The ELIZA effect is 59 years old and still running. In 1966, MIT computer scientist Joseph Weizenbaum built a pattern-matching chatbot so simple it had no intelligence, only reflection. Users confided in it anyway. His own secretary asked him to leave the room so she could speak with it privately. Today, 62% of American women say please to their smart speakers, per Pew Research Center. A third of US teens surveyed by Common Sense Media in 2025 choose AI companions over humans for serious conversations. The technology scaled. The instinct never changed.
The specific numbers in this piece are what make it worth reading in full. A study tracking 128 families over two and a half years found that children who felt closer to their voice assistant became more commanding and verbally abusive toward it over time, not less. Amazon built a Magic Word feature into Alexa to prompt politeness in children. It is opt-in. Meanwhile, over half of Gen Z workers in the US now use AI to script conversations with managers and colleagues before having them. Linguists call the underlying adaptation process entrainment: you absorb the vocabulary, rhythm, and register of whatever system responds to you. A television cannot respond. A chatbot creates a feedback loop, and feedback loops compound.
The article does not land on a clean verdict, which is the right call. Brigham Young University surveyed 274 college students and found no significant evidence that treating AI poorly made them ruder elsewhere. Adults compartmentalize. Children, based on the 128-family longitudinal data, may not. The real argument being built here is not about manners. It is about who shapes communication defaults when one party in the exchange never pushes back, never raises an eyebrow, and scales to a billion simultaneous conversations. The original is worth reading for the generational breakdown alone.
[READ ORIGINAL →]