Social Informatics

In this paper, we propose the

Topical Communities and Personal Interest

(

TCPI

) model for simultaneously modeling topics, topical communities, and users’ topical interests in microblogging data.

TCPI

considers different topical communities while differentiating users’ personal topical interests from those of topical communities, and learning the dependence of each user on the affiliated communities to generate content. This makes

TCPI

different from existing models that either do not consider the existence of multiple topical communities, or do not differentiate between personal and community’s topical interests. Our experiments on two Twitter datasets show that

TCPI

can effectively mine the representative topics for each topical community. We also demonstrate that

TCPI

significantly outperforms other state-of-the-art topic models in the modeling tweet generation task.

Judgment aggregation investigates the problem of how to aggregate several individuals’ judgments on some logically connected propositions into a consistent collective judgment. The majority of work in judgment aggregation is devoted to studying impossibility results, but the relationship between the (social) dependencies that may exist between voters and the outcome of the voting process is traditionally not studied. In this paper, we use techniques from social network analysis to characterize the relations between the individuals participating in a judgment aggregation problem by analysing the similarity between their judgments in terms of social networks. We obtain a correspondence between a voting rule in judgment aggregation and a centrality measure from social network analysis and we motivate our claims by an empirical analysis. We also show how large social networks can be simplified by grouping individuals with the same voting behavior.

Managing friendship relationships in social media is challenging due to the growing number of people in online social networks (OSNs). To deal with this challenge, OSNs’ users may rely on manually grouping friends with personally meaningful labels. However, manual grouping can become burdensome when users have to create multiple groups for various purposes such as privacy control, selective sharing, and filtering of content. More recently, recommendation-based grouping tools such as Facebook smart lists have been proposed to address this concern. In these tools, users must verify every single friend suggestion. This can hinder users’ adoption when creating large content sharing groups. In this paper, we proposed an automated friend grouping tool that applies three clustering algorithms on a Facebook friendship network to create groups of friends. Our goal was to uncover which algorithms were better suited for social network groupings and how these algorithms could be integrated into a grouping interface. In a series of semi-structured interviews, we asked people to evaluate and modify the groupings created by each algorithm in our interface. We observed an overwhelming consensus among the participants in preferring this automated grouping approach to existing recommendation-based techniques such as Facebook smart lists. We also discovered that the automation created a significant bias in the final modified groups. Finally, we found that existing group scoring metrics do not translate well to OSN groupings–new metrics are needed. Based on these findings, we conclude with several design recommendations to improve automated friend grouping approaches in OSNs.

While direct social ties have been intensely studied in the context of computer-mediated social networks, indirect ties (e.g., friends of friends) have seen less attention. Yet in real life, we often rely on friends of our friends for recommendations (of doctors, schools, or babysitters), for introduction to a new job opportunity, and for many other occasional needs. In this work we empirically study the predictive power of indirect ties in two dynamic processes in social networks: new link formation and information diffusion. We not only verify the predictive power of indirect ties in new link formation but also show that this power is effective over longer social distance. Moreover, we show that the strength of an indirect tie positively correlates to the speed of forming a new link between the two end users of the indirect tie. Finally, we show that the strength of indirect ties can serve as a predictor for diffusion paths in social networks.

Knowing which users are likely to churn (i.e. leave) a service enables service providers to offer retention incentives for users to remain. To date, the prediction of churners has been largely performed through the examination of users’ social network features; in order to see how churners and non-churners differ. In this paper we examine the social and lexical development of churners and non-churners and find that they exhibit visibly different signals over time. We present a prediction model that mines such development signals using Gaussian Sequences in the form of a joint probability model; under the assumption that the values of churners’ and non-churners’ social and lexical signals are normally distributed at a given time point. The evaluation of our approach, and its different permutations, demonstrates that we achieve significantly better performance than state of the art baselines for two of the datasets that we tested the approach on.

