arXiv 2502.125662025-09-18cs.AI

探索人格特质对LLM偏见与毒性的影响

Exploring the Impact of Personality Traits on LLM Bias and Toxicity

University of Macau（澳门大学）；Shenzhen Key Laboratory for High Performance Data Mining, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences（深圳高性能数据挖掘重点实验室，深圳先进技术研究院，中国科学院）；Nanyang Technological University（南洋理工大学）；Department of Chinese Language and Literature, University of Macau（中文语言文学系，澳门大学）

Shuo Wang, Renhao Li, Xi Chen, Yulin Yuan, Derek F. Wong, Min Yang

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本研究基于HEXACO人格框架，通过实验发现为LLMs分配不同人格特质会一致地影响其输出的偏见、负面情感和毒性，并表明调整人格水平可作为一种简单低成本的内容安全控制方法。

摘要：

随着AI被期望在人类生活中扮演不同角色，为大型语言模型（LLMs）赋予不同人格已引起越来越多的研究关注。虽然这种“人格化”增强了LLMs的交互性和适应性的人类体验，但它也引发了关于内容安全的关键担忧，尤其是关于LLM生成的偏见、情感和毒性。本研究探讨了为LLMs分配不同人格特质如何影响其输出的毒性和偏见。利用社会心理学中广泛接受的HEXACO人格框架，我们设计了实验上合理的提示，以测试三个LLMs在三个毒性和偏见基准上的表现。研究结果表明，所有三个模型都对HEXACO人格特质表现出敏感性，更重要的是，它们的输出在偏见、负面情感和毒性方面呈现出一致的变化。特别是，调整若干人格特质的水平可以有效降低模型表现中的偏见和毒性，这与人类人格特质与毒性行为之间的相关性相似。这些发现强调了在考察LLM人格化的训练或微调方法效率之外，还需要检查内容安全。它们还表明，调整人格有可能成为一种简单且低成本的方法，用于进行受控文本生成。

英文摘要：

With the different roles that AI is expected to play in human life, imbuing large language models (LLMs) with different personalities has attracted increasing research interests. While the "personification" enhances human experiences of interactivity and adaptability of LLMs, it gives rise to critical concerns about content safety, particularly regarding bias, sentiment and toxicity of LLM generation. This study explores how assigning different personality traits to LLMs affects the toxicity and biases of their outputs. Leveraging the widely accepted HEXACO personality framework developed in social psychology, we design experimentally sound prompts to test three LLMs' performance on three toxic and bias benchmarks. The findings demonstrate the sensitivity of all three models to HEXACO personality traits and, more importantly, a consistent variation in the biases, negative sentiment and toxicity of their output. In particular, adjusting the levels of several personality traits can effectively reduce bias and toxicity in model performance, similar to humans' correlations between personality traits and toxic behaviors. The findings highlight the additional need to examine content safety besides the efficiency of training or fine-tuning methods for LLM personification. They also suggest a potential for the adjustment of personalities to be a simple and low-cost method to conduct controlled text generation.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。