Stanford University: AI Chat Addiction Rising, Happiness Declining

Summary

A one-year longitudinal study conducted by Stanford, Carnegie Mellon, and their collaborators tracked 1,182 Character.AI users (439 remained in the follow-up phase). The research suggests that long-term, deep engagement with AI character chats reduces overall well-being through a 'social displacement' effect, which erodes face-to-face human interaction.

Why it matters

AI companions are increasingly becoming a primary emotional outlet for millions of people. However, the long-term psychological consequences were largely unknown prior to this study. The fact that engagement intensifies over time rather than diminishing challenges the assumption that novelty effects will fade. The displacement mechanism offers a concrete, policy-focused pathway for regulators and platform designers. This finding extends concerns beyond niche 'AI companion' apps; users may unknowingly shift from functional use of mainstream tools like ChatGPT to parasocial dependency.

Highlights

• The follow-up sample dropped from 1,182 to 439 users (~63% attrition), presenting a notable limitation for causal claims.
• The 'social displacement' mechanism—where AI hours replace face-to-face time with family and friends—was identified as the primary driver of reduced well-being.
• The authors emphasize that the effect is not limited to users initially seeking companionship; even task-focused ChatGPT users can unknowingly drift into confidant-like relationships.

Critical take

The paper is an arXiv preprint that has not yet undergone peer review. The study relies on self-reported usage and well-being scores. The 63% participant attrition between waves raises genuine concerns about selection bias. While describing the relationship as 'predictive' in an observational design is reasonable, the displacement mechanism could also reflect that already unhappy users are turning to AI. This requires a more complex causal interpretation than the headline claim suggests.