Callum J Hackett gives some good advice in Reading the Unexpected:

This is why I prefer to follow people rather than topics. I’m able to get a good sense of their character and interests, and while I know what kind of wonderful links and commentary to expect 90% of the time — all part of the initial attraction — I also look forward to that remaining 10% which I’d never have predicted or sought out myself, but which I still enjoy reading.

We need that kind of spontaneous discovery. We need to be exposed to the unfamiliar and the unexpected, even if it’s only truly interesting one time out of a hundred. If all our interesting content is redirected from individuals to subject-specific sources, we will inevitably place subtle, unnoticed restrictions on the things that we see, and we will continue to reinforce our prejudiced ideas and interests without thinking.

This ties in well with a very interesting discussion between Susan Greenfield, Maria Popova, and Evgeny Morozov with the New York Times, weirdly titled Are We Becoming Cyborgs? Here’s Maria Popova:

The Web by and large is really well designed to help people find more of what they already know they’re looking for, and really poorly designed to help us discover that which we don’t yet know will interest us and hopefully even change the way we understand the world. […]

When you think about so-called social curation — algorithms that recommend what to read based on what your friends are reading — there’s an obvious danger. Eli Pariser called it “The Filter Bubble” of information, and it’s not really broadening your horizons.

I think the role of whatever we want to call these people, information filters or curators or editors or something else, is to broaden the horizons of the human mind. The algorithmic Web can’t do that, because an algorithm can only work with existing data. It can only tell you what you might like, based on what you have liked.