Engagement-optimized feeds personalize reality, reward outrage, and quietly narrow what you see.
Recommendation systems optimize a measurable proxy — usually engagement — not your wellbeing or the truth. Because outrage, novelty, and tribal affirmation drive engagement, the feed learns to serve more of them. Personalization then creates a private information environment ('filter bubble') that feels like the world but is a model of your past clicks. The result is a persuasion loop: your behavior trains the model, and the model shapes your next behavior. Seeing the loop is the first step to stepping out of it.
Investigate with the AI detective