Setting the feed to "Most Recent" usually gives better content, but the option unsets itself every other day.

The trivial one: Allow me to turn off all links. All of them, globally. I don't care about the news article you just saw that reinforces your political views; I don't care about the clickbait editorial you just read; I don't even care about the moderately interesting piece you just shared because I probably already saw it on Twitter. No links. Just the pictures and status updates like before the age of inline expand.

The non-trivial one: Allow me to turn off all images with text in them. Why? Because these are, with stunningly high frequency, visual versions of the same political-view-reinforcement article links. I already know if a friend or family member identifies as a liberal or a conservative. I don't need a regular stream of smug SomeEECards or Memegenerator pictures to remind me. On the technical side, there's clearly some fuzziness, but I have to think simply detecting the presence of text would be a comparatively simple task to what else Facebook's deep learning systems do.

I'm not familiar with Snapchat but I expect they have a similar mechanic. It seems like the newer social networks are avoiding the problems caused by virality, and getting back to the original reason we wanted to join social networks: to see what our friends are up to.

http://www.fbpurity.com/

>F.B. Purity is a browser extension / add-on that lets you clean up and customise Facebook. It filters out the junk you don't want to see, leaving behind the stories and page elements you do wish to see. The list of story types that FBP hides is customizable to your taste.

Is it actually trustworthy?

And Facebook isn't interested because it's not intended to build something that's everyone's perfectly tweaked feed.

It's much more in Facebook's interests to be a somewhat uniform "public space" than to be everyone's ideal filter. To some extent, it forces people to care stuff they wouldn't otherwise simply because the stuff is being made public.

Part of this is that we aren't Facebook's customers, we are Facebook's product and Facebook's advertisers are its customers. But that was essentially true of newspapers as well (newsstand revenues were supplement to advertising revenues even before the Internet).

To some small extent that can be a good thing since a human society requires some mutual awareness of our activity. Like newspapers, Facebook and other social media generate something like a space where public reactions can be seen. An information filter couldn't do that - for ill or good.

Would hiding all of them work? According to the post, the algorithm treats hides as negative feedback and tries to avoid those posts in the future.

But that said, there is no substitute for a fixed "most recent first" setting that doesn't reset, and frankly, the fact that it DOES reset is extremely creepy and leads me to use facebook much less often that I would otherwise.

Combine that with the sometimes-relevant accuracy rate of the algorithm and a bottomless well of junk posts to draw from and you have a recipe for a constantly mediocre experience.

There is also no way to "remember this" unless you like the post.

Second Paragraph: "No one outside Facebook knows for sure how it does this, and no one inside the company will tell you."

I worked on a component of newsfeed for a while, and yet had no clue how the rest of the system worked despite trying. That's the level of complexity involved here.

And:

> The infrastructure and all of the components involved in generating and rendering newsfeed at this point (and for a while) have been far too large for any single person to understand.

Each component should have a clearly and strictly defined way to interact with other components, which you should have in nice diagrams. Tack the diagrams together, and you get a full definition of how the newsfeed interacts and which piece does what.

If that is not possible, then you have a very badly managed company, not even Healthcare.gov was that mismanaged.

But then that ranking component can change week over week so if you thought you understood all of the ranking mechanics last week, then you probably have some catching up to do this week.

There's also a lot of people working on this so stuff changes fast and often. Think about all of Facebook's different products vyying for attention in the feed. Think of all the PMs with agendas who want their product in the feed in certain ways.

Plus side: if it works well, they can just replace their current unspeakable Lovecraftian horror of a system with the meta feed they develop. A new dawn for all.

If you can't maintain a complete understanding of the process, you are playing a guessing game.

This isn't some special magical FB-centric problem.

Like actually break them.

The rest of the article does go into many details about the behind the scenes of Facebook's algorithm. And shares a number of interesting details about how it works, and how decisions about it are made. So I would say the title isn't that bad.

(i guess they changed it themselves, as the URL shows this old title)

"A small team of engineers in Menlo Park. A panel of anonymous power users around the world. And, increasingly, you."

And they not claim they know themselves.

The whole internet is sort of evolving into photo content these days it feels like. Twitter, tumblr, pinterest, reddit...all very photo focused.

I missed repeated posts from a good friend asking for help. If I had seen them, I would have been there in 10 minutes. They committed suicide exactly 1 month ago. Just goes to show how incredibly important some algorithms can be.

I don't need fucking "intelligent algorithms" to tell me what is important, I can read an arbitrary newspaper if I want filtered content.

I've unfollowed all but a couple of my close friends, the stories that show up are inconsistent (some date all the way back to December 25th but will show up every now and then). Most of their "liking and commenting" activities show up in the news feed but not in the close friends list. It's a mess.

I actually wonder how popular "unfollowing" is and if facebook is worried about it, because everyone I know is psyched about the ability to do it!

I LOVE some of my friends, but I hate 99% of any post on the NFL. I'll certainly never seek it out. I unfollow people because it's just so frustrating to start to dislike the person ranting about something I don't care about, just because Facebook isn't smart enough to understand how human relationships work. In real life, I can tell that person, "Look, I love you, but you've got to stop talking about the NFL around me! You're showing me you're not listening to me or valuing me as a person by doing so..." And that's what Facebook is not doing: valuing me as a person. So, I unfollow everyone; the NFL is just one example.

... only if you don't understand machine learning.

Bottom line is that it constantly changes just like Google's algorithms for search change, and if you find one person to explain it to you correctly one day they will likely have it wrong next week.

It wasn't even relevant to the article really but stuff like that takes you out of it and you lose interest.

As for recommendation algorithms, I still believe Last.fm is one of the best around, so there might be something to learn from those guys.