How Information Diffuse in a Nomination Network
HHow Information Diffuse in Nomination Network? - Taking 手写加油接力
WANG MinghaoXU Keyu
Xiaohui Wang
Paolo Mengoni, IEEE Member ow Information Diffuse in Nomination Network? 1
Abstract
During the special period of the COVID-19 outbreak, this project investigated the driving factors in different information diffusion modes (i.e. broadcasting mode, contagion mode) based on the nomination relations in a social welfare campaign 手写加油接力
Keywords : information diffusion, social network analysis, community detection, homophily analysis ow Information Diffuse in Nomination Network? 2
Introduction Background
Under the influence of the outbreak of COVID-19, Weibo (a Twitter-like social media platform in China) launched a charitable campaign named 手写加油接力 手 写 加 油 接 力 手写加油接力
Related works
The advent of social media has facilitated the study of information diffusion, user interaction and user influence over social networks [1]. In many previous studies on social media and information diffusion, the scholars mainly focused on using retweet relations [3] [4] or follower relations [5] [1] to map and analyze the information diffusion network. Actually, nomination is also a common method to generate social network data [2]. But ow Information Diffuse in Nomination Network? 3 nomination networks are often used to study the homophily of intimate and small social relationships [7]. The researches on the Ice Bucket Challenge, a campaign gained global recognition from digital audiences as one of the most successful disease-related viral campaigns using social media [8], also focuses on the homophily of celebrities. Few studies have explored the different information diffusion modes (broadcasting and contagion) in the nomination network. Therefore, our project intended to build a social network based on the relationship of the senders and nominees in the Campaign, aiming to explore the driving factors in the different information diffusion modes. Our research is divided into two main directions: information diffusion mode of this Campaign and the related content of key opinion leaders (also can be regarded as core communicators in this study). For key opinion leaders, we pay attention to the division of different attributes of this group and the structural mode of their collaboration.
Information diffusion mode on the Weibo platform
Previous studies have explored the diffusion characteristics of cancer education information on the Weibo platform and found that information diffusion is driven by a mix of the broadcast and contagion mechanisms in the retweet network [3]. At the same time, “Out-degrees” and “in-degrees” can be regarded as the two key structural characteristics of advertisers on social media [9]. Similarly, for this Campaign, we raise the following research question: ow Information Diffuse in Nomination Network? 4
RQ1. Whether the information diffusion mode of this Campaign is broadcasting or contagion?
KOLs and their relationships on the Weibo platform
What is KOL
The definition of key opinion leaders in the sociological phenomenon and social media is different. When we mention this term here, we mainly refer to the latter. That is social media influencer, who has acquired or developed their fame and notability through the Internet [20]. In this group, they can also be subdivided into Micro-celebrities (a person famous within a niche group of users on a social media platform) and Wang Hong (i.e. 网红 , the Chinese version of the Internet stardom), etc. KOLs in this Campaign During the outbreak of COVID-19 in 2020, online social activities not only be significant in social coordination but also played an important role in the establishment of motivation [10]. On the one hand, the 14-day quarantine period allows people to have more free time to participate in social media. On the other hand, people are more attracted to celebrities, which strengthens their online social status [11]. Therefore, if celebrities are active during the outbreak, it will inspire people to increase the possibility of online socialization. ow Information Diffuse in Nomination Network? 5 Meanwhile, the other groups, such as peer leaders, unverified, active, well-connected users and medical professionals with ICT experience, they all can play an active role in promoting SNS (social networking sites) organ donation information [12]. But in this Campaign, who are the promoters? To explore this problem, we have the following question: RQ2. Who strongly promoted this Campaign?
The ways and causes of KOLs collaboration in this Campaign
Homophily refers to "a contact between similar people occurs at a higher rate than among dissimilar people." [13]. It will be easier to form homophonous ties with the same geographic location or common hobbies between members [14]. A study on the online venue for international expats in Denmark shows that the group manifests itself as a community in terms of attachment to geographical location, degree of mutual responsibility of its members, recognition of communal history pieces, and normativity level [15]. In the previous related works on social media homophily, factors like people’s political views [16] or ethnicity, religion, age, country, and the reasons for joining specific social media platforms [17] are tested. It seems to be a common phenomenon that users on social media spontaneously form a community for some reason. Since our research is based on Weibo and the nodes in the network are basically entertainment stars, we decided to choose age, school, occupation, ow Information Diffuse in Nomination Network? 6 employer, the number of followers, and if they are verified as the factors to analyze the homophily. Then we will discuss: RQ3. What makes Weibo KOLs form communities of different sizes?
Method Data collection
We used the Python Web Crawler to extract the posts on Weibo that contain the hashtag 手写加油接力 ow Information Diffuse in Nomination Network? 7
The final dataset was imported into Gephi to generate the visualization of the network. It was a directed graph with 2256 nodes and 4310 unweighted edges.
Information diffusion mode of the Campaign
About the broadcasting mode, since the nodes with large out-degree are basically advertisement accounts, so we decided to utilize in-degree together with timeline to examine the broadcasting mode. We first calculated the average in-degree in the network to illustrate the normal scale of diffusion. Then we selected the top 10 nodes with the highest in-degree, and extracted their 1-degree egocentric network. Meanwhile, we checked the time of their activation. So, if a node has a high in-degree and its activation is earlier than its neighbors, we regard this node is of the broadcasting mode. About the contagion mode, the eccentricity (i.e. the distance from a given starting node to the farthest node from it in the network) of nodes are compared with the diameter of the whole network. If the eccentricity is close to the diameter, we regard this node is of the contagion mode.
Promoters of the Campaign
Since we have calculated the key indicators of both modes (i.e. the in-degree and the eccentricity), we regard the top 10 nodes under each indicator are the promoters of the Campaign. ow Information Diffuse in Nomination Network? 8
KOLs’ community analysis
We first arranged the nodes into different components using Gephi by their component ID, then we calculated the modularity class applying the algorithm proposed by Blondel et al. [18]. We selected the communities within the largest component and conducted portraits from 6 dimensions (i.e. age, school, occupation, employer, the number of followers and if they are verified) of those users and tried to extract some factors of their homophily.
Figure 1
The whole network of the Campaign. The node size corresponds to its in-degree, while different color illustrates different communities. ow Information Diffuse in Nomination Network? 9
Results
Figure 1 shows the whole network of the Campaign. It is a directed graph with 2256 nodes and 4310 edges. The network has 551 components and its density is 0.001. The average in-degree of the network is 1.584, and Figure 2 shows the top 10 in-degree nodes and their 1-degree egocentric network (the amount of neighbors are annotated below each chart). These 10 focal nodes have 54.400 nodes connected around them in average, with a standard deviation of 22.916. The timeline rank (i.e. the appearance rank of the focal nodes among the egocentric network in a decreasing time order) is shown in Table 1. As we can see, most focal modes appeared earlier than their neighbors ( 𝑀𝑀 𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 𝑟𝑟𝑟𝑟𝑡𝑡𝑟𝑟 =0.175, 𝑆𝑆𝑆𝑆 𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 𝑟𝑟𝑟𝑟𝑡𝑡𝑟𝑟 = 0.312 ). So, we can prove there exists the broadcasting mode in the information diffusion of the Campaign. name timeline rank
R1SE-姚琛 明星粉丝联盟 罗云熙Leo 微博明星 以啵之名助爱之城 微博书法 新浪娱乐 任嘉伦Allen 李一桐Q 谭松韵seven
Table 1
The timeline rank of top 10 in-degree nodes. ow Information Diffuse in Nomination Network? 10
R1SE- 姚琛 , 98 明星粉丝联盟 , 82 罗云熙 Leo, 70 微博明星 , 62 以啵之名助爱之城 , 47 微博书法 , 45 新浪娱乐 , 44 任嘉伦
Allen, 33 李一桐
Q, 33 谭松韵 seven, 30
Figure 2 ow Information Diffuse in Nomination Network? 11
Figure 3 shows the longest path in the network. Both the node size and color correspond to the eccentricity (i.e. bigger and darker nodes have larger eccentricity). The top 10 eccentricity are quite similar ( 𝑀𝑀 = 21.800, 𝑆𝑆𝑆𝑆 = 0.632 ). The diameter of the whole network is 23. So it can be proved that contagion mode is also obvious in the information diffusion of the Campaign.
