Birdcage liners – Joel on Software

Algorithms, tuned not to help readers but to help advertisers. Intermittent reinforcement tuned to maximize engagement/addiction. This is some scary shit, but it’s the web in 2018. We can do better.

But whereas Twitter sort of stumbled upon addictiveness through the weird 140-character limit, Facebook mixed a new, super-potent active ingredient into their feed called Machine Learning. They basically said, “look, we are not going to show everybody every post,” and they used the new Midas-style power of machine learning and set it in the direction of getting people even more hyper-addicted to the feed. The only thing the ML algorithm was told to care about was addiction, or, as they called it, engagement. They had a big ol’ growth team that was trying different experiments and a raw algorithm that was deciding what to show everybody and the only thing it cared about was getting you to come back constantly.

Source: Birdcage liners – Joel on Software

And now, Facebook has seen the light! Its former executives are lining up to denounce the horrible things that Facebook has done. Zuckerberg is now pivoting away from algorithmic news (because hey that’s evil) toward algorithmic posts-from-friends.

Good news! Except that won’t help. It will only tighten the feedback loop and prop up the bubble. If you’re more likely to see things from your friends, you’re less likely to see things serendipitously. You only see what you agree with. Therefore everyone agrees with you. Bubble intensifies