英文外文文献-分析wallop博客系统中社会关系与持续参与之间的关系（可编辑）（可编辑）

外文文献-分析wallop博客系统中社会关系与持续参与

之间的关系

附录英文资料翻译

英文原文

The Ties that Blog: Examining the Relationship BetweenSocial Ties and Continued Participation in the WallopWeblogging System

1 INTRODUCTION

Why do people maintain weblogs? Although existingresearch provides excellent insight into individualmotivations for creating weblogs, such work rarely focuses onchanges in behavior over time. Work on time-series data tendsto focus on linking patterns between weblogs and the flow ofinformation through these link structures. However, thequestion of why some people stop updating their weblogs aftera short period of time while others remain active for monthsor even years is not well researched. Is this a phenomenondependent upon social factors, the behavior of others as muchas individual qualities? Do people who receive responses andinteract with others remain active for longer than those who

do not?Or is it more personal, where some individuals are moreinternally motivated to continue posting content than others?Comparing social tie patterns with activity over time cananswer this question Technical developments and the growthof the user base have made weblogs a common medium for socialinteraction, and understanding this increasingly importantelement of network usage is critical to effective analysis ofthe medium. Recognizing the importance of social interaction,recent research ranging from studies of political opinions toanalyses of the conversational nature of weblogs has focusedon links and comments as indicators of interaction betweenusers. However, such research is necessarily limited toanalysis of individuals who actively post content, andgeneralizing these results to the full user population isdifficult since the bulk of the content produced in Internetdiscussion groups and some weblogging systems is contributedby a very small percentage of the user population. A systematicstudy of the factors contributing to continued user activitywill provide an important backdrop to future research on socialbehavior in weblogs, and could allow researchers to examinequestions which were previously intractable due to thedifficulty of generalizing results to a broader population.

Furthermore, understanding the factors which contribute tocontinued user activity could allow the creators of newweblogging systems to make more informed design decisionsThis study draws on rich longitudinal data taken from Wallop,a personal publishing and social networking system designed byMicrosoft Research, to compare users who remain active withthose who do not. The analysis focuses on the relationshipbetween social ties and continuedactivityas expressed throughvarious features of the system, and tests if people who remainactive are more socially connected to other users within thesystem. An additional dimension of the Wallop user base addsa cultural element to the analysis. Over the life of the Wallopsystem 65%of the active users were Chinese language users who,as we have shown elsewhere, contribute more content and remainactive in the system longer than non-Chinese language users.This suggests some difference between the two groups, whichcontributes to differences in their levels of continuedactivity in the system. Is this difference cultural? Is itsocial? How much of the variance between the two groups isexplained by differences in their respective social networks

2WEBLOGS AND SOCIAL NETWORKS

As the popularity of weblogs has increased, interaction

between authors has become more important to users, developers,and researchers. The popularity of services targeted atpromoting user interaction, such as LiveJournal, has led to thecreation of alternative weblogging systems which emphasizesharing content with others. Weblogging features have beenadded to popular social networking sites like MySpace andFriendster, and photo sharing tools like Flickr are increasingin popularity and beginning to draw competition from similarproducts. Meanwhile, researchers have been busily analyzingthe link structure taken fromcollections of weblogs, searchingfor clues to the social relationships between weblog authors.Analyses of these network structures have led to excellentstudies of deliberation within the medium, and the creation ofmetrics for authority and influence. Another study of thesocial structure of weblogs indicates that in some casesweblogpostings are conversational in nature, which may suggest astrong social connection between the participants Researchexamining the nature of social links contained within theweblog medium is rarely focused on explaining individualparticipation. Yet understanding why people upload content totheir weblogs, and why some continue to publish content monthsor even years later, may be a central element needed for

understanding the nature of the social phenomenapresent withinthe weblogging medium. Existing research on individualmotivations, based on content analysis and interviews,suggests that people publishweblogs formany different reasons,including a desire to publish a diary, share knowledge, expressa viewpoint, or become part of a community based on a particularshared interest. However, this research only touches on thesocial connections between the users and how they affectindividual decisions to participate Outside of the weblogliterature, research suggests that social ties are criticallyimportant for participation. Recruitment to political andsocial movements is oftendependent onpre-existing social ties.In e-learning settings, continued participation is dependenton social interaction with the professors and other students,even more so than the course contents and materials. Studiesof data taken from other online groups show that closelyconnected groups are more supportive of members, andindividuals with a strong sense of attachment to a group aremore likely to participate Despite the apparently strongconnection between social connections and participation inonline and offline groups, empirical research on therelationship between social network structure and individual

