JCMC 9 (2) January 2004
Collab-U  CMC Play  E-Commerce Symposium Net Law  InfoSpaces Usenet
  NetStudy VEs VOs O-Journ HigherEd Conversation Cyberspace Web Commerce
Vol. 6 No. 1  Vol. 6 No. 2  Vol. 6 No. 3  Vol. 6 No. 4  Vol. 7 No. 1  Vol. 7 No. 2  Vol. 7 No. 3
 Vol. 7 No. 4  Vol. 8 No. 1  Vol. 8 No. 2   Vol. 8 No. 3  Vol. 8 No. 4   Vol. 9 No. 1

Information Exchange in Virtual Communities: A Comparative Study

Gary Burnett and Harry Buerkle
Florida State University



Abstract

Burnett's (2000) typology of information exchange in virtual communities attempts to provide a framework for examining the range of activities undertaken by participants in such communities. This study is the first in a series to apply the typology to specific virtual communities, in an effort to assess its accuracy against the day-to-day interactions to be found in two online communities. Through a comparison of these two communities using the typology, revisions to the typology are proposed which will allow it to reflect more accurately activities found within the communities. By providing a metric through which to address such questions, the revised typology will allow a richer understanding of virtual communities as social information environments.


Introduction

Burnett's (2000) typology of information exchange in virtual communities attempts to provide a framework for examining the range of activities undertaken by participants in such communities. Derived from a substantial body of research as well as ongoing observation, this typology allows for the categorization of a virtual community's publicly posted messages, so that they may be analyzed to determine the degree to which that community can be thought of as an information (or purely social) environment for its participants. This study is the first in a series to apply the typology to specific virtual communities; thus, our primary focus here is on the typology itself, in an effort to assess its accuracy against the day-to-day interactions to be found in two online communities. Through a comparison of these two communities using the typology, we propose revisions to the typology which will allow it to reflect more accurately activities found within the communities, and propose further research directions.

A full description of this typology may be found in Burnett (2000). Table I provides a brief outline of the typology and its categories. While, as with other typologies, this categorization of behaviors in virtual communities presents a somewhat abstracted and simplified view of a complex set of human activities, its use can tell us a good deal about the relationship between information and socializing in online environments. By developing a structure through which individual posts may be categorized, the typology provides a mechanism for analyzing the extent to which a virtual community's public interactions can be examined to determine whether the exchange of messages emphasizes the sharing of information or more purely social activities. As would be expected in a situation in which information is exchanged informally as a part of normal and ongoing interactions, explicit information exchange is only one kind of interaction in a virtual community, as it is in the "real" world. In other words, although virtual communities include activities that are expressly aimed at gathering, sharing, or otherwise using information, it is not limited to those types of activities (c.f. Riva & Galimberti, 1998; Wellman, Salaff, & Dimitrova, 1996; Wellman & Gulia, 1999). Indeed, as Wellman and Gulia (1999, p. 172) have argued, "information is only one of many social resources that is exchanged on the Net." As group CMC environments become more ubiquitous in a wide variety of contexts, including education, virtual work, and civic participation, among others, it becomes increasingly important to understand how people behave and interact within those environments. Studies such as this will, potentially, have important implications for improvements in the design and management of virtual communities, helping them to become more fully functional and integrated parts of our social, informational, and work-related worlds.

As the typology notes, not all virtual community behaviors are part of the community's interactions. "Non Interactive Behaviors, "defined as "lurking" (or simply reading posts without ever writing one's own), are a significant activity, and "lurkers" may be, in fact, the largest single group within any virtual community (Smith, 1992). However, they cannot be directly observed, and, thus, "Interactive Behaviors," both hostile and collaborative, are of particular interest here.