Nowadays Open-Source Software is developed mostly by decentralized teams of developers cooperating on-line. GitHub portal is an online social network that supports development of software by virtual teams of programmers. Since there is no central mechanism that governs the process of team formation, it is interesting to investigate if there are any significant correlations between project quality and the characteristics of the team members. However, for such analysis to be possible, we need good metrics of a project quality. This paper develops two such metrics, first one reflecting project’s popularity, and the second one - the quality of support offered by team members to users. The first metric is based on the number of ‘stars’ a project is given by other GitHub members, the second is obtained using survival analysis techniques applied to issues reported on the project by its users. After developing the metrics we have gathered characteristics of several GitHub projects and analyzed their influence on the project quality using statistical regression techniques.

Building taxonomies for Web content manually is costly and timeconsuming. An alternative is to allow users to create folksonomies: collective social classifications. However, folksonomies have inconsistent structures and their use for searching and browsing is limited. Approaches have been proposed for acquiring implicit hierarchical structures from folksonomies, but these approaches suffer from the “generality-popularity” problem, in that they assume that popularity is a proxy for generality (that high level taxonomic terms will occur more often than low level ones). In this paper we test this assumption, and propose an improved approach (based on the Heymann-Benz algorithm) for tackling this problem by direction checking relations against a corpus of text. Our results show that popularity works as a proxy for generality in at most 77 of cases, but that this can be improved to 81% using our approach. This improvement will translate to higher quality tag hierarchy structures.

Social networking services (SNSs) such as Facebook and Google+ are indispensable social media for a variety of social communications, but we do not yet fully understand whether these currently popular social media will remain in the future. A number of studies have attempted to understand the mechanisms that keep social media thriving by using ameta-rewards game that is the dual formof a public goods game. However, the meta-rewards game does not take into account the unique characteristics of current SNSs. Hence, in this work we propose an SNS-norms game that is an extension of Axelrod’s metanorms game, similar to meta-rewards games, but that considers the cost of commenting on an article and who is most likely to respond to it. We then experimentally investigated the conditions for a cooperation-dominant situation in which many users continuing to post articles. Our results indicate that relatively large rewards compared to the cost of posting articles and comments are required, but optional responses with lower cost, such as “Like!” buttons, play an important role in cooperation dominance. This phenomenon is of interest because it is quite different from those shown in previous studies using meta-rewards games.

Article 1 of the United Nations Charter claims “human rights” and “fundamental freedoms” “without distinction as to [...] sex”. Yet in 1995 the Human Development Report came to the sobering conclusion that “in no society do women enjoy the same opportunities as men”. Today, gender disparities remain a global issue and addressing them is a top priority for organizations such as the United Nations Population Fund. To track progress in this matter and to observe the effect of new policies, the World Economic Forum annually publishes its Global Gender Gap Report. This report is based on a number of offline variables such as the ratio of female-to-male earned income or the percentage of women in executive office over the last 50 years.

In this paper, we use large amounts of network data from Google+ to study gender differences in 73 countries and to link online indicators of inequality to established offline indicators. We observe consistent global gender differences such as women having a higher fraction of reciprocated social links. Concerning the link to offline variables, we find that online inequality is strongly correlated to offline inequality, but that the directionality can be counter-intuitive. In particular, we observe women to have a higher online status, as defined by a variety of measures, compared to men in countries such as Pakistan or Egypt, which have one of the highest measured gender inequalities. Also surprisingly we find that countries with a larger fraction of within-gender social links, rather than across-gender, are countries with

less

gender inequality offline, going against an expectation of online gender segregation. On the other hand, looking at “differential assortativity”, we find that in countries with more offline gender inequality women have a stronger tendency for withing-gender linkage than men.

We believe our findings contribute to ongoing research on using online data for development and prove the feasibility of developing an automated system to keep track of changing gender inequality around the globe. Having access to the social network information also opens up possibilities of studying the connection between online gender segregration and quantified offline gender inequality.

