Social Computing

As computer plays an important role in our daily life, it is becoming also important to understand social interaction within a computerized society. Social Computing is an area of research intersecting computer science and social science to understand and support any sort of social behavior in or through computational systems.

Patterns of Social Relationship between Twitter Users

Twitter Follower-Following Study

Follower-Following Study (2012)

Following activity in Twitter is the most fundamental action that Twitter users need to perform to receive information. People choose followed for various reasons including information sharing and social interaction, and we are interested in characterizing the features associated with the following activities. In this research, we investigate the following behaviors among Twitter users and build quantitative models that can explain the dynamics of the Twitter following. We first characterize the following behavior of Twitter users by conducting online survey. We then collect more than three million tweets from the survey participants to extract quantitative features, with which we build a series of Generalize Linear Models (GLM) on Twitter following. We analyze the models and the results show that Twitter users exhibit different interaction behavior based on the type of their followee. The result informs that classification algorithms can be utilized to build intelligent Twitter applications such as personalization, filtering summarization, and recommendation systems.

  • 서봉원, 이준환, 오종환 (2012). “트위터 사용자 간 관계형성 패턴에 대한 연구” 한국언론학보 58권 5호 (2012): 88-133. Top Paper Awarded.

Measurement and Analysis of Online Social Networks

Facebook Homophily Study

Facebook Homophily Study (2013)

While most of the research on homophily has been focused on offline contexts, current research aims to test whether the traditional concept of homophily is applicable to “online homophily” with further speculation by collecting users’ profile information and activity log data, in addition to users’ survey response data. Results show that empirical evidence on homophily in the online space is not found based on the original concept of similarities on personal attributes, but through interaction behaviors between users. Though the significance of correlations between interaction behaviors and perceived intimacy was not so strong, our findings imply that further elaboration on “online homophily” is necessary, in a way more relevant to the computer-mediated context other than the traditional concept of homophily.

  • Jieun Wee (2013). “Interaction behavior as an indicator of online homophily : A study on intimacy among Facebook friendship,” Master’s thesis. Seoul National University, Seoul, Republic of Korea. Top Thesis Award (Korean Association for Broadcasting & Telecommunication Studies, 2013)
  • Jieun Wee, Joonhwan Lee and Woncheol Jang (2014). “Measuring Homophily Effect: Interaction Behaviors as Predictors of Intimacy among Facebook Friendship”

Crowdsourcing Media Indexing: Analyzing online activity data


Crowdsourcing Media Indexing (2013)

As people spend more time online, watching YouTube or playing games, a number of research studies arose in ways to make use of the time and energy from the crowd in doing such activities. In this research, we have explored the possibility of converting the collective resources from the crowd in making useful information back to people. We collected posts from the online forums about soap operas on the air, and extracted instances when the name of characters in the play has been mentioned. These crowdsourced indexes become good search keywords to find the scenes where the characters mentioned in the posts appear.

  • Seyong Ha, Dongwhan Kim, and Joonhwan Lee (2013). “Crowdsourcing as a method for indexing digital media.” in CHI ’13 Extended Abstracts on Human Factors in Computing Systems (CHI EA ’13). ACM, New York, NY, USA, 931-936.

Open Source Collaboration in Github

Github Scan

Github Scan (2014)

Open Source Software (OSS)’s success is often attributed to the large numbers of contributors who test the software and report bugs. However, coordination is needed to develop software collaboratively, and so contribution can be costly. This limitation justifies elitism in OSS projects. It is an open question whether the success of OSS comes from the effects of a large number of people each making small contributions, or if it is driven by a core group of elite users who do most of the work. To answer this question, this study visualized contributions in Github projects and then conducted large data analysis of the participants’ behavior in OSS projects. This study focuses on the effects of contributions to the growth of as project. The results suggest that OSS is driven by a subset of contributors who produce most of the code. An increase in elite developers has a positive effect on development, while an increase in common contributors has a negative effect. Also, direct contributions by project managers, which do not go through public discussion, increase the effectiveness of the project. These results reveal the importance of the control of elite users in effectively exploiting the “many eyes” in OSS development.

  • Inho Won and Joonhwan Lee (2014). “The Wisdom of Crowds is not Enough: Influence of a Select Few in Github Projects”

Online Emotion Expression

Facebook Emotion Study

Facebook Emotion Study

On Korean websites, people often use several Korean vowels or consonants solely, which is the way to express emotion. This way of expression has scarcely been examined seriously, though it is one of the most popular online slangs. This is an exploratory study that reveals how and how much these expressions are used on social media. Data crawling technique was used as a way to examine the usage of vowels and consonants in ordinary situations. 6 ‘Facebook pages’ popular among Korean users are selected, and about 2.5 million comments are collected from posts in those pages. 10 expressions, including ‘ᄏ(K)’, ‘ᄒ(H)’, ‘ᅮ(U)’, are counted and analyzed from the collected texts of comments. As a result, several patterns of using vowels and consonants are found. Especially, ‘ᄏ(K)’ is used much more frequently than other vowels or consonants, and used more continually. It is also found that the patterns of using vowels or consonants in writing comments changes as the mood of posts or the characteristics of Facebook pages.

