A social media platform uses an AI to sort posts in users’ feeds.
The goal is to show content that people find most interesting.
To do this, the AI measures how long users view a post,
whether they comment on it or share it.
After some time, researchers notice something strange:
More and more posts appear extreme, angry, or provocative.
Calm and factual content appears less often.
The platform never intentionally preferred such posts.
Yet the algorithm seems to push them to the top.
Question:
Why can an algorithm that simply wants to show interesting content start favoring extreme posts?
Solution will be unlocked tomorrow.