CODEXCard 3 of 9 · Modern Internet Patterns
Engagement-based algorithms (Engagement Maximization as Moral Position)
A platform that optimizes for time-on-app has already taken sides.
Engagement-based algorithms rank posts by how likely they are to hold your attention or provoke a reaction. Every recommendation system has to choose what to put first, and many platforms choose engagement because it ties directly to time spent and advertising revenue. That choice is often presented as neutral, as simply showing people what they want.
It is not neutral, because the target shapes the result. Material that triggers strong emotions, tribal loyalty or anger tends to keep people engaged, so an algorithm tuned for engagement surfaces more of it. Over time, the mix of what a whole society sees shifts toward whatever performs best on that one measure. Deciding what to optimise for is deciding what gets amplified.
The opposite mistake is to conclude that every platform is harmful and that the only answer is to log off forever. Platforms differ in what they reward, and some are built around goals closer to yours, such as learning, staying in touch or finding specific information. The practical skill is noticing what a service optimises for and adjusting how much of your attention it gets.
This is Goodhart's law at the scale of a culture. When engagement becomes the target, it stops being a good signal of what people actually value.
Example
You open a video app to look up a quick recipe and, forty minutes later, you are watching strangers argue about politics. Nothing forced you there, but each next suggestion was chosen because it was likely to keep you watching. Searching directly and closing the app afterward gets you the recipe without the detour.
Try it this week
This week, check one app's settings for a chronological or following-only feed and try it for a few days.
From Modern Internet Patterns (Internet Theory).
