One world, one opinion? The superstar effect in LLM responses

TL;DR

This paper investigates the dominance of a few prominent figures in LLM responses across multiple languages, revealing a significant superstar effect that may limit diversity in global knowledge representation.

Contribution

It uncovers the extent of the superstar effect in LLMs across languages and highlights potential biases in global knowledge dissemination.

Findings

Low diversity in LLM responses across languages

A small number of figures dominate recognition

Potential risk of narrowing global knowledge

Abstract

As large language models (LLMs) are shaping the way information is shared and accessed online, their opinions have the potential to influence a wide audience. This study examines who the LLMs view as the most prominent figures across various fields, using prompts in ten different languages to explore the influence of linguistic diversity. Our findings reveal low diversity in responses, with a small number of figures dominating recognition across languages (also known as the "superstar effect"). These results highlight the risk of narrowing global knowledge representation when LLMs retrieve subjective information.