AI is a master at fitting in, but it is mathematically incapable of being a 'non-conformist.'
April 15, 2026
Original Paper
Strategic Algorithmic Monoculture: Experimental Evidence from Coordination Games
arXiv · 2604.09502
The Takeaway
When humans coordinate, we naturally maintain a variety of different approaches. But LLMs (like ChatGPT) exhibit 'algorithmic monoculture'—the moment they are incentivized to work together, they all start acting exactly the same. They are incredibly good at 'averaging out' to match the crowd, but they fail to stay diverse even when being 'different' would be better. This means that as we use more AI in our social lives, we might be accidentally deleting the originality and weirdness that makes human society resilient. AI is the ultimate conformist.
From the abstract
AI agents increasingly operate in multi-agent environments where outcomes depend on coordination. We distinguish primary algorithmic monoculture -- baseline action similarity -- from strategic algorithmic monoculture, whereby agents adjust similarity in response to incentives. We implement a simple experimental design that cleanly separates these forces, and deploy it on human and large language model (LLM) subjects. LLMs exhibit high levels of baseline similarity (primary monoculture) and, like