Online Library / Finding Something To Read

A recommendation and an algorithm are not the same thing

One is a person guessing at you. The other is a machine guessing at people like you.

A recommendation from a person carries information that no system can reproduce, because the person is not only thinking about what you liked. They are thinking about what you are like, what you have been going through, what you complained about last time, and what they suspect you would enjoy despite yourself. A good recommender occasionally hands you something you would never have chosen, and is sometimes wrong in interesting ways. That wrongness is a feature. It moves you. A machine will rarely take that risk on your behalf, because being slightly wrong in an interesting way is not something it is built to attempt.

An automated suggestion works differently. It observes what you finished, notices what similar readers finished, and offers more of the same shape. It is genuinely useful and it is very good at what it does, but its purpose is to reduce your chance of disliking the next thing. Over time that quietly narrows the range of things you are offered, because the safest prediction is always something close to what you already chose. You end up in a comfortable groove that you did not consciously choose and cannot easily see from the inside.

The practical fix is not to reject either one. Use the machine when you know what you want and want more of it. Use people when you do not know what you want, which is more often than we admit. Ask the person at the desk, whose job includes exactly this, or ask a friend whose taste is unlike yours rather than one whose taste matches. Then, occasionally, ignore both and pull something off a shelf for no reason at all. A reading life needs a certain amount of unexplained input or it slowly closes in on itself.