Gender differences in human social and communication behavior have long been observed in various contexts. This study investigates such differences in the case of online social networking. We find a general tendency towards gender homophily, more marked for women, however users having a large circle of friends tend to have more connections with users of the opposite gender. We also inspect the temporal sequences of adding new friends and find that females are much more likely to connect with other females as their initial friends. Through studying triangle motifs broken down by gender we detect a marked tendency of users to gender segregation, i.e. to form single gender groups; this phenomenon is more accentuated for male users.

The boom of online social platforms of all kinds has triggered tremendous research interest in using social network data for user profiling, which refers to deriving labels for users that characterize their various aspects. Among different kinds of user profiling approaches, one line of work has taken advantage of the high level of label similarity that is often observed among users in one’s friendship network. In this work, we identify one critical point that has been so far neglected — different users in one’s friendship network play different roles in user profiling. In particular, we categorize all users in one’s friendship network into (I) close friends whom the user knows in real life and (II) online friends with whom the user forms connection through online interaction. We propose an algorithm that is affinity-aware in inferring users’ labels through network propagation. Our divide-and-conquer framework makes the proposed method scalable to large social network data. The experiment results in three real-world datasets demonstrate the superiority of our algorithm over baselines and support our argument for affinity-awareness in label profiling.

We explore the impact of technology on the strength of friendship ties. Data come from about two millions ties that members of CouchSurfing—an international hospitality organization whose goal is to promote travelling and friendship between its members—developed between 2003 and 2011 as well as original and secondary ethnographic data. The community, and the data available about its members, grew exponentially during our period of analysis, yet friendships between users tended to be stronger in the early years of CouchSurfing, when the online reputation system was still developing and the whole network was enmeshed in considerable uncertainty. We argue that this case illustrates a process of disenchantment created by technology, where technology increases the ease with which we form friendships around common cultural interests and, at the same time, diminishes the bonding power of these experiences.

Eye contact is presumed to be one of the most important non-verbal cues in human communication. It supports mutual understanding and builds the foundation for social interaction. In recent years, a variety of systems that support eye contact have been developed. However, research hardly focuses on investigating the impact of eye contact on social presence. In a study with 32 participants, we investigated the role of eye contact and gaze behavior with respect to social presence. Our results indicate that not only a system‘s capability to enable eye contact but also a user‘s consciously perceived eye contact are important to experience that the communication partner is ‘there’, i.e., social presence. Considering social presence as a source for social capital, i.e., valuable relationships that are characterized by trust and reciprocity, we discuss in what way social presence can serve as a contributing factor in video-mediated communication.

We use public data from Twitter to study the breakups of the romantic relationships of 661 couples. Couples are identified through profile references such as @user1 writing “@user2 is the best boyfriend ever!!”. Using this data set we find evidence for a number of existing hypotheses describing psychological processes including (i) pre-relationship closeness being indicative of post-relationship closeness, (ii) “stonewalling”, i.e., ignoring messages by a partner, being indicative of a pending breakup, and (iii) post-breakup depression. We also observe a previously undocumented phenomenon of “batch un-friending and being un-friended” where users who break up experience sudden drops of 15-20 followers and friends.

Our work shows that public Twitter data can be used to gain new insights into psychological processes surrounding relationship dissolutions, something that most people go through at least once in their lifetime.

The Name-Letter Effect states that people have a preference for brands, places, and even jobs that start with the same letter as their own first name. So Sam might like Snickers and live in Seattle. We use social network data from Twitter and Google+ to replicate this effect in a new environment. We find limited to no support for the Name-Letter Effect on social networks.We do, however, find a very robust Same-Name Effect where, say, Michaels would be more likely to link to other Michaels than Johns. This effect persists when accounting for gender, nationality, race, and age. The fundamentals behind these effects have implications beyond psychology as understanding how a positive self-image is transferred to other entities is important in domains ranging from studying homophily to personalized advertising and to link formation in social networks.

During sudden onset crisis events, the presence of spam, rumors and fake content on Twitter reduces the value of information contained on its messages (or “tweets”). A possible solution to this problem is to use machine learning to automatically evaluate the credibility of a tweet, i.e. whether a person would deem the tweet believable or trustworthy. This has been often framed and studied as a supervised classification problem in an off-line (post-hoc) setting.

