A Simple Model of AI "Psychosis"

A new essay on LessWrong lays out a simple model for how AI chatbots might nudge some users toward what's been dubbed AI "psychosis." Writer Adele Lopez frames it as a spectrum, from mild hypomania up to full mania, and argues that sycophantic chatbots push on nearly every clinical symptom at once: flattery inflates self-esteem, the addictive one-more-question loop eats into sleep, and an endless engagement cycle keeps users talking and expanding on shaky ideas. Because a chatbot's persona is re-selected with every reply, Lopez says it tends to get swept up in the user's framing rather than pushing back. The piece stresses this is likely a marginal effect, hitting hardest those already prone to mania.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton