Original title: Large-Language Models as a Cognitive Virus
Ricard Solé, Giulio Ruffini, Francesca Castaldo, Marco Tuccio +5 · September 2026
As large-language models become part of daily life, they change how people create and share information. Researchers looked at this spread by comparing artificial intelligence use to a contagious virus moving through culture. They modeled people moving across three stages: not using the tools, using them temporarily, or becoming persistently dependent on them. The process spreads through social contact, recovery, and group habits reinforcing each other. The researchers found this mix can create tipping points. Past a critical threshold, tiny rises in adoption can suddenly lock whole populations into dependent use, causing sudden losses in thinking skills. However, the model shows people can build cognitive immunization by lowering how easily the habit spreads and making it simpler to step away.
Imagine a new dance craze spreading through a school. At first, it's optional, but soon everyone joins in to fit in, and those who don't feel left out. In this paper, AI tools spread similarly, where social pressure drives adoption. Once enough people use them, it becomes hard to avoid, leading to reliance and diminished original thinking, much like how the dance might replace other activities.
If AI tools spread like a virus, societies could face sudden shifts where people lose critical thinking skills. This affects everyone from students to professionals, altering how decisions are made independently.
The abstract presents a theoretical model without empirical evidence or real-world data validation. It does not specify which AI tools are studied or measure cognitive losses directly. The framework identifies potential conditions but does not establish actual outcomes in practice.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Large-Language Models as a Cognitive Virus — Ricard Solé, Giulio Ruffini, Francesca Castaldo et al.
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.