AI幻觉挑战下循证科学传播的路径建设研究

A Study on Pathway Development for Evidence-Based Science Communication under the AI-Hallucination Challenge

中文关键词: AI幻觉 循证科学传播 知识生产 人工智能

英文关键词: AI hallucination  evidence-based science communication  knowledge production  Artificial Intelligence

中文摘要:

      循证科学传播作为以客观事实和系统性研究证据为基础的传播方式，在应对人工智能（AI）幻觉方面具有潜在优势。本研究以循证科学传播为切入点，分析AI与循证传播的双重关系。（1）AI对循证科学传播的赋能：AI在科学传播的数据汇聚、证据甄别和效果反馈等环节具有赋能潜力，可提升循证传播的效率与覆盖面；（2）AI幻觉对科学传播的挑战：AI幻觉对科学传播的挑战不仅体现在内容的真实性，更涉及知识生产过程的合法性，在内容根基、知识生产、知识稳定性与共识达成效率等维度均带来挑战，导致失实信息从个案失真扩展为结构性错误。基于此，本研究提出以证据性、可追溯性、适应性与公共性为原则的循证科学传播人机协同路径，明确每项原则下AI端干预机制与人机协同的关键节点，构建贯穿“内容生产—流程管控—节奏调控—效果反馈”的循证科学传播体系。

英文摘要:

      As an approach grounded in objective facts and systematic research evidence，evidence-based science communication holds potential advantages in mitigating Artificial Intelligence（AI）hallucinations. From the perspective of evidence-based science communication，this study analyzes the dual relationship between AI and evidence-based communication.（1）The empowering role of AI in evidence-based science communication：AI demonstrates enabling potential in stages such as data aggregation，evidence screening，and impact feedback，thereby enhancing the efficiency and reach of evidence-based communication.（2）The challenges posed by AI hallucinations to science communication：These challenges are reflected not only in content authenticity but also in the procedural legitimacy of knowledge production. They affect dimensions including content foundation，knowledge production，knowledge stability，and the efficiency of consensus-building，causing misinformation to escalate from isolated inaccuracies to structural errors. Based on these findings，this study proposes a human-AI collaboration pathway for evidence-based science communication，guided by the principles of evidentiality，traceability，adaptability，and publicness. By clarifying the AI-side intervention mechanisms and the key nodes of human-AI collaboration under each principle，this study constructs a comprehensive evidence-based science communication system spanning“content production—process control—pace regulation—impact feedback.”