  • 오종환, 장수연, 이준환 (2014). “한글 자음 및 모음 사용을 통해 드러나는 온라인에서의 정서 표현에 대한 탐색적 연구” 한국멀티미디어연구 17권 7호 (2014): 866-878


The quality improvement has been an important issue in crowdsourcing management. Prior studies have tried to improve the quality by amount or scheme of payment, filtering results, or voting but the ways increased quality only in restricted conditions. In this paper, we suggested communication as a new way to manage quality. From prior studies, types of work-related communication were feedback and goal and each type consisted of 4 categories. Before starting the main experiment, prior study was performed to find out the effect of communication in crowdsourcing. When given message that third party would evaluate results, workers tended to spend more time for their work, but the quality was under expectation. The finding was that work-related communication was needed to be specific for better result. With this insight, the main experiment was conducted from May 11 to June 10, 2015. In the experiment, each participant was assigned to one of 8 communication styles and one of 2 payment schemes were set to find the effect of payment on quality management. Dependent variables were final answer’s length, work time, and cost. As results, negative feedback was more effective to improve the quality than positive feedback as it reduced the length, time, and cost but the positive one didn’t. Evaluative feedback was found to improve the quality better than descriptive feedback by reducing length and cost. In terms of goals, proximal & distal goal reduced the length and time more than distal goal, and the amount of reducing length was similar so proximal & distal goal was proved to be more effective for improving quality. Achievment goal was shown to improve quality better than target goal. We found that the most appropriate communication in crowdsourcing environment need to be negative and evlauative, the goal should be specific with proximal & distal goals, and context is required to improve the work quality. Negive feedback, evaluative feedback and proximal & distal goal were effective because they can help workers judge their results clearly. And workers can be engaged more when they understand the reason and context of works. Also from the results, the amount of payment affected the quality improvement as prior researches insisted but the effect was not absolute. Rather the effect was seen to be different by communication conditions. With these findings, this research found the possibility of communication as a solution for quality management in crowdsourcing.

  • Jae-Eun Lim and Joonhwan Lee (2014). “They are going to look at you: the effect of evaluation on crowdsourcing quality at Amazon Mechanical Turk”

Exploring the Usage of Social Networking Service by Depression and Emotional Stability

As social media become popular communication tools, the behavior on social media has been an interest of researchers. One of the topic regarding social media is to analyze the social media use by affective state, such as loneliness or depression. This research focuses on depression, one of the common negative affects, and investigates how communication behavior on social media is affected by depression. Facebook, which has been a popular social networking service in Korea, is selected among many other social media. Based on examining previous researches, two research questions are established. The first question is how the level of depression influences Korean users’ communication behavior on Facebook. The second question is if emotional stability, one of the personality traits, influences the relationship between depression and communication behavior on Facebook. To solve these questions, web application and crawling application are developed, using Facebook API(Application Programming Interface). Activity data of Korean Facebook users, who agreed to participate in this research, are collected with these applications. Participants also completed two types of survey, which measures depression and emotional stability, and those scores are also collected. In order to solve the first research question, Poisson regression is conducted between depression scores and each type of Facebook communication behavior. The result shows the frequency of broadcasting on one’s wall, given ‘Like’ on others’ posts or comments, and received ‘Like’ on one’s posts or comments becomes less with increasing depression score. The frequency of given comment and tag, received comment and tag, on the other hand, grows as depression score increases. Another Poisson regression is conducted to investigate the second research questions, which analyzes the interaction effect of depression and emotional stability. It is revealed that the interaction effect has negative coefficients through every types of communication behavior. This result suggests that one shows less Facebook activity when one’s emotional stabilty is lower than certain levels. This study collected and analyzed Korean Facebook users’ activity data, and can be differentiated from other Facebook studies which mainly focused on English users. The result of this study implies that further research for investigating Facebook activity data in relation to affective states is necessary and available.

Social Watching (VoteDial project)

Televised political debates among election candidates draw a lot of attention from the potential voters and television events often create a huge buzz on social media. Many people infer the general opinion climate on the timeline of social media while they are watching an event on TV. This study aim to examine the influence of opinions on social media on individuals’ cognitive process.