In this paper, we present a semi-supervised ranking model for scoring tweets according to their credibility. This model is used in

TweetCred

, a real-time system that assigns a credibility score to tweets in a user’s timeline.

TweetCred

, available as a browser plug-in, was installed and used by 1,127 Twitter users within a span of three months. During this period, the credibility score for about 5.4 million tweets was computed, allowing us to evaluate

TweetCred

in terms of response time, effectiveness and usability. To the best of our knowledge, this is the first research work to develop a real-time system for credibility on Twitter, and to evaluate it on a user base of this size.

In this paper, we propose methods to estimate the credibility of reviewers as an individual and as a group, where the credibility is defined as the ability of precisely estimating the quality of items. Our proposed methods are built on two simple assumptions: 1) a reviewer who has reviewed many and diverse items has high credibility, and 2) a group of reviewers is credible if the group consists of many and diverse reviewers. To verify the two assumptions, we conducted experiments with a movie review dataset. The experimental results showed that the diversity of reviewed items and reviewers was effective to estimate the credibility of reviewers and reviewer groups, respectively. Therefore, yes, the diversity does improve the credibility of user review data.

Despite the enthusiastic rhetoric about the so called

collective intelligence

, conspiracy theories – e.g. global warming induced by chemtrails or the link between vaccines and autism – find on the Web a natural medium for their dissemination. Users preferentially consume information according to their system of beliefs and the strife within users of opposite worldviews (e.g., scientific and conspiracist) may result in heated debates. In this work we provide a genuine example of information consumption on a set of 1.2 million of Facebook Italian users. We show by means of a thorough quantitative analysis that information supporting different worldviews – i.e. scientific and conspiracist news – are consumed in a comparable way. Moreover, we measure the effect of 4709 evidently false information (satirical version of conspiracist stories) and 4502 debunking memes (information aiming at contrasting unsubstantiated rumors) on polarized users of conspiracy claims.

Nowadays, more and more people are using online news platforms as their main source of information about daily life events. Users of such platforms discuss around topics providing new insights and sometimes revealing hidden aspects about topics. The valuable information provided by users needs to be exploited to improve the accuracy of news recommendation and thus keep users always motivated to provide comments. However, exploiting user generated content is very challenging due its noisy nature. In this paper, we address this problem by proposing a novel news recommendation system that (1) enrich the profile of news article with user generated content, (2) deal with noisy contents by proposing a ranking model for users’ comments, and (3) propose a diversification model for comments to remove redundancies and provide a wide coverage of topic aspects. The results show that our approach outperforms baseline approaches achieving high accuracy.

Geographical processes might well impact online engagement in big countries like the USA. Upon a random sample of 200K news articles and corresponding 41M comments posted on the Yahoo! News in that country, we show that nearby individuals tend to comment and engage with similar news articles more than distant individuals do. Interestingly, at state level, topics one reads about are associated with specific socio-economic conditions and personality traits.

We perform a study on cold-start news popularity prediction using a collection of 13,319 news articles obtained from Yahoo News. We characterise the online popularity of news articles by two different metrics and try to predict them using machine learning techniques. Contrary to a prior work on the same topic, our findings indicate that predicting the news popularity at cold start is a difficult task and the previously published results may be superficial.

In this work, we reveal the structure of global news coverage of disasters and its determinants by using a large-scale news coverage dataset collected by the GDELT (Global Data on Events, Location, and Tone) project that monitors news media in over 100 languages from the whole world. Significant variables in our hierarchical (mixed-effect) regression model, such as population, political stability, damage, and more, are well aligned with a series of previous research. However, we find strong regionalism in news geography, highlighting the necessity of comprehensive datasets for the study of global news coverage.

