Loading...
In 2017, researchers at the Yale Human Dynamics Lab made a disturbing discovery. They tracked Twitter users over time and found that people who received more likes and retweets for expressing moral outrage gradually increased the outrage in their posts — even when the topics they were responding to hadn't become more outrageous. The platform was training people to be angrier. The mechanism was simple reinforcement learning. A user posts a measured opinion: 15 likes. The same user posts the same opinion with outraged language: 500 likes, 200 retweets, 30 replies. The user learns, unconsciously, that outrage is the currency of attention. Over weeks and months, their baseline emotional register shifts. What began as genuine engagement with issues becomes a performance optimized for engagem...
Popular framing: Social media has made people angrier because bad actors spread inflammatory content and weak-willed users can't resist engaging with it. The solution is better content moderation and more personal discipline. The role of 'Social Proof'—how the crowd's reaction *defines* the individual's emotional state, rather than just reflecting it.
Structural analysis: Outrage is an emergent attractor state produced by a feedback loop between user behavior and engagement-maximizing algorithms. Reinforcement learning operating below conscious awareness gradually shifts users' emotional baseline — not through manipulation by bad actors but through ordinary reward signals. The system is a Goodhart's Law failure at scale: engagement was chosen as a proxy for value, but the metric was captured by its most reliable driver (emotional arousal), which then became the system's de facto optimization target. The Red Queen dynamic ensures that any intervention short of restructuring the core incentive architecture will be adapted around. The 'Zero-Sum' nature of the attention market, which forces 'Outrage' to be the only viable survival strategy for a digital identity.
The gap matters because popular framing generates solutions (bans, personal discipline campaigns, content warnings) that treat symptoms while leaving the feedback architecture intact. These interventions are not merely ineffective — they may accelerate the dynamic by creating outrage about censorship and training users to find new outrage-generation strategies within the new constraints. Accurate structural diagnosis would redirect effort toward metric redesign, business model alternatives, and architectural friction that breaks the reinforcement loop before conditioning occurs.
