Work Paper

  • Online Dating

  1. Time: 2011 June-Now
  2. Intro:The figure above is one of the samples, which indicates how people interact with each other. Using ERG model, I focus on how individual attributes influence the interactions. I encounter one question of sampling from the huge population. It’s always a hard job of sampling from network. In this case, I randomly sample from the name list, and further investigate how these sampled people interact with each other. Here comes the problem, that is, I focus on the interactions between these sampled people, while their interaction with other people are not included.
  3. I am reanalyzing the data set to study how self-disclosure influence the messaging behavior within the online dating website, which is a much better indicator than clicking behavior.

Note: For the this network, triangle denotes female, while blue square denotes male.

  • Mobile Internet Use

  1. Time: 2011 Jan-2011 July
  2. Introduction: In this study,  I claim that fashion attentiveness moderate the relationship between demographic variables and mobile Internet use, and new forms of mobile technologies mediates the relationship between fashion attentiveness and mobile Internet use.

  • News diffusion on Digg

  1. Time: 2011 Feb-Now
  2. The purpose of this study is twofold: first, in perspective of threshold models, to propose a two-step flow model of news diffusion on social media, and second, using network data of 3883 news diffusions made by 139,409 distinct users on Digg’s social network (including 1,731,658 friendship links) to identify the influences of online friends and social media respectively. Individual’s opinion thresholds are calculated, and the relationship between adoption time, opinion threshold, and vector centrality are analyzed.
  3. The results suggest, on social media, friends have only slight influence on individuals’ adoption of news, and most adopters’ thresholds are 0, which implies most users are not influenced by their friends, and news jumps in the friend networks of Digg.com, individuals with high vector centrality tend to adopt news earlier. If the findings are any indications, epidemic models of social contagion (or information cascade) may overestimate online friends’ influence, and social media’s direct influence should be highlight.

  • Online Discussions on Micro-blog

  1. Time: 2010 Dec-Now
  2. Introduction: This is the mention/reply/follow network of twitter for the event of wikileaks. This is only a small scale data set, one third of the edges (see green color edges) are mention/reply relationship which indicated the collaborative conversation or discussion , and two third of the edges (see lime color edges) are the relationship of the following behavior between this 1051 people who tweets the term of ‘wikileaks’ at 11/12/2010.

Obviously, this data set demonstrate a fairly sparse mention/reply network (without the lime edges) with only 204 edges forming within 10 minutes (1:02-1:13, 11/12/2010). While all the green color lines which indicate the following relationship are collected 12 hours later.

Intuitively, the data set describe the interaction/conversation (green color edges) in social network (lime color edges). With the individual attributes (tweets, followers, followed, favorites, time zone) and the measure of social network structure, it’s possible to run ergm model and test how individual attributes give rise to global network.

Following this line of thought, I study how twitter users discuss occupying Wall Street along the time. The findings confirm that the online discussion of social movement on twitter is highly unequal, relatively stable with strong cyclical fluctuations, but there is no evidence that it’s influenced by extreme emotions. In addition to that, although the discussion initiators come and go along the time, the people who are spoken to are fairly stable. The findings implies that the online discussion is purposeful, aiming at the media and the elites as fixed target, thus twitter as social media effectively serves as the tool of social mobilization.

  • Agent-based Modeling of Spiral of Silence

  1. Time: 2010 July-Now
  2. Intro: This study is titled “The Emergence of Spiral of Silence from the Individual Behavior: Agent-based Modeling of Spiral of Silence”. By proposing an agent-based model of spiral of silence, Chengjun’s study shows that the stable existence of spiral of silence is contingent upon the relative strength of mass media over reference groups.

sos

convergence

  • Interconnected Pattern of WWW

  1. 2011 July-Now
  2. Intro: Based on the clickstream data of 1000 most popular websites,this study employs the ERG model to study how attributes of website influence the clickstream among the website.
  •  Video Diffusion Through YouTube

 

  1. How do the videos get popular on YouTube is a long-standing research question. This project aims in uncovering how different diffusion channels influence the popularity of videos (or the diffusion range) by controlling the influence of the content.
  2. The findings confirm the strong relationship among view, comment, favorite, rating, further although YouTube search engine, related videos, together with being featured in someone’s channel contribute to most of views for videos, the views from a subscriber most strongly predict the popularity of the video, thus the social influence on YouTube still matters. Similar to the news diffusion offline, the interpersonal  networks plays in a crucial role in spreading the the most popular information.

Burst leads to small popularity. See the video Charlie bit my finger – again below, you will find the linear growth of the number of views without big burst.