Publications

人工智能对媒体行业未来就业的挑战是确有其事还是杞人忧天？基于未来就业研究视角，本研究建立机器学习模型并发现：（1）内容生产类媒体行业处于中低风险区间，覆盖17%的媒体从业者；（2）内容生产类媒体行业以认知型技能为主，现阶段侧重写作、理解与表达沟通，未来技能更加偏向决策、服务、咨询；（3）计算机化水平、教育和职业类型显著影响媒体行业的收入。本研究厘清了媒体行业对于未来就业的困扰：媒体行业短期内不会被自动化所替代，但是在中期（广播、广电、广告）和长期（主编、记者、评论员）存在较强的被替代的风险。

基于有限容量模型，论文认为社会加速与传播游戏之间的张力导致了移动阅读时间的碎片化，进而影响用户的阅读内容和行为模式。通过分析某移动阅读应用三个月的大规模用户行为数据，论文测量了阅读时长和阅读时点的碎片化，发现二者均显著提升阅读内容的多样性和阅读行为的持续性。论文的理论贡献在于阐明数字媒体使用具有社会加速和传播游戏的双重属性，强调通过重塑数字媒体中的社会互动和娱乐模式，以真正实现罗萨所说的现代主体与世界之间的“共鸣”。

信息扩散是一种普遍存在的传播现象，在人类生活中扮演着重要角色。社交媒体平台及其与公众之间的相互作用，对信息扩散的机制产生了深刻的影响，并带来很多新的、亟须解决的问题。本书聚焦于信息扩散研究领域存在一个重要困惑：与分析式模型的预测相反，实证研究发现大规模扩散罕见且脆弱。限制信息扩散的主要因素是什么？本书从格兰诺维特所提出门槛模型（1978）出发，系统地分析了不同社交媒体上的信息扩散及其瓶颈；在此基础上，逐渐走向注意力流动和计算叙事研究；最后，围绕拉图尔所提出来的“计算中心”概念对计算社会科学的逻辑进行讨论。

Yaotian Zhang, Mingming Feng, Ke-ke Shang, Yijun Ran, Cheng-Jun Wang * (2022) Peeking strategy for Online News Diffusion Prediction via Machine Learning. Accepted by Physica A. In Press.

[引用本文] 张伦、彭泰权、王成军、梁海、祝建华（2021）。〈从边陲到主流的一条自然路径：华人计算传播学者的参与和体验〉。李立峰、黄煜（编），《传承与创新：中华传播研究40年》（页399-419）。香港：香港中文大学出版社。 https://computational-communication.com/blog/2021/11/01/ccr-history/

Li, T., Cui, H. *, & Wang, C.J. * (2021) Breaking the Boundaries of Knowledge Space: Knowledge Spanning on the Q&A Website. Manuscript. Data and Code https://github.com/socrateslab/knowledge-spanning

Although popular users are relatively more influential on social media, there may exist a popularity fallacy in terms of their civic participation. We quantify the popularity, influence, and participation of 233 most influential verified users (big Vs) of Sina Weibo in 63 most salient public issues (2013-2016). We find that the popularity of big Vs has a negative influence on their participation in public issues. We explain this finding in the perspective of social identity and impression management. The results of this study shed light on our understanding of the most popular users on social media.

Social influence plays a crucial role in determining the size of information diffusion. Drawing on threshold models, we reformulate the nonlinear threshold hypothesis of social influence. We test the threshold hypothesis of social influence with a large dataset of information diffusion on social media. There exists a bell-shaped relationship between social influence and diffusion size. However, the large network threshold, limited diffusion depth and intense bursts become the bottlenecks that constrain the diffusion size. The practice of viral marketing needs innovative strategies to increase information novelty and reduce the excessive network threshold. In all, this research extends threshold models of social influence and underlines the nonlinear nature of social influence in information diffusion.

Using the word embedding techniques, we reveal the constructed emotional dependency of female characters on male characters in stories. We call this narrative structure “Cinderella complex,” which assumes that women depend on men in the pursuit of a happy, fulfilling life. Our analysis covers a substantial portion of narratives that shape the modern collective memory, including 7,226 books, 6,087 movie synopses, and 1,109 movie scripts. The “Cinderella complex” is observed to exist widely across periods and contexts, reminding how gender stereotypes are deeply rooted in our society.

