AI vs. Human Judgment of Content Moderation: LLM-as-a-Judge and Ethics-Based Response Refusals

作者
Stefan Pasch
出处
期刊:Cornell University - arXiv [Cornell University]
被引量:10
标识
DOI:10.48550/arxiv.2505.15365
摘要

As large language models (LLMs) are increasingly deployed in high-stakes settings, their ability to refuse ethically sensitive prompts-such as those involving hate speech or illegal activities-has become central to content moderation and responsible AI practices. While refusal responses can be viewed as evidence of ethical alignment and safety-conscious behavior, recent research suggests that users may perceive them negatively. At the same time, automated assessments of model outputs are playing a growing role in both evaluation and training. In particular, LLM-as-a-Judge frameworks-in which one model is used to evaluate the output of another-are now widely adopted to guide benchmarking and fine-tuning. This paper examines whether such model-based evaluators assess refusal responses differently than human users. Drawing on data from Chatbot Arena and judgments from two AI judges (GPT-4o and Llama 3 70B), we compare how different types of refusals are rated. We distinguish ethical refusals, which explicitly cite safety or normative concerns (e.g., "I can't help with that because it may be harmful"), and technical refusals, which reflect system limitations (e.g., "I can't answer because I lack real-time data"). We find that LLM-as-a-Judge systems evaluate ethical refusals significantly more favorably than human users, a divergence not observed for technical refusals. We refer to this divergence as a moderation bias-a systematic tendency for model-based evaluators to reward refusal behaviors more than human users do. This raises broader questions about transparency, value alignment, and the normative assumptions embedded in automated evaluation systems.

科研通智能强力驱动
Strongly Powered by AbleSci AI
科研通是完全免费的文献互助平台,具备全网最快的应助速度,最高的求助完成率。 对每一个文献求助,科研通都将尽心尽力,给求助人一个满意的交代。
实时播报
peir完成签到,获得积分10
1秒前
科研努力版完成签到 ,获得积分10
2秒前
风吹麦浪发布了新的文献求助10
2秒前
3秒前
小二郎的应助被嗷嗷ao采纳,获得10
3秒前
欧阳峰发布了新的文献求助10
3秒前
宋宋完成签到 ,获得积分10
5秒前
5秒前
6秒前
水煮电吹风的应助被剪刀手采纳,获得10
6秒前
大碗完成签到,获得积分10
7秒前
8秒前
CAOHOU的应助被吱吱采纳,获得10
8秒前
欧阳峰完成签到,获得积分10
8秒前
9秒前
9秒前
1618完成签到 ,获得积分10
9秒前
Ning_完成签到,获得积分10
9秒前
一只鱼完成签到 ,获得积分10
9秒前
10秒前
zivi完成签到,获得积分20
10秒前
易宋发布了新的文献求助10
10秒前
Owen的应助被惠1采纳,获得10
10秒前
meimei完成签到 ,获得积分10
11秒前
和道一文字完成签到,获得积分10
12秒前
kiiso发布了新的文献求助10
14秒前
14秒前
江枫渔火完成签到 ,获得积分0
15秒前
tang发布了新的文献求助10
15秒前
Juvenilesy的应助被墨尘采纳,获得10
15秒前
pmh完成签到,获得积分10
17秒前
和谐完成签到 ,获得积分10
17秒前
D_DAY完成签到,获得积分10
18秒前
19秒前
19秒前
19秒前
hudirudi完成签到,获得积分10
21秒前
神经蛙完成签到,获得积分10
21秒前
学术小白完成签到,获得积分10
21秒前
22秒前
高分求助中
(应助此贴封号)【重要!!请各用户(尤其是新用户)详细阅读】【科研通的精品贴汇总】 10000
拉伸高分子熔体中缠结行为的模拟和小角中子散射研究 1700
The World Atlas of Wine 9th edition 1000
奈特基础组织学:配套组织病理学(第4版)(Netter's Essential Histology With Correlated Histopathology) 800
八旗制度の研究 600
Boundaries and Bodhisattvas 500
高分子基功能界面水蒸发材料与应用研究 500
热门求助领域 (近24小时)
化学 材料科学 医学 生物 纳米技术 计算机科学 工程类 有机化学 化学工程 生物化学 物理 复合材料 内科学 催化作用 光电子学 人工智能 细胞生物学 无机化学 物理化学 数学
热门帖子
关注 科研通微信公众号,转发送积分 8181839
求助须知:如何正确求助?哪些是违规求助? 9571584
关于积分的说明 21818707
捐赠科研通 7778124
什么是DOI,文献DOI怎么找? 3433524
关于科研通互助平台的介绍 2557710
邀请新用户注册赠送积分活动 2479621