What Do I Read Next?
Who hasn’t encountered Amazon’s famous “People who bought this product also liked…” recommendations? More often than not, those recommendations aren’t annoying shots in the dark guessing what you might like; they are actually pretty good. But the algorithm wasn’t always this good. In fact, in the beginning, when Amazon didn’t have a history of your previous purchases, they had human critics recommending “the Amazon voice” on its site. Sounds so quaint now, doesn’t it? Then, when the algorithms replaced the humans, they weren’t exactly smart. As James Marcus described the initial algorithms in his book, Amazonia : “It felt as if you had gone shopping with the village idiot.” Then Amazon realized it didn’t have to customize suggestions based on individuals ; rather they could just identify correlations between products ; and make their recommendations based on the latter! But even as the algorithms got better, there remained several problems for Amazon to improve their recommen...