Based on collaborative information filtering and information aggregation services, the social news website carries out a collective gatekeeping function, supplying an empirical ground to test the bandwagon effect on information diffusion. Modeling the interpersonal effect and bandwagon effect in collective gatekeeping with threshold models, we confirm the bandwagon effect of collective gatekeeping on news diffusion with two novel datasets.

本文以双重意见气候作为研究沉默的螺旋理论的框架，采用多主体建模的方法，将参考群体重新引入对沉默的螺旋的研究当中，重新考察了沉默的螺旋理论的边界条件，并分析了沉默的螺旋在不同的区域和时间中的差异。研究发现：在引进参考群体的情况下不能确保沉默的螺旋必然出现；沉默的螺旋效果是否能够稳定地出现要视大众媒体对其参考群体的影响力大小而定；沉默的螺旋效果的大小依赖于参考群体规模以及社区总人口规模，其演化受个体的空间分布的影响，而其增长率随着时间递减。

人类社会已经进入了计算社会科学时代，新的新闻生产方式和研究范式的确立给传统新闻学研究带来了挑战和机遇。计算社会科学从数据、计算方法、网络科学理论等多个层面丰富了新闻学研究。本文尝试梳理计算社会社会科学的发展脉络，总结代表性成果，分析其对新闻学研究的影响。本文最后讨论了计算传播学，强调了从研究问题、数据、模式、机制和基本原理的角度来理解新闻学研究。

Human attention becomes an increasingly scarce resource in the age of information explosion. To better understand the flow of collective attention, we construct the attention flow network using anonymous smartphone data of 100,000 users in a major city of China. In the constructed network, nodes are websites visited by users, and links denote the switch of users between two websites. We find a strong concentration of collective attention allocated to popular websites for smartphone users, and strong scaling relationships for the flow of collection attention in the flow networks. Especially, there is a centralized flow structure, the website of large traffic can control the circulated collective attention. Finally, we discussed the benefits and limitations of using the flow network analysis for computational communication research.

With the widespread use of mobile computing devices in contemporary society, our trajectories in the physical space and virtual world are increasingly closely connected. Using the anonymous smartphone data, we can construct the communication network, mobility network, and attention network, and further analyze the data in the perspective of network science.

互联网众筹公益项目的筹款过程是一个社会动员的过程。本文从资源动员理论的角度切入，利用从腾讯乐捐网站抓取的数据，考察了影响众筹公益项目的社会动员能力（筹款能力）的因素。本文研究发现：互联网众筹公益项目的筹款能力差异很大，筹款金额与排名的分布具有明显的长尾特征；项目自身的属性特点如项目类型、目标金额、项目发起人和地点均显著影响其社会动员能力；对社交媒体使用，能够大大地提高公益项目的资源动员能力，尤其是腾讯乐捐特有的“一起捐”机制，能够有效地发挥社会关系网络中强大的人际传播能力，以超线性增长的速度带来捐助人数的提升，从而大大提升项目的资源动员能力。

Contrary to popular belief that member turnover is harmful to organizational performance, evidence from millions of software development projects suggests that teams of a high turnover rate can also achieve success, if not more successful (Klug & Bagrow, 2016). Targeting on this novel observation, we argue that task allocation is important in understanding team success. Our analysis reveals two types of tasks, one of low uncertainty and requires the commitment of a small team, or a small group of core members in teams of considerable size, the other of high uncertainty and takes the collective wisdom from a large pool of team members. The “parallel path” effect, which means that the probability of a challenging sub-task being solved by a “walk-in” user increases with the pool size, ensures benefits offset the cost of maintaining a big team. The lessons from the studied big teams, who have been trying to adapt to and survive from the dynamic online environment, provide valuable insights into the management practices of teams in general.