Hostile behaviors, which have received considerable attention in the research (Kennedy, 2000; Smith, McLaughlin, & Osborne, 1998), include flames, trolling, spam, and (in extreme cases) cyber-rape. The presence of such behaviors, while they are sometimes embraced by specific communities,1 can reflect a range of situations, from momentary anger or conflict between participants, through chronic disagreements or community-wide enmity, to, in extreme cases, a community self-destructing from within through the weight of animosities growing from endless flaming.

Collaborative behaviors, on the other hand, can be seen as signs of a healthy, or, at least, functional, community and may, as some researchers argue, be more common than hostile behaviors such as flaming (Lea, O'Shea, Fung, & Spears, 1992; Rafaeli & Sudweeks, 1998). These behaviors range from informationally inconsequential but socially important activities such as the exchange of pleasantries, gossip, and jokes, through active emotional support, to more explicitly information-oriented exchanges, including announcements, queries and requests for information, and replies to such queries and requests.

The typology suggests that, while each of these behaviors will most likely be found across all virtual communities, specific communities may vary widely in the particular emphasis they place on certain types of behaviors; such a community-specific emphasis will be reflected in the relative frequency of those behaviors. The variance between communities can provide a way to assess the "small worlds" of those communities, and to compare the differences between their social norms and normative information behaviors (Burnett, Besant, & Chatman, 2001), and can also allow us to determine the relationship between social activities, play, conflict, and information exchange in virtual communities generally. Thus, while two virtual communities may be nominally focused on the same subject, such as, for example, a particular genre of music, they may differ radically in the amount of attention they give to simple social pleasantries or to the more substantive exchange of information related to their area of interest. Analysis of the differences in types of posts across virtual communities may give us a mechanism through which we can understand more clearly the role that information and information behavior in particular plays in the social environments we call virtual communities. The current study is the first in a series designed to apply the typology to specific virtual communities.


Health Information and Support on Usenet

The Internet is increasingly becoming a channel both for information about health issues and for geographically dispersed social support networks for individuals with a wide range of health problems, as well as for any others with an interest in those problems (Ferguson, 1997). Indeed, as recently as August, 2001, Massachusetts General Hospital and Women's Hospital in Boston announced that it was creating a web-based service through which patients could receive second opinions online from participating practitioners (Kowalczyk, 2001). Along with such increased utilization of the Internet for health information and support, there has also been increased concern regarding the helpfulness and even the accuracy of health-related online information sources (Preece, 1999).

In addition to the question of accuracy of information found in these online information sources, another source of concern is created when individuals utilize these open forums to deliberately misrepresent information to mislead other members by pretending to have the shared illness with which they are not personally afflicted. They intentionally divert the focus of the group upon themselves, sharing their feigned battles with the ailments shared by the group. Unfortunately, discovery of such deception by the members of the group can be devastating to the point of eliciting feelings of being violated or "raped" after having placed trust in the forum's intruder. It is for this reason that forum members must balance feelings of empathy with caution. Participants should be especially careful about basing their own health care decisions on uncorroborated information supplied in groups (Feldman, 2000).

The outlets for health information and support on the Internet, as might be expected, range from authorized channels (such as the second opinion service, cited above) to more informal forums, including numerous virtual communities related to specific health issues. For the current study, two Usenet newsgroups were chosen, both of which were established to provide health information and emotional support to participants who were suffering from common, but very serious, illnesses.2 Of these two newsgroups, one focuses on a life-threatening physical illness, while the other is devoted to a debilitating emotional disorder that can sometimes become life-threatening if untreated.3 Because both of these support communities are on Usenet, they are accessible to anybody with Internet access and either newsreader software4 or access to Web-based Usenet portals such as the Google search engine. Neither of these newsgroups is "moderated." That is, there is no individual in either group who takes responsibility for vetting the posts made to the group; as a result, anybody may post anything he or she wishes to either group, and both groups rely largely on the normative processes of their own internal social norms to define and enforce the acceptable behavior of group members (Burnett et al., 2001). The first of these groups has created and maintains an extensive FAQ ("Frequently Asked Questions") document, to which participants and newcomers may refer both for information on the group's area of interest and or guidance on the norms of expected behavior in group interactions. The other, which appears to be a spin-off from a related group, has no FAQ of its own.

