智能算法如何重构新闻价值?——基于智能推荐平台算法审计的研究

Reconstituting News Value in the Algorithmic Context:A Computational Audit of a Personalized Recommendation Platform

有关推荐算法重构新闻价值的讨论已在学术界兴起,但这种影响是否真实存在?如存在,有哪些影响?相关问题迄今鲜有实证研究。本研究采用计算实验范式中的虚拟代理测试法,对国内一款头部算法短视频平台展开算法审计,针对不同情境下新闻类视频的新闻价值展开测量与比较。研究发现,在不同的实验场景中,时新性、新颖性、接近性、显著性、重要性,以及情绪极性与强度等要素均产生显著的变化,这些价值要素在算法中介、用户偏好的作用下时强时弱,表现出弹性。结果表明,各类新闻价值要素在承受了技术、专业机构、用户等多方“推拉”后发生显著“位移”,这种“位移”展现出一定本土特色,部分价值要素的变动或将挑战甚至颠覆行业与受众的新闻观念。本研究为算法语境下的新闻价值研究提供了一套切实可行的新方法,也为观察当今哪些主体参与新闻价值之重塑、各新闻价值要素如何排列等问题提供了实证窗口。

The study explores the impact of recommender systems on news value,an area that has received attention but lacks empirical research.Guided by a reverse engineering approach,the audit focuses on a leading Chinese algorithmic short video platform as a sociotechnical news curation system.The research employs agent-based testing and computational communication methods to measure and compare classical news value elements like prominence and timeliness across different algorithmic contexts.Findings indicate a notable shift in news value elements within the recommender system where proximity,importance/influence,emotional polarity,and intensity encounter a tug-of-war from various forces,including automation technology,professional institutions,and platform users.These elements show flexibility and variation under different algorithmic settings.We argue that the observed shift in news value on Chinese algorithmic platforms has distinctive local characteristics,potentially challenging or overturning existing news perceptions within the Chinese journalism industry and among the general audience.Overall,the research contributes a viable methodological approach to studying news value in the algorithmic context,providing an empirical insight into how elements of contemporary news value are rearranged,and which entities are involved in reshaping news values.