比较人类在虚拟世界和现实空间的移动行为是科学研究的一个重要方面。本文的研究为完成这一个目标,构建并比较了注意力网络和移动网络。将人类的行为的数学结构表达为网络的形式,并采用网络科学的视角进行分析。通过对网络进行重整化,我们发现注意力网络是分形的,而移动网络是小世界的。Zhao 等(2014)的研究 认为人类在虚拟空间和现实世界的移动行为具有很强的相关;而我 们的研究则进一步发现这两类行为其实属于两类普遍的人类行为。更为重要的是空间约束不仅仅在物理空间发挥作用,在虚拟空间里同样发挥着重要作用,甚至更强。局部结构缺失空间约束使得移动网络表现出了小世界的特征并具有正的度相关;随着重整化的粗粒化过程空间约束开始发挥作用,度相关发生由正到负的变化。手机浏览的网站和应用之间具有明显的自相似的特点,使得移动网络空间约束更强,强制其嵌入几何网络的结构当中,并表现为很强的分形特点和与之相对应的负的度相关。通常我们认为虚拟空间解放了人的行为,因而可以表现出更多近乎随机的行为,而实际的发现则恰好与之相反。

The rich-club is a closely connected sub-community in networks that has important impacts on the online information diffusion. To investigate the rich-club effect in information diffusion on social media and its origins, we studied the mechanisms of information diffusion on the biggest microblog of China—Sina Weibo. The results demonstrated that there are obvious rich-club effects in both influential users and randomly sampled users; For the information diffusion within the rich-club, social selection, geographic proximity, and social influence were found to have significant influence, and the impact of social influence is overwhelmingly strong. The findings outline the landscape of information diffusion on Microblogs and the theoretical generalizations help us understand the hierarchical feature of rich-club phenomenon in information diffusion.

With the widespread use of mobile computing devices in contemporary society, our trajectories in the physical space and virtual world are increasingly closely connected. Using the anonymous smartphone data of $1 × 10^5$ users in a major city of China, we study the interplay between online and offline human behaviors by constructing the mobility network (offline) and the attention network (online). Using the network renormalization technique, we find that they belong to two different classes: the mobility network is small-world, whereas the attention network is fractal. We then divide the city into different areas based on the features of the mobility network discovered under renormalization. Interestingly, this spatial division manifests the location-based online behaviors, for example shopping, dating, and taxi-requesting. Finally, we offer a geometric network model to help us understand the relationship between small-world and fractal networks.

We suggest that the tree-like, stable structure of clickstream networks reveals the time-sensitive preference of users in online browsing. To test our assumption, we discuss three models on individual browsing behavior, and compare the simulation results with empirical data.

计算传播是指数据驱动的、借助于可计算方法所进行的传播过程,而分析计算传播现象的研究领域就是计算传播学。文章首先分析了计算传播的起源、概念和应用,然后从计算社会科学的角度对计算传播学的理论脉络进行了介绍。基于以上内容,文章介绍了一个计算传播学研究的例子新闻地图研究项目。最后讨论了计算传播学研究中存在的从数据到模式,再从模式到机制的研究思路。

本文探求根据大数据策划专题的可能性,目的是看大数据计算出来的受众爱好,是否能够为新闻IP变现建立受众需求的基础。王成军的这篇文章,以美剧《纸牌屋》作为对象来展现大数据运作的过程,这个流程本身对于新闻IP变现有着十分重要的借鉴意义。我们可以借文章了解计算传播过程(获取数据、挖掘数据、互联网实验、推荐系统、分析反馈)对集体智慧的挖掘和运用,并最终将此方法推广到新闻IP变现领域的实践当中。

The parent–child relationship normally experiences a significant change during the transition from adolescence to adulthood. However, there is much left to understand about how this transition affects and is affected by the communication between parents and emerging adults. A survey conducted among 490 Hong Kong university students examined their self-disclosure to their parents as an interpersonal process centered on perceived parental responsiveness and the role of separation–individuation in this self-disclosure process. The results support the idea that perceived parental responsiveness mediates the link between self-disclosure and relationship quality in the context of parent–child relationships. Dysfunctional independence predicts less self-disclosure, perceptions of less parental responsiveness, and poorer parent–child relationship quality. Significant gender differences were found on dysfunctional independence, self-disclosure, perceived parental responsiveness, and parent–child relationship quality. Young women with dysfunctional dependence disclosed less positive information, perhaps driven by an excessive need for attention and care.

