Ideally, Usenet members wouldmake efficient use of bandwidth,participating actively but judi-ciously in newsgroups, ensuringtheir comments are posted only torelevant newsgroups, and abidingby the local norms and culture thatgovern decorum. Everyone is betteroff if all behave insuch a manner, butthere is the temptation tofree-ride on the efforts of others.Thus, some participants post articlesthat are unnecessarily long, or lurkrather than contribute to the give-and-take that is an essential featureof any newsgroup, or post off-topicarticles, or violate the local rules ofdecorum. The more people free-ride, the more difficult it is to pro-duce useful information andinteraction. In the language of theUsenet, the signal-to-noise ratiodeteriorates. Newsgroups silt upwith commercial messages, fraud,and gibberish. The challengebecomes a question of how a groupof individuals can “orga-nize and governthemselves to obtaincollective benefits insituations where thetemptations tofree-ride and tobreak commit-ments are sub-stantial” [5].A key find-ing of collec-tive actionstudies showsthat mutualawareness of otherparticipants’ histories andrelationships is critical to a coop-erative outcome. The challengesof cooperation are heightened fur-ther when people are able to drawfrom a resource without contribu-Usenet, the globally distributed database ofconversation and other material, is a power-ful example of anarchic social organization.No one is in charge of Usenet, no centralauthority controls its borders or content in theway so many commercial services are ruled by sin-gle companies or groups. In 2000, at least 8 mil-lion unique participants created more than 150million messages unevenly distributed over50,000 or more newsgroups devoted toevery topic of possible human interest,from reselling Taiwanese householdgoods, to debating religion, to tradingsoftware. Like many related conversa-tional media, including email lists andWeb discussion boards, the problem isoften not finding others who shareyour interests. Instead, the challengeis dealing with too much content ofmixed value. TOOLS FORNAVIGATINGLARGE SOCIALCYBERSPACESMarc SmithCOMMUNICATIONS OF THE ACM April 2002/Vol. 45, No. 4 51The Netscan project helps online participantsform cooperativerelationships by offering a better sense of the otherplayers involved.LISA HANEY52 April 2002/Vol. 45, No. 4 COMMUNICATIONS OF THE ACMtion and when contributors are difficult to identify.It would make many of the social processes thatbind collective projects together more effective ifparticipants in social cyberspaces could easily createand access histories of one another and the spacesthey interact within [2]. But the tools available tointeract with these social spaces compound theproblem. Interfaces, like email and news browsers, that pro-vide access to social cyberspaces such as discussionboards, email lists, and chat rooms, present limited, ifany, information about the social context of the inter-actions they host. Basic social cues about the size andnature of groups are missing, making discovery, navi-gation, and self-regulation an increasing challenge asthe size and scope of these spaces expand. While peo-ple can eventually develop a refined sense of therhythms, leaders, and fools in a particular social cyber-space, the information does not come easily or easilytransfer to other spaces. With little sense of the pres-ence of other people, individuals have a difficult timeforming cooperative relationships. The goal of the Netscan project at MicrosoftResearch is to generate social accounting metrics of arange of dimensions of social cyberspaces in generaland the public Usenet feed in particular. Netscan is apublicly available Web service located atnetscan.research.microsoft.com. Social accountingmetrics are measures of the social dimensions ofonline spaces, like the size of newsgroups in terms ofmessages and participants along with the histories ofthe participating authors in each, such as how longthey have been active in the group, in what othernewsgroups they participate, to what threads of con-versation they contribute, and which other partici-pants they most often engage in discussion [4].Standard news browsers that present informationabout the messages themselves—their posting date,their subject, how many lines of text they contain—force users to pay more attention to the structure ofthe medium than the qualities of the participants whowould more naturally draw the user’s focus.A news browser would serve users far better, forexample, if it allowed users to sort and search news-group content by such salient behavioral attributesas the number of days on which each author con-tributed a message to the newsgroup, or the fractionof each author’s messages that were replies to exist-ing messages versus initial turns or postings. Com-bined with information about the structure anddevelopment of threads, the chains of turns andresponses created in these spaces, these metrics canbe used to extract content of likely value out of large,active social cyberspaces. Various aspects of people and conversations can beused as guides through the welter of content depend-ing on the kind of task pursued. Such measures canbe used to separate active groups from dormant ordead ones and to distinguish discussion newsgroupsfrom job listings and other non-interactive news-group types. The existing process of newsgroup dis-covery is hit-or-miss since as much as two-thirds of allnewsgroups are desolate or mostly inactive. Morethan 100,000 newsgroups exist but only about50,000 receive more than a single message a day.Only 20,000 receive more than one a day. Most newsbrowsers offer no indication of the population ofnewsgroups or their general levels of activity overtime. As a result, finding an active newsgroup on anappropriate topic can be a chore. To address this, the Netscan Web interface pro-vides a newsgroup search engine that takes a string asinput and returns a list of all the newsgroups contain-ing that string in their name (see Figure 1). Mostnews browsers provide a similar feature but return alist of newsgroups without any indication of the socialproperties of the space. The Netscan interface reportsthe number of messages and authors that are presentin each newsgroup returned by the query. These mea-sures help distinguish active from inactive news-groups and broadcast newsgroups (which have manymessages but few authors) and more diverse news-groups with