The post claims to analyze how people start attributing “personas” or agency to LLMs — but the language used throughout (“persona X”, “the model wants”, “something gets activated”) actively reinforces exactly that framing. From a psychological perspective, this is classic induced anthropomorphism + agency projection. You’re describing a cognitive illusion, but at the same time narrating it in agentic terms, which predictably strengthens the illusion in readers. LLMs don’t have personas. They don’t have intentions, continuity of self, or internal goals. What people are responding to are stable statistical patterns + their own hyperactive agency detection. Ironically, the article itself becomes part of the memetic mechanism it’s trying to warn about. If the goal is epistemic hygiene, I’d strongly recommend much plainer language: “stable output patterns” instead of “personas”, “user projection” instead of “entities”, etc. Otherwise this easily slides into techno-mysticism dressed up as rational analysis.
Just my 2 cents — from someone who spends a lot of time dealing with projection and meaning-making in humans.
The post claims to analyze how people start attributing “personas” or agency to LLMs — but the language used throughout (“persona X”, “the model wants”, “something gets activated”) actively reinforces exactly that framing. From a psychological perspective, this is classic induced anthropomorphism + agency projection. You’re describing a cognitive illusion, but at the same time narrating it in agentic terms, which predictably strengthens the illusion in readers. LLMs don’t have personas. They don’t have intentions, continuity of self, or internal goals. What people are responding to are stable statistical patterns + their own hyperactive agency detection. Ironically, the article itself becomes part of the memetic mechanism it’s trying to warn about. If the goal is epistemic hygiene, I’d strongly recommend much plainer language: “stable output patterns” instead of “personas”, “user projection” instead of “entities”, etc. Otherwise this easily slides into techno-mysticism dressed up as rational analysis.
Just my 2 cents — from someone who spends a lot of time dealing with projection and meaning-making in humans.