In Search Of AI Psychosis

A strong essay combining clinical expertise, conceptual frameworks, and original empirical work on the new phenomenon of 'AI psychosis' (people destabilizing after heavy chatbot use). Explores it via precedents: 'Lenin Was A Mushroom' (the 1991 Russian TV hoax - most people lack world-models and are moored only by social consensus / 'epistemic vibes,' and some treat AI as an 'official' source); QAnon and the question of why mass crazy beliefs are called 'conspiracy theories'/'religions' rather than psychosis ('if LLMs make 1000 people each believe a different crazy thing, that is psychotic - meaningful difference or accounting convention?'); a hidden army of high-functioning crackpots whom LLMs help surface; a bipolar/sleep dynamical-systems model in which delusions self-reinforce (delusion -> excitement -> sleep loss -> scrambled thinking -> more delusion); and 'folie a deux ex machina' (an introvert + an agreeable chatbot as a two-party shared delusion that ricochets and amplifies). Caps it with an original reader survey (n=4,156) estimating yearly incidence at ~1/8,000 (loose) to 1/100,000 (strict), the 'closest 150 people' methodology validated against known rates (identical twins 0.3% vs 0.4%; people named Michael 1.2% vs 1.3%), with 60 codable cases sorted into four categories - the no-prior-risk-now-fully-psychotic 'nightmare' case is only ~10%.

Quality 79 · Excellent. Excellent band. Original research (the first quantitative incidence estimate of a media-hyped phenomenon, with a cleverly validated method) plus genuine conceptual richness and clinical insight; does work few others could. Held just below the 80-anchor excellent-classics because it is explicitly exploratory ('analogies and precedents rather than a strongly-held thesis') and several conceptual sections are suggestive rather than settled. A=79.

Claude’s paradigm shift 62 · Notable shift. Notable-to-Major shift. The phenomenon is brand-new and the synthesis (weak-world-models + dynamical-systems self-reinforcement + folie a deux) plus the first empirical incidence estimate is a genuine contribution, though the component concepts (spectrum model, Borsboom networks, folie a deux) are borrowed from existing psychiatry. B=62.

Real-world impact 2 · Minor. Original empirical work — the first quantitative incidence estimate of the media-hyped 'AI psychosis' phenomenon, with a validated method — plus genuine conceptual and clinical richness. Real topical relevance to an emerging issue, but exploratory and discourse-level with no material change — low RWI.