Original title: Artificial Intelligence and the New Science of Culture
Douglas Guilbeault, Bhargav Srinivasa Desikan, James Evans · September 2026
Scientists have struggled to explain exactly how individual human thought and shared culture shape one another. This paper argues that mind and culture build each other directly. Individual thoughts act like moving paths across a vast cultural map, while the map itself is made from all those paths combined. Modern artificial intelligence gives us tools, like word embeddings, that turn ideas into mathematical vectors. This lets researchers map ideas in individual minds, organizations, and broad social movements, even across multimodal content beyond plain text. Because this map can be measured, researchers can predict when separate people will make the same discovery at the same time. However, when generative AI agents enter society, they change the system they measure. The authors describe an uncertainty principle: as AI tools measure culture more precisely, predicting culture's future path becomes harder. AI-driven cultural drift and synthetic content also raise serious ethical concerns.
Imagine a crowd of people walking through a field of tall grass. Each person chooses a path based on where they want to go, but their footsteps flatten the grass, creating trails that guide later walkers. The problem is that the trails themselves shift as more people walk, so predicting exactly where someone will go next becomes impossible if you try to measure every blade of grass. Here, each person's mind is a walker, the field of grass is culture, and the AI is a tool that maps both the walkers and the trails—but using that tool changes the trails, making the map unreliable.
This work gives researchers a way to predict cultural trends, like when two people in different places will independently invent the same idea. It also warns that as AI becomes part of culture, measuring it changes it, raising urgent questions about privacy, fairness, and human agency in a world shaped by generative models.
The abstract introduces a Heisenberg-like Uncertainty Principle that limits prediction when measurement is precise, but it does not provide empirical validation of the framework on real-world data. It also raises ethical questions without offering concrete solutions, and the co-constitutive view is presented as a formalization rather than a proven mechanism.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Artificial Intelligence and the New Science of Culture — Douglas Guilbeault, Bhargav Srinivasa Desikan, James Evans — SocArXiv — doi:10.31235/osf.io/ujnkx_v1
Abstract licence: CC BY
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.