it because it’s the status quo, and they have no better ideas. A “School” in this context (as far as I can tell) is an American college that has a Facebook network. All of the ones that I saw on the list are 4 year colleges, but it seems feasible that a 2 year college could be included.

While it’s an interesting data mining exercise, I wholeheartedly agree that it produced a largely meaningless set of results. The SAT tests your ability to take a standardized test, and is only a good predictor insofar as standardized test taking is concerned. Since the data are grouped by school, average SAT scores are used; this introduces an incredible amount of imprecision. Without considering the standard deviation (or a higher granularity) of scores from a given school (for example, my SAT score was nearly 350 points higher than the average at my school), it’s silly to attempt correlation. It’s also silly to take the top 10 books from a given school, since this method effectively limits us to popular fiction and required high school texts.

This worst assumption is that enjoying certain books and being smart have a causal relationship. This doesn’t take into account the fact that some books impart knowledge to the reader without the reader’s enjoyment. It begs the question of intelligence and artistic taste, rather than investigating it properly.

The actual data samples are so opaque (it’s like comparing averages to averages) that the end result is slightly humorous, but ultimately meaningless. I’d be loathe to put my name on a study like this.

]]>