网络舆论的同音效应——基于选择螺旋和真基模拟的分析

Monotonization through Selective Spiral : Evidence from Analysis of Big Content by People-Instructed Machine (ABC-PIM) and Simulation on Real Data (SORD)

Research output: Journal Publications and ReviewsRGC 21 - Publication in refereed journalpeer-review

View graph of relations

Author(s)

  • 赵心树
  • 敖颂
  • 陆宏驰
  • 邢季成
  • 刘莉萍
  • 王子君
  • 廖圣清
  • 曾振华

Related Research Unit(s)

Detail(s)

Original languageChinese (Simplified)
Pages (from-to)23-57
Journal / Publication国际新闻界
Volume44
Issue number12
Publication statusPublished - Dec 2022

Abstract

本项研究报告一个此前未知的网络舆论现象:若无外来力量干涉或外来信息注入,且读者群体和转者群体的偏好维持稳定,则选择螺旋的数学本性使其趋向“同音”。即,经过选择发布、选择接收、选择转发、再接收、再转发,环生若干轮之后,接收和转发这两个群体之一所偏好的某类信息成为几乎唯一被转被读的信息,从而独霸特定信息场和舆论场,尽管另一群体和发布群体的偏好可能相异甚至相反于主导群体。本项研究分析的三个案例中,都是读者群体偏好的一类信息最终“霸场”,导致读者意愿主导的同音。发布群体和转发群体的偏好与读者群体的偏好相异甚至相反。发布群体偏好的内容甚至可能在螺旋早期占据信息场的大部空间。群体间的偏好相异或相反影响同音所需的螺旋轮数,从而影响同音速度,但未能改变读者群体主导、同音和霸场的趋向和结局。本项研究所依据的数据是2015年1月-2017年6月间近30个月中今日头条的帖文。在此期间,作者从该平台下载了全部可爬取的帖文标题与用户响应,从中选取了含有关键词“日本”的全部共47,224条标题和对应的5,237,308,573次点读数据及28,129,768次转发数据。经过“人指机大内容分析”,提取了三个情绪效价 (sentiment valence) 变量作为可能影响发布、点读和转发行为的选择因素。这三个效价变量分别测量标题内容对中国、对日本、或就事件性质而言的正负方向,形成本研究所基于的三个案例。
This study found that, selective spiral, due to its mathematical nature, tends to monotonize, a phenomenon previously undocumented in academic literature. Through selective release, selective reception, selective relay, and some repetitions of the subsequent reception-relay recycle, a certain type of information preferred by receivers or relayers will dominate, monopolize, and eventually monotonize, i.e., becoming virtually the only information relayed and read, even when the preferences of the other users differ or counter that of the dominant group.
The data for the study came from nearly five million postings and the related user responses crawled in almost 30 months from January 2015 to June 2017, from www.toutiao.com, which constitute all crawlable content and behavior data from the platform during the time period. From the dataset all postings with the Chinese keyword “ri ben,” i.e., Japan, were selected, giving us about 47,224 titles, 5,237,308,573 clicks to read, and 28,129,768 clicks to relay, aka share or forward, for further analysis. The titles were content-analyzed following the procedure of Analysis of Big Content by People-Instructed Machine (ABC-PIM). Three valence variables were coded, valence from the standpoint of China, i.e., positive, neutral, or negative for China, from the standpoint of Japan, or from the standpoint of journalism textbooks. The valence variables represented three selecting factors that may affect users’ behavior of release, read, and relay, making three cases for this study.
The direct observation of this study was on round zero, i.e., the initial round, of communication spiral, according to the theory of selective spiral. Relay is the product of round zero negation between releasers, receivers, and relayers, according to the theory of negotiated communication, derived from sociology of knowledge. Guided by the integrated theory of negotiated spiral, the study reports that, for each of the three cases, the releasers were found to be the most selective (s1=.79), thereby the most influential (ct=.82), in round zero. The readers, representing receivers (s1=.37), were far less selective than the releasers but far more selective than the relayers (s1=.02). Consequently, the receivers (ct=.17) were between the releasers and the relayers (ct<.01) in influencing the distribution of the relayed content at the conclusion of the round zero. The relayers were minimally selective (s1=.02), thereby minimally influential (ct<.01).
With valence defined from Chinese standpoint, the study also reports that readers favored the negative information the most (58.43%), positive information (28.93%) the second, and the neutral information (12.94%) the least. The pattern of selection is explained by an agitated and alarmed mentality toward Japan among the Chinese readers. By contrast, the released information was overwhelmingly neutral (86.00%), which competed with a small share of positive information (13.87%) and very little negative information (0.14%). The selective pattern of the releasers is explained by a tendency of the Chinese authors, editors, media managers and web regulators to calm and counter the alarm and agitation of the readers aided by the active but unselective relayers.
Devising a novel method, simulation on real data (SORD), the study also forecasts the process and product of the selective spiral, assuming that read per release and relay per read, measured in round zero to represent readers’ and relayers’ preferences respectively, would remain stable in the subsequent rounds. The results show that, in each of the three cases under study, it was readers’ preference that would dominate, monopolize, and eventually monotonize. The speeds of the monotonization, however, differed significantly. The negative info from Chinese standpoint needed over six rounds to climb to the 50% share of voice, and nine rounds to reach the 90% line. The counterpart yardsticks are respectively 26 rounds and 51 rounds for positive info from Japanese standpoint, and six and 11 rounds for negative events by journalism texts. The overwhelming neutral content released at the onset of the round zero appears to have decelerated, delayed and dissuaded the powerful monotonization of agitation and alarm.

Research Area(s)

  • selective spiral, monotonization, message negotiation, reception-dominated spiral, message involution, 选择螺旋, 同音化, 信息协调, 收者主导, 信息内卷

Citation Format(s)

网络舆论的同音效应——基于选择螺旋和真基模拟的分析. / 赵心树; 敖颂; 陆宏驰 et al.
In: 国际新闻界, Vol. 44, No. 12, 12.2022, p. 23-57.

Research output: Journal Publications and ReviewsRGC 21 - Publication in refereed journalpeer-review