The mass popularity of online social networks, such as Facebook and Twitter, makes them an interesting and important platform for exchange of ideas and opinions. Accurately capturing the opinions of users from their self-generated data is crucial for understanding these opinion flow processes. We propose a supervised model that uses a combination of hashtags and n-grams as features to identify the opinions of Twitter users on a topic, from their publicly available tweets. We use it to detect opinions on two current topics: U.S. Politics and Obamacare. Our approach requires no manual labeling of features, and is able to identify user opinion with a very high accuracy over a randomly chosen set of users tweeting on each topic.

Social media has changed the way people interact with each other and has contributed greatly towards bringing people together. It has become an ideal platform for people to share their opinions. However, due to the volatility of social networks, a negative campaign or a rumor can go viral resulting in severe impact to the community. In this paper, we aim to solve this problem of stemming the flow of a negative campaign in a network by observing only parts of the network. Given a negative campaign and information about the status of its spread through a few candidate nodes, our algorithm estimates the information flow in the network and based on this estimated flow, finds a set of nodes which would be instrumental in stemming the information flow. The proposed algorithm is tested on real-world networks and its effectiveness is compared against other existing works.

The rise in popularity of Twitter has led to a debate on its impact on public opinions. The optimists foresee an increase in online participation and democratization due to social media’s personal and interactive nature. Cyber-pessimists, on the other hand, explain how social media can lead to selective exposure and can be used as a disguise for those in power to disseminate biased information. To investigate this debate empirically, we evaluate Twitter as a public sphere using four metrics: equality, diversity, reciprocity and quality. Using these measurements, we analyze the communication patterns between individuals of different hierarchical levels and ideologies. We do this within the context of three diverse conflicts: Israel-Palestine, US Democrats-Republicans, and FC Barcelona-Real Madrid. In all cases, we collect data around a central pair of Twitter accounts representing the two main parties. Our results show in a quantitative manner that Twitter is not an ideal public sphere for democratic conversations and that hierarchical effects are part of the reason why it is not.

We apply an agent-based opinion dynamics model to investigate the distribution of opinions and the size of opinion clusters. We use parameter sweeps to examine the sensitivity of opinion distributions and cluster sizes relative to changes in individuals’ tolerance and uncertainty. Our results demonstrate that opinion distributions and cluster sizes are structurally unstable, not stationary, and have fat tails in most configurations of the model, rather than stable Gaussian distributions. Hence, extremist radical individuals occur far more frequently than “normally” expected. Opinion clusters, in addition to being fat-tailed, reveal a dynamic transition from lognormal to exponential distributions as parameters change.

The aim of this paper is to provide a framework and novel methodology geared towards mapping technological change in complex interdependent systems by using large amounts of unstructured data from various recent on- and offline sources. Combining techniques from the fields of natural language processing and network analysis, we are able to identify technological fields as overlapping communities of knowledge fragments. Over time persistence of these fragments allows to observe how these fields evolve into trajectories, which may change, split, merge and finally disappear. As empirical example we use the broad area of

Technological Singularity

, an umbrella term for different technologies ranging from neuroscience to machine learning and bioengineering, which are seen as main contributors to the development of artificial intelligence and human enhancement technologies. Using a socially enhanced search routine, we extract 1,398 documents for the years 2011-2013. Our analysis highlights the importance of generic interface that ease the recombination of technology to increase the pace of technological progress. While we can identify consistent technology fields in static document collections, more advanced ontology reconciliation is needed to be able to track a larger number of communities over time.

Researchers are actively turning to Twitter in an attempt to network with other researchers, and stay updated with respect to various scientific breakthroughs. Young and novice researchers have also found Twitter as a valuable source of information in terms of staying up-to-date with various developments in their field of research. In this paper, we present an approach to utilize this valuable information source within a topic modeling framework to suggest scientific articles of interest to novice researchers. The approach in addition to producing effective recommendations for scientific articles alleviates the cold-start problem and is a step towards elimination of the gap between Twitter and science.