Before our discussion of these two Usenet newsgroups, we review the literature about health-related support groups online.


Literature Review: Online Support Groups

As Wellman and Gulia (1999) note, there are relatively few full ethnographic studies of virtual communities. However, as they suggest, virtual communities can "open a door for people who would not ordinarily reach out for help," because they typically do not limit their interactions within specialized or narrow boundaries, but provide their participants with the benefits of both support groups and information services (Wellman & Gulia, 1999, p. 173). It should be noted that the distinction between those who are able to "reach out" and those who may not readily seek help from strangers is central to the nature of the virtual connectivity. "One of the empowering features of self-help groups is that members experience autonomy, control of the group, and a sense that they are experts on their problem" (Humphries & Rappaport, 1994, p. 219). The anonymity offered to members of these virtual communities can make it difficult for researchers to determine whether ethnicity or cultural influences contribute to those who are "reaching out."

A short generation ago support groups were a rarity, and support issues were dealt with in the nuclear or extended family, or perhaps the church. The success of Alcoholics Anonymous in the early twentieth century led to face-to-face groups such as Weight Watchers and, ultimately, to a plethora of Twelve-Step programs based on the AA structure, dealing with everything from overeating to sexual addiction. Psychologists discovered the efficacy of group therapy in the mid-20th Century. Given such a century-long trend, the recent evolution of self-help communities for those concerned with either addictions or health on the Internet is not a surprise. Such groups, which are, typically, populated by individuals in need of medical support and advice arguably represent a virtual self-healthcare revolution. In an analysis of electronic support groups for recovering addicts, King (1994) found that there are advantages and disadvantages to the sharing of recovery tools and experiences via these types of support groups. Advantages identified were the thoughtfulness of the replies to issues of the commonalities of the recovering addicts. Writing in an electronic forum provides an alternative to spoken conversation where the thoughts of the addicts can be formed more slowly and edited more carefully. Experiences and personal accomplishments expressed in a written form may allow the writer to express a message more clearly and profoundly than a similar message shared in a face to face setting (King, 1994, p. 48).

A study by Sparks (1992, p. 62) found that benefits of recovering addicts use of computer-mediated communication were that they have access to "24 hour availability, selective participation in entering and responding to messages, anonymity and privacy, immediate and/or delayed responding, and recording of transmissions." This is very different from traditional face-to-face support groups where members typically share only once, or not at all, during the hour they are together as a group. In addition, while users may easily disguise themselves either by using pseudonyms or, in extreme cases, faking illnesses they do not have, especially on Usenet, ongoing text-based interactions between a group of people who become "known" to each other over time can provide CMC support groups with a strong sense of presence.

These online forums, each of which is dedicated to a single health-related topic, offer technical medical information, practical advice on coping, emotional support through a sense of virtual presence, and online second opinions. Finn (1999), in a study of online self-help groups, found that such groups can provide many of the processes used in face-to-face self-help and mutual aid groups, with an emphasis on mutual problem-solving, information-sharing, expression of feelings, catharsis, and mutual support and empathy. Moreover, they can encourage patients to take an extremely responsible role in their own care. Most of these support communities are available to anybody who has a home computer and a modem, are often free, and sometimes feature volunteer health professionals who are experts on the condition in question (Ferguson, 1997).

Culver, Gerr, and Frumkin (1997) examined concern about the accuracy of information shared in medical support groups. The goal of their study was to assess medical information provided in a medically-oriented Internet discussion group, in terms of the professional status of the individuals providing information, the consistency of the information with standard medical practice, and the nature of the evidence cited in support of specific claims or recommendations. Their findings indicated that medical information available on Internet discussion groups may come from nonprofessionals and may be unconventional, based on limited evidence, and/or inappropriate. (While issues related to the trust of participants in information providers and to the accuracy of information provided in an online environment is important, it is beyond the scope of this article.)