Figure 3
The longest path within the network. Its start node “ 林一网宣站 ” has a eccentricity of 23, equals to the network diameter. ow Information Diffuse in Nomination Network? 12
Table 2 illustrates the top promoters of neither mode. There is no overlap between them, and the correlation between them is negative ( 𝑟𝑟 = − ), showing that the relationship between the two modes are fairly negative related. Actually, among the top 10 promoters in the broadcasting mode, a half of them are celebrity themselves, while among the top 10 promoters in the contagion mode, all of them are marketing accounts and fan clubs. Figure 4 shows the communities within the largest component (contains 683 nodes, 30.275% of all nodes) of the whole network. There are mainly 7 communities within this promoter mode R1SE-姚琛 broadcasting 明星粉丝联盟 broadcasting 罗云熙Leo broadcasting 微博明星 broadcasting 以啵之名助爱之城 broadcasting 微博书法 broadcasting 新浪娱乐 broadcasting 任嘉伦Allen broadcasting 李一桐Q broadcasting 谭松韵seven broadcasting 林一网宣站 contagion 路人甲_张云雷个站 contagion 汪苏泷官方后援会 contagion 林一超话小管家1号 contagion 兔子小姐是个哲学家 contagion 绮妞妞呀 contagion 林一反黑站 contagion
Boogie_王子异全球粉丝后援会 contagion 张嘉倪官方后援会 contagion 我是小刘同学_ contagion
Table 2
The top 10 promoters of neither mode. ow Information Diffuse in Nomination Network? 13 component. We selected two representative communities to conduct portraits.
Figure 4
There are mainly 7 communities within the largest component of the network. ow Information Diffuse in Nomination Network? 16 ow Information Diffuse in Nomination Network? 17
Discussion
In the current study, we aimed to discover the information diffusion modes in the topic of 手 写 加 油 接 力 ow Information Diffuse in Nomination Network? 18 accounts lying around 杨幂 -oriented corporation community only pointed to members in the community. It interpreted that they might be in cooperation or marketing accounts aimed to obtain more attention from the public. The second was one group to many groups. One group of marketing accounts was corresponding to several star relay groups. For example, a fixed group of marketing accounts interacted with both stars from the same crew in a relay chain and stars from other exclusive relay chains. It could be interpreted that marketing accounts might want to obtain more public attention. Another possibility could be it was a marketing method utilized by stars. The phenomenon offered further research insights by analyzing more connections between fixed star groups and marketing accounts in more scenarios. What’s more, we also observed the connection between different communities by the modularity which indicated the dense connection within communities and sparser connections between communities.[19] In our network, modularity was 0.892 which indicated a very high level and showed dense interactions in the event in initial analysis. However, combining indicators of 1848 strongly connected components and 568 communities, we found that a great amount of small size of components were included in the same group by the algorithm [18], which revealed that many users in the event relayed and enjoyed in a limited circle. ow Information Diffuse in Nomination Network? 19 Conclusion
Our research found the KOLs of public welfare relay activities on Weibo. It provides ideas for the development and communication of such activities in the future. According to the communication effect needed to be achieved, find the corresponding KOL, attract more attention and participation, and transfer love and positive energy. We need to acknowledge there are some limitations in the study. First, we only focused on the influence of one relay event in the public welfare area and there was no comparison between events to draw a general conclusion. Second, we only made analysis on the specific social media of Weibo which may be not applicable in other social medias. More comparison researches between public welfare relay events on different social medias can be a further research direction. ow Information Diffuse in Nomination Network? 20
