Ricard Solé Giulio Ruffini Michael Levin And Six Co-Authors arXiv 2609.03344 Submitted September 3 Physics And Society Three-State User Model: Uncoupled Coupled Persistently Dependent Tipping Points And Technological Lock-In Runaway Dynamics: Small Adoption Increase Triggers Abrupt Population-Level Shift Cognitive Competence Lost At Scale Cognitive Immunisation: Reduce Transmission Facilitate Reversibility
A paper submitted to arXiv on September 3, 2026, by Ricard Solé, Michael Levin, David Krakauer and six co-authors frames the spread of large-language model use not as a technology adoption story but as an epidemiological one and the model they construct suggests the consequences could be irreversible at the population level.
The authors propose a three-state user framework: uncoupled users who don’t use LLMs, coupled users who use them without dependency, and persistently dependent users who can no longer function without them. Using epidemiological and dynamical systems mathematics; the same tools used to model infectious disease and ecosystem collapse; they show that the interaction between social transmission, recovery and collective reinforcement can generate tipping points and technological lock-in.
The paper’s most provocative finding is the possibility of runaway dynamics: once a critical adoption threshold is crossed, small incremental increases in LLM use can trigger abrupt, population-scale transitions toward persistent dependence, accompanied by measurable losses in collective cognitive competence. This is not a gradual slide. It is a phase transition.
The framework is not entirely pessimistic. The same mathematics that describes the contagion also identifies conditions for what the authors call “cognitive immunisation”, reducing the transmission rate and maintaining pathways for reversibility so populations can recover cognitive independence. But the window for immunisation narrows as adoption deepens.
The most uncomfortable implication of the paper is not that AI is bad. It is that the system dynamics of mass AI adoption may be structurally identical to those of an epidemic and equally indifferent to individual intent.
We are building the very dependency the paper describes. The question it leaves open is whether we will notice the tipping point before we have already crossed it.
Check out our previous coverage of artificial intelligence on The Trusted Times.