Lebow (1998) shares a similar view of online medical support groups, stating that although marriage and family therapists should be aware that chat groups and newsgroups are not a substitute for therapy, they should understand this territory and provide guidance to those who use the resources. Lebow further suggests that the Internet can serve as an adjunct to marriage and family therapy if carefully moderated.

Schmetzke (1996), in an examination of online resources including e-mail, mailing lists, newsgroups, Gopher sites, and the World Wide Web, found electronic discussion groups to be the most important resource. Unlike Gopher and the WWW resources, electronic discussion groups, which include e-mail-based discussion forums and Usenet newsgroups, are not limited to the distribution of information and permit an intensity of personal support unmatched by any other Internet application. Madara (1997) concurs, suggesting that, although most mutual aid groups still offer help through face-to-face groups, an increasing number of people find it easier and more convenient to have their needs met through participation in online support networks. He contends that there has been a dramatic increase in the volume of online mutual help networks and people taking part in online groups, and attributes this growth in online groups to the fact that such groups are satisfying needs not often met in face-to-face groups. Indeed, this suggests that CMC environments may be more important for the social and emotional benefits users find there than for whatever informational role they may play.

Content analysis has been found to be an appropriate methodology for the study of online support groups. Drawing a sample from 27 online discussion groups (both on the Internet and commercial services such as CompuServ), Savicki, Lingenfelter, and Kelley (1996) coded 2692 messages for language content, using a modified version of the ProjectH Codebook (Rafaeli & Sudweeks, 1998). The typology developed by Savicki et al. based on the ProjectH Codebook differs from the typology used in the current study, and includes the following items: Although the work by Savicki et al. focuses on gender language style and group composition in Internet discussion groups (both issues beyond the scope of the current study), their typology has interesting implications for the study of information exchange in medical support groups. It may be that a single typology may not be appropriate for evaluating all discussion groups. For example, the groups studied for this report did exchange many of the types of messages identified by Savicki et. al. (1996), as well as other types that are not included in their work. Comparative analysis of multiple typologies may lead to further insights into the workings of virtual communities.

In the content analysis of an online cancer support group, Klemm, Hurst, Dearholt, and Trone (1999) identified an alternative typology, which includes the following categories: In the study by Klemm et al., these four categories accounted for approximately 80% of responses across groups. Groups did differ, however, in the emphasis they gave to these categories, with one (mostly male) group focusing primarily on the exchange of information and another (mostly female) focusing strongly on encouragement and support.

Rafaeli and Sudweeks (1998), developing a theory of interactivity, argued that studying "threads" or "chains" of interrelated and interdependent messages can provide a representative snapshot of communication in virtual communities online. By examining threads instead of individual messages, they found that the content on the net is less confrontational than is popularly believed: conversations are more helpful and social than competitive. Interactive messages seem to be more humorous, contain more self-disclosure, display a higher preference for agreement and contain many more first-person plural pronouns. As will be seen, this is clearly the case in one of the two groups discussed in the current study.


