Comment by pixl97

There are a few particular problems with this. Let's go with the most extreme example I can think of.

You walk up to the interface of an infinite digital library. In this library there are an infinite amount of books that you'd like to read, or even contain anything coherent for you to read. The problem is there is also an infinite amount of books you don't want to read, can't read, or aren't even coherent information. And this second infinity is so much bigger than the first infinity that it is infinitely improbable that you will ever pick out a coherent book.

Everything you do in a coherent informational world is directed by sets of biases. Algorithmic recommendations are just another one of the tools used to traverse this infinite world, and if you understand the biases they impart you can use these tools successfully.