Despite the burgeoning research on social entrepreneurship (SE), SE strategies remain poorly understood. Drawing on extant research on the social activism and social change, empowerment and SE models, we explore, classify and validate the strategies used by 2,334 social entrepreneurs affiliated with the world’s largest SE support organization, Ashoka. The results of the topic modeling of the social entrepreneurs’ strategy profiles reveal that they employed a total of 39 change-making strategies that vary across resources (material versus symbolic strategies), specificity (general versus specific strategies), and mode of participation (mass versus elite participation strategies); they also vary across fields of practice and time. Finally, we identify six meta-SE strategies―a reduction from the 39 strategies―and identify four new meta-SE strategies (i.e., system reform, physical capital development, evidence-based practices, and prototyping) that have been overlooked in prior SE research. Our findings extend and deepen the research into SE strategies and offer a comprehensive model of SE strategies that advances theory, practice and policy making.

对2015年出版的761部新闻传播学论著进行内容分析及评述，借助作者、商品简介、出版社、销量排行等结构化指标，可以窥探新闻传播学科的出版形势。中国新闻传播学界正在形成由科研院校为主导的学科论著出版格局，但政府干预和资本市场依然是主要出版动力，导致学术出版良莠不齐，为此应及时建立独立的学科论著评价机制和知识生产激励机制。

We use clicks as a proxy of collective attention and construct networks to study the temporal dynamics of attention. In particular we collect the browsing records of millions of users on 1000 Web forums in two months. In the constructed networks, nodes are threads and edges represent the switch of users between threads in an hour. The investigated network properties include the number of threads $N$, the number of users $UV$, and the number of clicks, $PV$. We find scaling functions $PV \sim UV^{\theta_1}$, $PV \sim N^{\theta_3}$, and $UV \sim N^{\theta_2}$, in which the scaling exponents are always greater than 1. This means that (1) the studied networks maintain a self-similar flow structure in time, i.e., large networks are simply the scale-up versions of small networks; and (2) large networks are more “productive”, in the sense that an average user would generate more clicks in the larger systems. We propose a revised version of Zipf’s law to quantify the time-invariant flow structure of attention networks and relate it to the observed scaling properties. We also demonstrate the applied consequences of our research: forum-classification based on scaling properties.

随着今日头条的崛起,聚合媒体开始进入公众的视野。这类媒体都是基于数据挖掘技术,筛选和推荐新闻。它为用户推荐有价值的、个性化的信息,提供连接人与信息的新型服务,是国内移动互联网领域成长最快的产品服务之一。自从2012年3月创建以来,今日头条至今已经累计激活用户3.1亿,日活跃用户超过3000万。本文尝试从技术层面分析今日头条的传播机制和相关原理。

计算社会科学为传播学发展提供了新的动力。通过文本挖掘和引文网络分析,本文梳理了计算社会科学的相关文献,证实了计算社会科学发展过程中存在的两大研究脉络:多主体建模和网络科学,而网络科学的发展和互联网大数据的广泛使用真正促进了计算社会科学的发展。基于此,本文尝试从可计算化的视角定义了计算传播学,强调了人类传播行为的可计算性基础和对于大数据背后的模式、机制及普适原理的挖掘。同时,本文简要讨论了传播学可计算化的工具、目标及应用。

本文旨在回顾和讨论新兴的计算社会科学在新闻传播研究中的应用。按照新闻传播研究中经典的5W模型,本文分别介绍了计算社会科学在谁(传播者),通过什么(渠道),对谁(受众),说了什么(内容),并产生了什么(效果)等五个领域的主要应用案例,并讨论了计算社会科学和网络大数据对这些研究领域的主要贡献和现存问题。

To evaluate the quality of public discussion about social movements on Twitter and to understand the structural features and evolution of longitudinal discussion networks, we analyze tweets about the Occupy Wall Street movement posted over the course of 16 days by investigating the relationship between inequality, emotion, and the stability of online discussion. The results reveal that (1) the discussion is highly unequal for both initiating discussions and receiving conversations; (2) the stability of the discussion is much higher for receivers than for initiators; (3) the inequality of online discussions moderates the stability of online discussions; and (4) on an individual level, there is no significant relationship between emotion and political discussion. The implications help evaluate the quality of public discussion, and to understand the relationship between online discussion and social movements.