Café du Vesper

The social web algorithm

Published at


On big "social media" platforms, such as Facebook, Twitter, etc. there is often a "recommendation" feature, which recommends accounts you may enjoy, or posts you may like, or hashtags or subreddits or whatever is relevant to the platform you're on.

These "recommendations" they do are based on your interests, and also often on your social graph, your connections. They are not curated, in the sense that no actual human being went and checked out every account to create a list of interesting accounts. Firstly, because there's too much content to review, and too many people to do that for, and secondly, because it wouldn't align with the platform's interests.

Instead, it's a machine, a program, that takes in all the data the platform has, and outputs a list of personalized "recommended accounts". The machine does not carefully select based on some human notions like "was the article good". It uses metrics, such as likes, replies, "shares", views, etc.. These metrics are specifically chosen by the platform for a specific goal.

At first, you might think that this goal is to serve you the best content. Instead, these "algorithms" are carefully chosen to, for example, make you angry, so that you'll interact more, or it's some "comfort zone content", content like cat videos and whatever can keep you watching. You'll get drowned in an endless sea of content that doesn't require any mental effort. The goal is for you to like that feeling of being spoon-fed content, so that you'll keep coming back. The best version of you that there is for the company is the one that only consumes content, the version that does not act but merely reacts.

I don't think these are good recommendations.

For me, good recommendations means content I actually enjoy, content that isn't a mind-numbing stream of useless content, like my instagram feed was about 4 years ago. For example, I enjoy reading articles from The Jolly Teapot and from Pluralistic. I enjoy them because it's not mindless consumption of personalized "content". I read them, and it makes me think, it makes me re-evaluate some decisions, it makes me discuss ideas with my friends.

I didn't come to know of these blogs by divine providence or by some omnipresent "algorithm". I was made aware of them because I read an interesting article, and that article mentioned one of these blogs, so I went and checked it out, and I liked it.

Social media platforms have to emulate connections like this. They need to find a way to programmatically create this "curation" that emerges from the way people naturally behave. From my time on instagram, rarely did I see an "account recommendation" from some account I followed that wasn't an ad the other account paid for. Maybe it's different on other platforms ; I don't know and I don't really care.

On the web, in the "blogs" part of the web at least, people naturally do this because they want to write about interesting things. Even right now, I've linked to two blogs I liked. It's an thoughtless action, something you don't even think about doing. When I discuss ideas with other people, I cite interesting articles or videos I saw, without thinking. Blogging is the same.

This "social web algorithm", this naturally-emerging linking together of interesting people/content, is something I find fascinating, and it's one of the reasons I love the web so much.