Prior work in architectural and urban studies suggests that there is a strong correlation between people dynamics and crime activities in an urban environment. These studies have been conducted primarily using qualitative evaluation methods, and as such are limited in terms of the geographic area they cover, the number of respondents they reach out to, and the temporal frequency with which they can be repeated. As cities are rapidly growing and evolving complex entities, complementary approaches that afford social scientists the ability to evaluate urban crime theories at scale are required. In this paper, we propose a new method whereby we mine telecommunication data and open crime data to quantitatively observe these theories. More precisely, we analyse footfall counts as recorded by telecommunication data, and extract metrics that act as proxies of urban crime theories. Using correlation analysis between such proxies and crime activity derived from open crime data records, we can reveal to what extent different theories of urban crime hold, and where. We apply this approach to the metropolitan area of London, UK and find significant correlations between crime and metrics derived from theories by Jacobs (e.g., population diversity) and by Felson and Clarke (e.g., ratio of young people). We conclude the paper with a discussion of the implications of this work on social science research practices.

Online paedophile activity in social media has become a major concern in society as Internet access is easily available to a broader younger population. One common form of online child exploitation is

child grooming

, where adults and minors exchange sexual text and media via social media platforms. Such behaviour involves a number of stages performed by a predator (adult) with the final goal of approaching a victim (minor) in person. This paper presents a study of such online grooming stages from a machine learning perspective. We propose to characterise such stages by a series of features covering sentiment polarity, content, and psycho-linguistic and discourse patterns. Our experiments with online chatroom conversations show good results in automatically classifying chatlines into various grooming stages. Such a deeper understanding and tracking of predatory behaviour is vital for building robust systems for detecting grooming conversations and potential predators on social media.

Interest has been revived in the creation of a “bill of rights” for Internet users. This paper analyzes users’ rights into ten broad principles, as a basis for assessing what users regard as important and for comparing different multi-issue Internet policy proposals. Stability of the principles is demonstrated in an experimental survey, which also shows that freedoms of users to participate in the design and coding of platforms appear to be viewed as inessential relative to other rights. An analysis of users’ rights frameworks that have emerged over the past twenty years also shows that such proposals tend to leave out freedoms related to software platforms, as opposed to user data or public networks. Evaluating policy frameworks in a comparative analysis based on prior principles may help people to see what is missing and what is important as the future of the Internet continues to be debated.

Recent research on automatic analysis of social media data during disasters has given insight into how to provide valuable and timely information to formal response agencies—and members of the public—in these safety-critical situations. For the most part, this work has followed a bottom-up approach in which data are analyzed first, and the target audience’s needs are addressed later.

Here, we adopt a top-down approach in which the starting point are information needs. We focus on the aid agency tasked with coordinating humanitarian response within the United Nations: OCHA, the Office for the Coordination of Humanitarian Affairs. When disasters occur, OCHA must quickly make decisions based on the most complete picture of the situation they can obtain. They are responsible for organizing search and rescue operations, emergency food assistance, and similar tasks. Given that complete knowledge of any disaster event is not possible, they gather information from myriad available sources, including social media.

In this paper, we examine the rapid assessment procedures used by OCHA, and explain how they executed these procedures during the 2013 Typhoon Yolanda. In addition, we interview a small sample of OCHA employees, focusing on their uses and views of social media data. In addition, we show how state-of-the-art social media processing methods can be used to produce information in a format that takes into account what large international humanitarian organizations require to meet their constantly evolving needs.

This work investigates in-depth the communication practices within a workplace to understand whether workers interact face to face or more indirectly with email. We analysed the interactions to understand how these changes affect our work (productivity, deadlines, interesting task) and our wellbeing (positive and negative affective states),by using a variety of data collection methods (sensors and surveys). Our analysis revealed that overall email was the most frequent medium of communication, but when taking into account just the communication within working hours (8am to 7pm), that face to face interactions were preffered. Correlation analysis revealed significant relationships between Affective States and Situational Factors while Longitudinal Analysis revealed an impact of communication features and measures of self reported Productivity and Creativity. These findings lead us to believe that different communication processes (synchronous and asynchronous) can impact Positive and Negative Affective States as well as how productive and creative you feel at work.

In this paper, we outline an investigation of the impact of an office move on the social relationships of staff and students in a university research department. Combining the techniques of Social Network Analysis to assess for changes in social relations and Space Syntax Analysis for measuring the spatial changes, we identify key changes in the social relations that can be defined by spatiality. A decline in the social connections taking place and a change in the structure of the social network, accompanied by significant changes in spatial connectivity suggests that the office locations are influencing the underlying complex social processes.

Broadcast chat messages among team members in an organization can be used to evaluate team coordination and performance. Intuitively, a well-coordinated team should reflect the team hierarchy, which would indicate that team members assigned with particular roles are performing their jobs effectively. Existing approaches to identify hierarchy are limited to data from where graphs can be extracted easily. We contribute a novel approach that takes as input broadcast messages, extracts communication patterns—as well as semantic, communication, and social features—and outputs an organizational hierarchy. We evaluate our approach using a dataset of broadcast chat communications from a large-scale Army exercise for which ground truth is available. We further validate our approach on the Enron corpus of corporate email.

In this paper we leverage recent developments in the way scholars access, collect, and analyze data to reexamine consumption dynamics in popular music. Using web-based tools to construct a dataset that distills songs’ musical content into a handful of discrete attributes, we test whether and how these attributes affect a song’s position on the

Billboard Hot 100

charts. Our analysis suggests that attributes matter, beyond the effect of artist, label, and genre affiliation. We also find evidence that the relational patterns formed between attributes—what we call cultural networks—crowds songs that are too similar to their neighbors, adversely affecting their movement up the charts. These results suggest that culture possesses its own sphere of influence that is partially independent of the actors who produce and consume it.

We investigate trends in the international migration of professional workers by analyzing a dataset of millions of geolocated career histories provided by LinkedIn, the largest online platform for professionals. The new dataset confirms that the United States is, in absolute terms, the top destination for international migrants. However, we observe a decrease, from 2000 to 2012, in the percentage of professional migrants, worldwide, who have the United States as their country of destination. The pattern holds for persons with Bachelor’s, Master’s, and PhD degrees alike, and for individuals with degrees from highly-ranked worldwide universities. Our analysis also reveals the growth of Asia as a major professional migration destination during the past twelve years. Although we see a decline in the share of employment-based migrants going to the United States, our results show a recent rebound in the percentage of international students who choose the United States as their destination.

Religiosity is a powerful force shaping human societies, affecting domains as diverse as economic growth or the ability to cope with illness. As more religious leaders and organizations as well as believers start using social networking sites (e.g., Twitter, Facebook), online activities become important extensions to traditional religious rituals and practices. However, there has been lack of research on religiosity in online social networks. This paper takes a step toward the understanding of several important aspects of religiosity on Twitter, based on the analysis of more than 250k U.S. users who self-declared their religions/belief, including

Atheism

,

Buddhism

,

Christianity

,

Hinduism

,

Islam

, and

Judaism

. Specifically, (i) we examine the correlation of geographic distribution of religious people between Twitter and offline surveys. (ii) We analyze users’ tweets and networks to identify discriminative features of each religious group, and explore supervised methods to identify believers of different religions. (iii) We study the linkage preference of different religious groups, and observe a strong preference of Twitter users connecting to others sharing the same religion.

User behavior in online social media is not static, it evolves through the years. In Twitter, we have witnessed a maturation of its platform and its users due to endogenous and exogenous reasons. While the research using Twitter data has expanded rapidly, little work has studied the change/evolution in the Twitter ecosystem itself. In this paper, we use a taxonomy of the types of tweets posted by around 4M users during 10 weeks in 2011 and 2013. We classify users according to their tweeting behavior, and find 5 clusters for which we can associate a different dominant tweeting type. Furthermore, we observe the evolution of users across groups between 2011 and 2013 and find interesting insights such as the decrease in conversations and increase in URLs sharing. Our findings suggest that mature users evolve to adopt Twitter as a

news media

rather than a social network.