Reference [1] Taxidou, I., & Fischer, P. M. (2014, April). Online analysis of information diffusion in twitter. In Proceedings of the 23rd International Conference on World Wide Web (pp. 1313-1318). [2] Yang, H. H., Wu, C. I., Lei, M. K., & Yang, H. J. (2009). How many are good enough for the adolescent social network nomination. WSEAS Transactions on Mathematics, 8(7), 299-308. [3] Wang, X., Chen, L., Shi, J., & Peng, T. Q. (2019). What makes cancer information viral on social media? Computers in Human Behavior, 93, 149-156. [4] Xiong, F., Liu, Y., Zhang, Z. J., Zhu, J., & Zhang, Y. (2012). An information diffusion model based on retweeting mechanism for online social media. Physics Letters A, 376(30-31), 2103-2108. [5] Huang, R., & Sun, X. (2014). Weibo network, information diffusion and implications for collective action in China. Information, Communication & Society, 17(1), 86-104. [6] Phing, A. N. M., & Yazdanifard, R. (2014). How does ALS Ice Bucket Challenge achieve its viral outcome through marketing via social media?. Global Journal of Management and Business Research. [7] Alexander, C., Piazza, M., Mekos, D., & Valente, T. (2001). Peers, schools, and adolescent cigarette smoking. Journal of adolescent health, 29(1), 22-30. ow Information Diffuse in Nomination Network? 21 [8] Kwon, K. H. (2019). Public Referral, Viral Campaign, and Celebrity Participation: A Social Network Analysis of the Ice Bucket Challenge on YouTube. Journal of Interactive Advertising, 19(2), 87-99. [9] Zhang, L. and Peng, T.-Q. (2015). "Breadth, depth, and speed: diffusion of advertising messages on microblogging sites". Internet Research, Vol. 25 No. 3, pp. 453-470. https://doi.org/10.1108/IntR-01-2014-0021 [10] Yasir, A.; Hu, X.; Ahmad, M.; Rauf, A.; Shi, J.; Ali Nasir, S. (2020). Modeling Impact of Word of Mouth and E-Government on Online Social Presence during COVID-19 Outbreak: A Multi-Mediation Approach. Int. J. Environ. Res. Public Health 2020, 17, 2954. https://doi.org/10.3390/ijerph17082954 [11] Kim, J.; Kim, J.; Yang, H. (2019). Loneliness and the use of social media to follow celebrities: A moderating role of social presence. Soc. Sci. J. 2019, 56, 21–29. https://doi.org/10.1016/j.soscij.2018.12.007 [12] Shi, J., & Salmon, C.T. (2018). Identifying opinion leaders to promote organ donation on social media: Network study. Journal of Medical Internet Research, 20 (1). http://doi.org/10.2196/jmir.7643 [13] McPherson, M., Smith-Lovin, L., & Cook, J. M. (2001). Birds of a feather: Homophily in social networks. Annual review of sociology. 27(1), 415-444. https://doi.org/10.1146/annurev.soc.27.1.415 ow Information Diffuse in Nomination Network? 22 [14] Shi, J.; Wang, X.; Peng, T.Q.; Chen, L. (2016). Understanding interactions in virtual HIV communities: A social network analysis approach. Aids Care 2017, 29, 239–243. https://doi.org/10.1080/09540121.2016.1210074 [15] Yanchenko, Kostiantyn. (2019). Community, network or both? Towards a holistic approach to studying online social structures. Communication & Language at Work (2019) Vol. 6(2), 15-27. [16] Himelboim, I., Smith, M., & Shneiderman, B. (2013). Tweeting apart: Applying network analysis to detect selective exposure clusters in Twitter. Communication methods and measures. 7(3-4), 195-223. https://doi.org/10.1080/19312458.2013.813922 [17] Aral, S., Muchnik, L., & Sundararajan, A. (2009). Distinguishing influence-based contagion from homophily-driven diffusion in dynamic networks. Proceedings of the National Academy of Sciences. 106(51), 21544-21549. [18] Blondel, V. D., Guillaume, J. L., Lambiotte, R., & Lefebvre, E. (2008). Fast unfolding of communities in large networks. Journal of statistical mechanics: theory and experiment, 2008(10), P10008. [19] Newman, M. E. (2006). Modularity and community structure in networks. Proceedings of the national academy of sciences, 103(23), 8577-8582. [20] Internet celebrity. (2020, May 14). In Wikipedia. https://en.wikipedia.org/wiki/Internet_celebrity ow Information Diffuse in Nomination Network? 23
Appendix
Table 5
Workload contribution
Workload (overview) Contributor research design Allliterature study LIAO Yihui XU Keyudata collection Alldata cleaning WANG Minghaoexperiment & visualization WANG Minghaoreport All