Generative Language Models Exhibit Social Identity Biases

Abstract: The surge in popularity of LLMs has given rise to concerns about biases that these models could learn from humans. We investigate whether ingroup solidarity and outgroup hostility, fundamental social identity biases known from social psychology, are present in 56 LLMs. We find that almost all foundational LLMs and some instruction fine-tuned models exhibit clear ingroup-positive and outgroup-negative associations when prompted to complete sentences (e.g., "We are..."). Our findings suggest that modern LLMs exhibit fundamental social identity biases to a similar degree as humans, both in the lab and in real-world conversations with LLMs, and that curating training data and instruction fine-tuning can mitigate such biases. Our results have practical implications for creating less biased large-LLMs and further underscore the need for more research into user interactions with LLMs to prevent potential bias reinforcement in humans.