夜雨聆风Lexical Traces of AI: Linguistic Impact of Generative Tools in Academic Abstracts
人工智能的词汇痕迹：生成式工具之于学术摘要的语言层面影响
Rubén Comas Forgas et al.
Abstract
The widespread availability of generative AI tools is reshaping scholarly discourse at a linguistic level. This study investigates the diffusion of AI-associated vocabulary in abstracts indexed in Scopus, focusing on post-2022 patterns that emerged after the public release of tools such as ChatGPT. Using a combined bibliometric and content-analysis approach, we identified 17 lexical items with unusually steep growth trajectories between 2022 and 2024. These terms were analysed across disciplines, countries, and journals as probabilistic markers of an AI-era linguistic shift rather than as diagnostic proof of direct AI use in individual articles. The results show a marked post-2022 lexical acceleration consistent with machine-mediated or AI-influenced writing environments, although the underlying causes are likely multifactorial and may include direct use, indirect imitation, editorial normalization, and broader stylistic drift. The gap between explicit AI disclosures and the observed diffusion of these markers is therefore interpreted as an illustrative reporting asymmetry rather than a definitive measure of undisclosed AI use. Overall, the findings point to a substantial transformation in academic expression and call for clearer disclosure taxonomies, stronger editorial guidance, and more nuanced debate about authorship, integrity, and linguistic heterogeneity in scholarly communication.
Abstract
生成式人工智能工具的普及正从语言层面重塑学术话语体系。本研究以爱思唯尔文献数据库（Scopus）收录论文摘要为研究对象，探究人工智能相关词汇的扩散特征，重点分析 ChatGPT 等工具面向公众开放后 2022 年以来形成的词汇使用规律。研究综合采用文献计量法与内容分析法，筛选出 2022—2024 年间使用频次增速异常显著的 17 个词汇特征项。本研究分学科、国家、期刊维度对上述词汇展开分析，将其视作人工智能时代语言变迁的概率性标识，而非判定单篇论文直接使用 AI 工具的决定性依据。研究结果显示，2022 年后相关词汇使用频次大幅攀升，印证当前写作环境普遍受到机器辅助或人工智能影响，但该现象背后成因具有多重性，既包括作者直接使用生成工具，也存在间接模仿、期刊编辑行文规范引导与整体学术文风演变等因素。据此，本研究认为，论文中明确标注 AI 使用情况的比例与上述词汇标识的广泛扩散程度存在明显落差，该现象仅能体现 AI 使用披露行为的信息不对称，无法作为衡量未报备 AI 使用行为的绝对标准。总体而言，研究证实学术文本表达已发生显著转变，呼吁学界建立更清晰的 AI 使用披露分类标准、强化期刊编辑规范指引，并围绕学术传播中的作者归属、学术诚信与语言多样性展开更细致深入的讨论。
以上翻译由人工智能完成，可能存在歧义或不正确之处，如有建议，请后台留言，谢谢。
其他详细内容请点击阅读原文
本文原载LEARNED PUBLISHING; 2026;39(3)
DOI：10.1002/leap.2067
版权归作者及LEARNED PUBLISHING所有
学术传播学报
http://ojs.inforvidor.org/jsc
