我国数字阅读APP个性化推荐算法透明度实证分析  

Empirical Analysis of Transparency of Personalized Recommendation Algorithm Used in Digital Reading APPs in China

作　　者：杨菲 唐凡舒 郑凯丽 Yang Fei;Tang Fanshu;Zheng Kaili

机构地区：[1]中国计量大学法学院

出　　处：《国家图书馆学刊》2024年第3期37-48,共12页Journal of The National Library of China

基　　金：国家社会科学基金青年项目“区块链视角下数字创意产业著作权保护及交易问题研究”(项目编号:19CFX063)阶段性研究成果。

摘　　要：个性化推荐算法生成的阅读建议便捷了人们的阅读生活,却也带来认知窄化等问题,因此向公众准确披露算法风险、保障算法透明愈发重要。文章以23款主流数字阅读APP为研究样本,从技术标准、用户感知和规范合规三个维度设计透明度检测要点,通过分析相关隐私政策文本及算法功能,评估算法透明度的履行差距。基于此,提出完善算法透明标准体系、推进数字阅读行业自律发展、增设“算法标签”、在算法设计中纳入透明度合规考量等对策。The algorithm-generated personalized reading recommendations have facilitated peoples reading life but also brought about issues such as cognitive narrowing.Therefore it is increasingly important to accurately disclose algorithm risks and ensure algorithm transparency to the public.At a practical level this article takes 23 mainstream digital reading apps as research samples and designs transparency detection points from three dimensions technical standards user perception and normative compliance to evaluate the performance gap of algorithm transparency through analysis of relevant privacy policy texts and algorithm functions.Finally countermeasures are put forward in terms of improving the algorithm transparency standard system addingalgorithm labels and incorporate transparency into the design considerations in algorithm development.

关 键 词：算法透明 个性化推荐 数字阅读 算法标签 

分 类 号：G250.7[文化科学—图书馆学] TP391.3[自动化与计算机技术—计算机应用技术]