Wednesday, September 22, 2010

What Intelligence Tests Miss, Keith Stanovich

Keith Stanovich's What Intelligence Tests Miss does a reasonable job of arguing that we have a couple of different things in mind when we talk about how "smart" a person is, and that some of the important aspects are very different from what IQ tests measure. His goal seems to be to convince us that the other parts are important and we would do better if we either found good ways to measure them (though there are caveats there) or reduce the societal importance of IQ tests and their ilk.

Most of the areas that Stanovich is interested in could loosely be called rationality skills. He starts out the book with the example of George Bush, whose apparent IQ (estimated from various of his tests results that are on the record) is about 120, but who is agreed to not have conventional smarts, or be a thorough, consistent, or deep thinker. The main point here is talking about how people are surprised, but shouldn't be, that IQ is separate from what we call smart. The book is mainly a riff on Kahneman and Tversky's work on human decision making, and all the kinds of rationality traps that we fall for.

Apparently, Stanovich's own research is in how the various layers of processing—the Autonomous mind, the Algorithmic mine, and the Reflective mind— interact and override one another in order to determine the kinds of processing we do. We spend most of our time in autonomous mode, with occasional incidents propelling us into slower Algorithmic thinking, and only rarely do we have a reason to actually think about what we're doing reflectively. Stanovich has a detailed model showing the interactions, and pointing to the Reflective mind as the director that gives the signal for when to invoke the Algorithmic level. His argument seems to be that people who don't "act smart" fail to engage their Reflective layer, and so end up on auto-pilot most of the time.

The rest of the book is mainly a rehash of the literature on rationality errors, and a plea for approaches like Thaler and Sunstein's Libertarian Paternalism, which are intended to provide support for people so they can get smarter results without having to reason more clearly.

In the end, I guess I'd say that there are some interesting ideas here, but not enough to make the book worthwhile.

No comments: