摘要
[目的/意义]随着GAI技术的迅猛发展,信息生产与传播范式发生了结构性变革,信息污染问题随之凸显,主要表现为低质量内容的泛滥、虚假幻觉信息的持续传播等,严重危害信息生态环境,文章旨在揭示AIGC环境下信息污染的多维成因,并提出对应治理路径。[方法/过程]提出系统性治理框架,以信息污染的底层逻辑为切入点,深入探究AIGC环境下信息污染的多维成因,并构建数据层—模型层—传播层三阶污染链框架,系统剖析信息污染的生成逻辑与扩散路径。重点探讨数据筛选困境、大模型幻觉、用户认知偏差等核心问题。[结果/结论]提出对应的治理框架:通过训练数据净化与标准重构优化源头质量,依托算法修正与哨兵监督抑制模型幻觉,建立多模态检测机制、开展AI信息素养教育阻断污染传播。最后预想AIGC信息污染问题的发展方向,呼吁治理问题亟待解决。
[Purpose/significance]With the rapid development of GAI technology,the paradigms of information production and dissemination have undergone structural changes,and the problem of information pollution has become prominent,mainly mani⁃fest as the proliferation of low-quality content and the continuous spread of false and illusory information,seriously endangering the information ecological environment.This article aims to reveal the multi-dimensional causes of information pollution in the AIGC envi⁃ronment and propose corresponding governance paths.[Method/process]The research proposes a systematic governance framework,taking the underlying logic of information pollution as the entry point,deeply explores the multi-dimensional causes of information pollution in the AIGC environment and constructs a three-stage pollution chain framework of data layer-model layer-propagation layer,systematically analyzing the generation logic and diffusion path of information pollution.Among them,the core issues such as the dilemma of data screening,the illusion of large models,and user cognitive biases are mainly discussed.[Result/conclusion]Propose a corresponding governance framework:optimizing the source quality through training data purification and standard recon⁃struction,suppressing model illusions by relying on algorithm correction and sentinel supervision,blocking the spread of pollution by establishing a multimodal detection mechanism,and carrying out AI information literacy education.Finally,this paper antici⁃pates the development direction of the AIGC information pollution and emphasizes the urgent need for governance problem.
作者
储节旺
樊鑫鑫
Chu Jiewang;Fan Xinxin(School of Management,Anhui University,Anhui Hefei 230601)
出处
《情报理论与实践》
北大核心
2026年第2期38-46,共9页
Information Studies:Theory & Application
基金
国家社会科学基金一般项目“数智创新生态系统知识生成动力、扩散逻辑与治理机制研究”的成果,项目编号:23BTQ055。
