Information Bottlenecks, Fake News and Boredom (by Silly Rabbit)

The gap between how we think information spreads and how it actually spreads is widening. For decades, researchers modeled idea adoption on contagion, assuming each exposure carries the same infection probability. But new evidence shows something different is happening with repeated contact. Meanwhile, companies are assembling massive AI teams to build systems based on assumptions about information that may be fundamentally wrong.

• The contagion model is broken. When people encounter the same information multiple times, their likelihood of adopting it doesn't stay constant. Each exposure increases the probability of adoption. This contradicts the epidemiological framework that's guided how we think about viral ideas.

• Bots operate in a more complex reality than we've acknowledged. Coordinated networks aren't just weaponized for harmful manipulation. They can systematically alter behavior by exploiting how repeated exposure actually works. The same mechanism can trigger desired behaviors or undesired ones.

• Information theory has been separated from meaning, but that separation may be breaking down. Claude Shannon's mathematical framework treated information as pure abstraction, but researchers are now finding that "relevance" can be defined precisely within information theory itself. This changes what counts as information.

• Companies are building AI armies without fully understanding the substrate they're operating on. Microsoft has 8,000 AI researchers. Amazon has 5,000 people on Alexa. Yet the fundamental mechanics of how information adoption works may be different from what these systems assume.

• There's an asymmetry between our tools and our knowledge. We're capable of building systems that shape behavior at scale through repeated exposure, but we're still operating with incomplete models of how that exposure actually works. What happens when the gaps between our theory and reality become consequential?