activity is limited. This lackof empirical research is largelya result of the difficulty of collecting network data. Even inthe case of weblog research, where data on the links connectingone weblog to another is both archived and publicly available,the sheer number of weblogs, the variety of different links,and the tendency of users to change hosts makes comprehensivedata collection difficult. Furthermore, defining differentlevels of participation while controlling for the enormousvariation in the different types of weblogs is a massiveundertaking, making any study of social network attributes andcontinued participation either exceedingly difficult ordramatically simplified This study seeks to address theseconcerns by using a unique dataset to examine the relationshipbetween social network ties and continued participation in aweblogging system. The dataset, which is taken from thedatabase of activity within the Wallop weblogging and socialnetworking system, includes full information about each users'invitation and entry into the system, activity in the system,and interactions with other users. These activity logs can beused to generate dataon the full networkof social interactionswithin theWallop system, andby combining the activity networkdata with changes in activity over time it is possible to

examine the relationship between social ties and continuedparticipation It is important to remember that Wallop is apartially closed unit of weblogging activity bounded by asnowball sample invitation mechanism. Although application ofthese findings may be applicable to weblogs and other contentcreation platforms even when they lack Wallop’ s membershipfeatures, the features that enable some aspects of thisresearch simultaneously set directly comparable to othersimilarly bounded social network systems such as Orkut3DATAANDMETHODS

Datawere taken from the Wallop back-end database, whichincludes arecord of all invitations anduploads for eachuser,along with date and time information. The database alsoincludes data associating responses or comments with theoriginal object. This makes it possible to reconstruct thenetwork of interactions created by users sending comments toeach other. It also makes it possible to reconstruct theinvitation tree created by existing Wallop users invitingpotential new users to join the system Three types ofnetworks, a comment network 4,285 nodes, 14,884 edges, aninvitation network 4,514, nodes, 4,255 edges, and a combinationnetwork 3, 119 nodes, 4,323 edges, were created for this study.

in the analysis, are defined as comments sent from one user toanother. The comment network was constructed from data takenin a five weekperiod spanning the entire month of November 2004.This timeperiodwas chosenbecause itwas an interestinggrowthperiod in the Wallop lifecycle, occur-ring shortly after thesystem went public. It also has the advantage of producingnetworks which, while large, are substantially smaller thannetworks taken from similar periods later in the life cycle.A five week period was chosen as short enough to capture arelatively brief moment in time, while still being long enoughto provide rich data. The invitation network contains all nodesincluded in the comment network, plus any additional inviters.An invitationwhichdidnot needtobe accepteddefines anedge,and each user could only be invited once. The combinationnetwork is a subset of the comment network, where edges aredefined for any two nodes connected by both a comment and aninvitation The network data generated from the raw Wallopdata was used to generate variables, which were then includedin a logistic regression model. The regression model alsoincluded individual level variables, such as activitymeasuresand language group affiliation. Finally, the network data was

used to generate visualizations, whichwere used to illustrategeneral trends in user behavior and show some distinctionsbetween those who remained active and those who did not3. 1Network Variables

Except where noted, network variables for the analysiswere constructed from the comment network In-degree is thenumber of others who commented on a person’ s weblog. Thisvariable captures interaction received from an individual.Out-degree was not included because, as a measure of messagessent, it may not be representative of social interaction fromthe sender' s perspective. Receipt of a message is a slightlystronger indication to the recipient that there is a socialconnection. Reciprocal ties would be the ideal measure, but dueto difficulties with the database they were not available atthe timeof thisanalysis Densityis theratioof theobservednumber of ties divided by the imum possible ties [n-1*n-2] .Density is calculated for each node as the density of a subsetof the comment network, consisting of all nodes and edges withina two degree radius of ego. This is a measure of theconnectedness of an individual' s local network. The moreconnected an individual' s network, more likely it is that shewill feel she is part of a social group, as opposed to an