FAIRGAMER: Evaluating Social Biases in LLM-Based Video Game NPCs

Bingkang Shi, Jen-tse Huang, Long Luo, Tianyu Zong, Hongzhu Yi +4 more
8/25/2025
cs.AI

Official code: Anonymous999-xxx/FairGamer (1 star, GPL-3.0, last commit 2025-12-29).

Abstract

Large Language Models (LLMs) have increasingly enhanced or replaced traditional Non-Player Characters (NPCs) in video games. However, these LLM-based NPCs inherit underlying social biases (e.g., race or class), posing fairness risks during in-game interactions. To address the limited exploration of this issue, we introduce FairGamer, the first benchmark to evaluate social biases across three interaction patterns: transaction, cooperation, and competition. FairGamer assesses four bias types, including class, race, age, and nationality, across 12 distinct evaluation tasks using a novel metric, FairMCV. Our evaluation of seven frontier LLMs reveals that: (1) models exhibit biased decision-making, with Grok-4-Fast demonstrating the highest bias (average FairMCV = 76.9%); and (2) larger LLMs display more severe social biases, suggesting that increased model capacity inadvertently amplifies these biases. We release FairGamer at https://github.com/Anonymous999-xxx/FairGamer to facilitate future research on NPC fairness.

View on arXivView PDF

Code Implementations(1)

Anonymous999-xxx/FairGamerOfficial100%
GPL-3.0

FairGame: First benchmark exposing LLMs' decision biases in real-game NPC scenarios, revealing critical balance-breaking effects across languages

10PythonMay 10, 20259 months agoGPL-3.0

Similar Papers

Benchmarking Bias in Large Language Models during Role-Playing

85% match2024

BBQ-V: Benchmarking Visual Stereotype Bias in Large Multimodal Models

83% match2025

Prompting Fairness: Artificial Intelligence as Game Players

82% match1 repo22024

Biased Heritage: How Datasets Shape Models in Facial Expression Recognition

82% match1 repo12025

VLBiasBench: A Comprehensive Benchmark for Evaluating Bias in Large Vision-Language Model

82% match1 repo162024

Breaking the Benchmark: Revealing LLM Bias via Minimal Contextual Augmentation

82% match2025

Cite this paper

@article{shi2025fairgamer,
  title  = {FAIRGAMER: Evaluating Social Biases in LLM-Based Video Game NPCs},
  author = {Bingkang Shi and Jen-tse Huang and Long Luo and Tianyu Zong and Hongzhu Yi and Yuanxiang Wang and Songlin Hu and Xiaodan Zhang and Zhongjiang Yao},
  year   = {2025},
  eprint = {2508.17825},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2508.17825v3}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2508.17825.svg)](https://codeofpaper.com/paper/2508.17825)

Discussion