Categorization of Newsgroup Posts

For this study, we examined two health-related Usenet newsgroups for a period of one month. During this time, both groups displayed slow but steady traffic, with the number of posts ranging from slightly more than 20 per day for one group to slightly fewer than 40 for the other.5 To gather posts for analysis, we used the Web-based newsreader provided by the Google search engine (http://groups.google.com/). Because of the ways in which posts to Usenet newsgroups are distributed (forwarded from server to server around the world, rather than being sent to individual email accounts), a number of posts appeared more than once on the Google server; these posts were removed prior to our analysis. Because the typology suggests that participants' behavior is to be seen in their posts to the community as a whole, our unit of analysis, for gathering data, was the individual post itself. Each post was examined by a research assistant, coded according to the terminology of the typology, and placed into one of the categories. A random sample of posts were further examined by both authors to confirm the accuracy of coding; all posts that were found by the research assistant to be ambiguous, to fit into multiple categories, or to fall outside of the typology's categories were further examined and discussed by both authors. Because a significant number of posts were found that did not fit easily into the typology, two new categories were added during the analysis, as follows:
  1. The original typology, while it does account for queries to the community, does not contain a category for responses to those queries. Thus, a category was added for such posts.
  2. A category to account for Responses to Flames was added. Posts placed into this category responded directly to Hostile Interactive Behaviors such as Flames and Trolls, without themselves becoming Flames.
Table II presents these new categories under the heading of "Posts Undefined in Original Typology." The implications of these additions, and a couple of others, will be discussed in the following section. Our analysis of the activities within the two newsgroups here limits itself to an examination of the cumulative results of this classification. That is, here we do not directly examine individual posts, but only the total number of posts made in each category.

In this section, we discuss the general outcomes of this analysis, disregarding ambiguities and other subtleties that were discovered during our analysis. We discuss these issues in the next section, together with suggested changes to the typology to reflect such subtleties.

Table 2 summarizes our categorization of posts made to the two newsgroups. In this discussion, "Group A" refers to the newsgroup focused on a life-threatening illness, and "Group B" refers to the newsgroup focused on an emotional disorder. In addition to simply reporting the numbers, we provide some commentary and speculation regarding the implications of our categorizations.

Message Type
Postings Group A
%
Postings Group B
%
Hostile Interactive Activities
Flaming
3
0.26
61
9.53
Trolling
8
0.70
73
11.41
Spamming
3
0.26
35
5.47
Cyber-rape
0
0.00
1
0.16
 
Collaborative Interactive Behaviors: Non-Informational
Neutral
330
28.87
68
10.63
Humorous
40
3.50
37
5.78
Emotional
85
7.44
36
5.63
 
Collaborative Interactive Behaviors: Informational
Announcements
109
9.54
69
10.78
Queries
93
8.14
24
3.75
Group Projects
20
1.75
0
0.00
 
Posts Undefined in Original Typology
Responses to Flaming
0
0.00
96
15.00
Responses to Queries
452
39.55
140
21.88
 
TOTAL
1143
100.00
640
100.00

Table 2. Newsgroup posts categorized according to typology.


Hostile Interactive Behaviors

Hostile Interactive Behaviors-flames, trolls, spam, and cyber-rape-are those behaviors which do suggest interaction between members of the virtual community, but which emphasize overt aggression and conflict rather than either congeniality or the social exchange of information. The two groups, clearly, varied radically in terms of their tendencies to engage in such interpersonal sniping and hostile interactions. Given the subject focus of the two groups, such discrepancy seems to be predictable. The group brought together by a shared interest in a serious issue of physical health tends to avoid flames almost entirely, with only slightly more than one percent of all posts showing characteristics of flames, and no posts at all responding to those posts. On the other hand, the group discussing an emotional disorder seems, by comparison, to have chosen flaming as one of its predominant activities, with more than 41% of all posts being categorized either as being hostile, as responses to such hostility, or spam. Indeed, in this group, one post was overtly violent and sexually explicit to the point of being a "cyber-rape."

On the basis of their differences in engaging in-and supporting-hostile interactions, the two groups can, thus, be defined as quite different environments, with one being much more congenial and supportive than the other. Our analysis of posts in other categories supports this conclusion.


Collaborative Interactive Behaviors: Non-Informational

Collaborative interaction that is not explicitly informational in nature includes not only neutral "small-talk," but also the exchange of jokes and, significantly, expressions of emotional support. Thus, it might be expected that groups that are primarily social environments, rather than forums for information exchange, would emphasize such "chit-chat" and socio-emotional interactions.

