I think recommendation algorithms provide a useful service, but I think the main issue is that for a lot of sites there's no way to explore the content _without

I think recommendation algorithms provide a useful service, but I think the main issue is that for a lot of sites there's no way to explore the content _without_ that algorithm which is heavily biased in your favour. Things you dislike are minimised and slowly hidden from you to make it seem like no one likes these things. Things you enjoy are proliferated to make it seem like everyone likes these things. This gives a warped view of the world. Showing a random article 10% of the time isn't going to help with that. An experimental interface I worked on last year (and am working on) is using classification systems to let people explore the books of Open Library, https://openlibrary.org/explore https://openlibrary.org/explore (beta warning :) ). The idea being that you can quickly dive deep into a topic you're interested in, but just because you're interested in "Science", that doesn't mean the "Arts" or the "Religion" bookcases adjacent to it will become any smaller. You're always in the context of the bigger picture of the entirety of human knowledge.