It is widely recognized that centralization, such as one sees in most social media (Facebook, Twitter, etc.) has strengths which bring concomitant dangers. Possibly the biggest danger, and certainly the most pressing on people’s minds, is censorship.

Distributed media, such as blogs, make censorship much harder. However, this has (so far) been at the cost of making discovery much more difficult. Finding new blogs is a very haphazard things, to a great degree relying on cross-promotion in blogs. By contrast YouTube is able to leverage its centralized information to provide a list of recommended videos to the user after each video. Given the massive data available to them of what people watched and how long they watched it for, this enables them to make recomendations for videos which are often good. Almost every YouTube channel I’m subscribed to I found through recommended videos.

It probably goes without saying, but unfortunately recomendation engines are extraordinarily succeptible to manipulation by hosts with an agenda. Moreover, it would be virtually impossible to discover such manipulation as it’s only relevant to people who are not aware of particular video makers anyway.

In order to make distributed media truly competitive with centralized media, what we really need is a system for making distributed recomendations. It’s not immediately obvious that this is doable, unfortunately. A system of distributed recomendations would be a spammer’s dream ifi they could figure out how to manipulate it. In fact, most parties would be deeply desirous of manipulating this system. This guarantees that a lot of effort would be put into trying to figure out how to game the system. Worse, the system would almost certainly need to be anonymous, so as not to track people’s reading habits, which makes fake recomendations all the harder to defend against.

The most obvious approach to avoiding spam would be to attach micro-payments to the recomendation system. That brings with it its own problems, but it also has benefits. There are probably other options, too. Especially if one were to somehow include negative reviews as well as positive reviews, the correlations required to spam people might need to be far too complex to allow for useful spamming.

Anyway, if you know anyone who likes to develop algorithms, try planting the seed of a distributed recomendations network. It may not be doable, but the internet would benefit tremendously from it if it is.