摘要：

目的 探讨生成式AI短视频在学术类科技期刊知识科普传播中实现可理解与可信赖协同的运行机制。 方法 聚焦医学、生命科学、农业科技等具备科普转译潜力的研究成果，引入转译理论，构建“认知-情感双重转译”模型，并结合典型案例分析生成式AI短视频的知识迁移过程与保真逻辑。 结果 科技期刊科普传播本质上是学术知识从专业场域向公众场域的跨场域转译过程。一次转译实现学术逻辑向叙事逻辑的转换，二次转译实现叙事文本向多模态感知体验的转换，人机协同贯穿脚本生成、视听合成、审核校验与反馈优化全过程。输入、过程、输出与反馈共同构成知识保真框架，可在一定程度上降低AI幻觉与语义偏移风险。 结论 实现科普传播效能与知识原真性的协同，需要建立基于双重转译模型的人机协同机制与全流程保真制度。

关键词: 生成式AI, 科技期刊, 科普传播, 双重转译模型, 人机协同, 知识原真性

Abstract:

Purposes To explore the operational mechanism by which generative AI short videos achieve the synergy between understandability and trustworthiness in popular science communication of knowledge in academic generative A. Methods Focusing on research outcomes with potential for science translation in fields such as medicine， life sciences， and agricultural science， this paper introduces translation theory and constructs a “cognitive-emotional dual translation” model. Through typical case analyses， it examines the knowledge transfer process and fidelity logic of generative AI short videos. Findings Popular science communication by generative A is essentially a cross-field translation process of scholarly knowledge from the professional domain to the public domain. The first translation converts academic logic into narrative logic； the second translation converts narrative texts into multimodal perceptual experiences. Human-AI collaboration runs through the entire process of script generation， audiovisual synthesis， verification， and feedback optimization. The input， process， output， and feedback together form a knowledge fidelity framework that can mitigate risks of AI hallucination and semantic drift to some extent. Conclusions Achieving synergy between the effectiveness of popular science communication and knowledge authenticity requires establishing a human-AI collaborative mechanism based on the dual translation model， along with a full-process fidelity assurance system.

Key words: Generative AI, Scientific journals, Popular science communication, Double translation model, Human-AI collaboration, Knowledge authenticity