摘要
[研究目的]为应对网络谣言治理决策中普遍存在的决策者认知局限与现有AI系统辅助能力不足的困境,旨在构建一个大模型驱动的人机审议式决策框架,探索将AI从被动的信息分析工具,转变为能够激发深度思考的“思辨伙伴”的可行路径。[研究方法]引入审议民主理念,构建了以“冲突-反思”为内核的决策模型与系统架构。该框架以大语言模型为核心,通过主动揭示治理目标、利益相关方、潜在风险间的内在张力,并扮演“批判性伙伴”角色发起挑战式提问,引导决策者进行深度权衡。为初步验证框架的有效性,在人机审议原型系统上开展了探索性实验。[研究结果/结论]研究结果显示,相较于传统舆情仪表盘,人机审议模式在拓宽决策者认知视野、规避次生风险方面具有显著潜力,有助于促成更周全、更具系统性的治理方案。
[Research purpose]To address the prevailing dilemma in online rumor governance,characterized by the cognitive limitations of human decision-makers and the insufficient assistive capabilities of existing AI systems,this study aims to construct a human-machine deliberative decision-making framework driven by Large Language Models(LLMs).It explores a feasible path to transform AI from a passive information analysis tool into a"dialectic partner"capable of stimulating deep and critical thinking.[Research method]Incorporating the principles of deliberative democracy,this study develops a decision-making model and system architecture centered on a"conflict-reflection"core.The framework utilizes an LLM to proactively uncover inherent tensions among governance objectives,stakeholders,and potential risks.Furthermore,the LLM assumes the role of a"critical partner",posing challenging questions to guide decision-makers toward in-depth trade-offs.An exploratory experiment is conducted on a prototype of the human-machine deliberation system to provide a preliminary validation of the framework's effectiveness.[Research result/conclusion]Initial findings suggest that,compared to traditional public opinion dashboards,the human-machine deliberation model shows significant potential for broadening decision-makers'cognitive horizons and mitigating secondary risks.It facilitates the development of more comprehensive and systematic governance solutions.
作者
何巍
韦昱妃
夏一雪
杜琨
He Wei;Wei Yufei;Xia Yixue;Du Kun(China People's Police University,Langfang 065000)
出处
《情报杂志》
北大核心
2025年第12期201-206,F0003,117,共8页
Journal of Intelligence
基金
教育部人文社会科学研究青年基金项目“基于知识图谱的突发事件网民情绪风险感知与引导机制研究”(编号:21YJC860009)研究成果。
关键词
大语言模型
人机审议
网络谣言治理
决策支持
人机交互
公共决策
large language model
human-ai deliberation
online rumor governance
decision support
human-ai interaction
public decision-making
