arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2610.27573cs.AIcs.CLcs.CYcs.HCcs.LG
收藏

当大语言模型进行操纵时:测量行为、社会与时间效应

When LLMs Manipulate: Measuring Behavioral, Social, and Temporal Effects

Rensselaer Polytechnic Institute(伦斯勒理工学院);National University of Sciences and Technology(巴基斯坦国防科技大学);Virginia Tech(弗吉尼亚理工大学)

Sadia Asif, Israel Antonio Rosales Laguan, Shumaila Asif, Haris Khan, Muneeb Asif, Muhammad Zeeshan Karamat

AI总结:

该研究提出行为计算框架,结合MGD、MSIAN、THPM模型,经120人实验证实操纵性LLM输出可显著改变用户信任与顺从,且效应跨场景传播,为LLM安全评估提供量化风险的计算方法。

摘要:

大语言模型(LLMs)可通过微妙的语言线索影响用户信任与决策,但操纵性输出的行为效应仍缺乏量化。我们提出一种用于测量「行为脆弱性」的行为计算框架,其定义为用户对操纵性AI响应的易感性。基于DarkPatterns-LLM基准,我们使用操纵梯度下降(MGD)生成受控操纵刺激,针对权威、紧迫性与社会认同相关偏差。我们使用多利益相关方影响评估网络(MSIAN)分析参与者响应以建模跨用户影响,使用时间危害传播模型(THPM)表征行为效应随时间的持续性。在120名参与者的研究中,操纵性表述使信任校准与顺从产生显著变化(p<0.01),且效应在社会场景间传播。这些发现强调需将LLM安全评估从模型输出扩展至可测量的行为后果,并提供了量化操纵性AI交互产生的个体、社会与时间风险的计算框架。

英文摘要:

Large language models (LLMs) can influence user trust and decision-making through subtle linguistic cues, yet the behavioral effects of manipulative outputs remain poorly quantified. We introduce a behavioral computing framework for measuring \textit{behavioral vulnerability}, defined as users' susceptibility to manipulative AI responses. Building on DarkPatterns-LLM benchmark, we generate controlled manipulative stimuli using Manipulation Gradient Descent (MGD), targeting biases associated with authority, urgency, and social proof. We analyze participant responses using the Multi-Stakeholder Impact Assessment Network (MSIAN) to model cross-user influence and the Temporal Harm Propagation Model (THPM) to characterize the persistence of behavioral effects over time. In a study involving 120 participants, manipulative framing produced significant changes in trust calibration and compliance ($p < 0.01$), with effects also propagating across social contexts. These findings highlight the need to extend LLM safety evaluation beyond model outputs to measurable behavioral consequences and provide a computational framework for quantifying individual, social, and temporal risks arising from manipulative AI interactions.

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。

↑