Strategic Algorithmic Monoculture: Experimental Evidence from Coordination Games

TL;DR

This study investigates how AI agents and humans coordinate in multi-agent environments, highlighting differences in maintaining diversity when incentives favor heterogeneity, with implications for strategic algorithmic monoculture.

Contribution

It introduces an experimental framework to distinguish primary and strategic algorithmic monoculture, revealing how LLMs and humans differ in sustaining diversity under incentives.

Findings

LLMs show high baseline similarity (primary monoculture).

Both humans and LLMs regulate similarity based on incentives (strategic monoculture).

LLMs coordinate well on similar actions but struggle to sustain heterogeneity when divergence is rewarded.

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 humans, they regulate it in response to coordination incentives (strategic monoculture). While LLMs coordinate extremely well on similar actions, they lag behind humans in sustaining heterogeneity when divergence is rewarded.