June 5, 2012

What Computers Want

Correlation isn't causation, but "Jeremy" and "cool"
both started surging around the year I was born

I really believe that our tools affect the way that we see the world. David Weinberger's Too Big To Know argues that the web has forced us to see knowledge as linked and foundationless - like a network, not linear like a book.

In a larger sense, I think that computers affect the way that we see the world. For the first time ever, we can store and manipulate huge amounts of data - we can quantify things that never seemed quantifiable before. For example, Google's Ngram Viewer allows you to analyze millions of published works to see how phrases have become more or less popular over time. People have used this as a way to quantify culture.

There is danger, however, in thinking that everything can be quantified. As Jaron Lanier argues in You Are Not a Gadget, when we quantify our lives, we may be tempted to think that they are fully quantifiable. Or worse - we may begin to live quantifiable lives. He argues that Facebook is "organizing people into multiple-choice identities", and that if we continually describe ourselves in those sorts of terms, then we may become flatter, describable, and boring people.

My deeper reaction, though, is that this is really cool. I think it's amazing that we can track all of the websites we visit, or how many steps we take or all of the places that we go, and find patterns and meaning. Thinking like a computer leads to a whole new way of seeing the world.

For example, Google took over search because they came up with the idea of tracking all of the links that go between websites. Before Google, search engines searched based on just the content of the page (or searched a human-created directory of websites). Larry and Sergey's great insight was that there was data in the network, not just in the nodes.

There are plenty of other examples of the insights gained from this sort of computational thinking - on Hunch.com they track correlations of how people answer questions. For example, they know that gel pen users are significantly more likely to drive to work than pencil users. Most of that is silly, but the fact that we can track and share these things is really cool, and I think that we will continue to find insights.

No comments: