One of my favorite—and incredibly goofy—examples of how the ways we model our environment help us make sense of the world around us comes from Amazon.
When you search for an item, Amazon makes suggestions of other things, such as in their “Frequently bought together” links.
Here, Amazon in the UK has served up items frequently sold with that particular baseball bat.
We humans get the joke right away. Yet it is a great example of how a theory can explain what correlation alone cannot.
People had bought these items together frequently enough—which is undefined—that the algorithm served up more things a mugger might want to obtain. Of course, a human would immediately understand the entire context. And I think it’s unlikely that they would want to offer recommendations for muggings.
The strength of Amazon’s algorithms is in linking things that seem unrelated, by simple correlations. The strength of humans is our ability to create a model of the world around us, test that model, and make corrections.