Not surprisingly given their differences in hostile behaviors, the two groups also showed major differences in more positive social interactions. Group A, facing the travails of serious physical illness, engages in purely social interactions in almost 40% of its posts, while Group B does so in only slightly more than 22% of its posts. Both groups make substantial numbers of "small-talk" or neutral posts (almost 29% for Group A, compared to almost 11% for Group B). However, Group A shows a preference for emotional support over humor, while Group B included almost equal numbers of both. Interestingly, even though Group A made more than 500 more total posts than Group B over the month of observation, members of Group B made only three fewer humorous posts than Group A during that time period. Thus, we would infer that Group B, in addition to engaging in a high degree of conflict, also values joking; indeed, it may be that some of the posts that were categorized as flames may have been intended humorously, but were not perceived as such by other community members, and were thus treated as flames.


Collaborative Informational Behaviors: Informational

As would be expected from their differences in other categories, the two groups also differed significantly in terms of their overtly information-related activities, with Group A spending almost 20 % of its time explicitly exchanging information, compared to Group B, which spent only approximately 14 % of its time in such activities. Both groups emphasized the making of announcements in their information activities. However, Group A demonstrated a community-oriented approach to information activities, with a small but important number of their posts (20 posts, 1.75 % of the total) clearly being made as parts of ongoing group projects, as a way of building the community's information resources collaboratively.

When categorized according to the original typology, the posts made to Group A tend to suggest that this virtual community is primarily oriented toward socializing rather than the exchange of information. However, when a category is added to the typology to include responses to information queries, an important trend emerges, since nearly 40% of all posts to Group A fall into this category. When these posts are included, it becomes clear that Group A centers around the exchange of information, with almost 60% of all posts being directly related to overtly information-related activities. Thus, Group A can, we infer, be described as a virtual community that is very strongly oriented toward the exchange of information, with a very significant component of primarily social activities supporting it.

The addition of a category for responses to queries also makes a difference in the appearance of Group B. When such responses are added to the fairly small number of other information-related activities, it also can be seen as a community with a much stronger orientation toward the exchange of information than it might otherwise seem, with more than 36% of all posts being directly related to such activities. However, because of its clear preference for hostile interactions, Group B can be generally described as a virtual community oriented toward conflict, with subordinate but still substantial interests in more positive social activities and the exchange of information.

The typology, as currently configured, counts all queries that are posted to the community in the same way. However, in our analysis of the posts made to the two newsgroups, it was found that a significant number of queries posted to the groups were explicitly directed by their authors to specific individuals within the group rather than to the group as a whole. While the current typology does not reflect this distinction, the typology should be revised accordingly.


Implications and Revisions to the Typology

This project began as a test of Burnett's (2000) typology of information exchange in virtual communities, which was developed based on existing research in the virtual community literature and general observations of virtual communities, using it to analyze the interactions within specific online communities. Through this process, we hoped not only to come to some understanding about information-related interactions within two newsgroups, but also to assess the accuracy of the typology against the actuality of these specific virtual communities, and, thus, to determine ways in which the typology might need revision. Ultimately, our goal is to be able to use the typology to further our understanding of the nature of virtual communities as social information environments.

Although we found that, in general, the typology adequately accounted for the different types of posts to be found within the two newsgroups, three areas of concern did emerge in our analysis:
  1. Category Limits and Ambiguities.
  2. The Temporal Dimension. And
  3. Definitional Limits.
While none of these issues threatens to wholly undercut the usefulness of the typology, the challenges they pose must be met with revisions in the typology if it is to reflect accurately the function of information behavior within virtual communities. We will discuss each of the issues in turn.


Category Limits and Ambiguities

In our analysis of the two newsgroups, it was found that almost all posts were easily placed into the categories outlined in the typology. However, two categories, announcements and spam, proved to be difficult to distinguish in all cases, and another, flames, also proved to be problematic in unforeseen ways.

Announcements and Spam

In our analysis of the two newsgroups, we found that posts we have classified as "announcements" in fact contained two quite different types of information and, thus, played very different roles as mechanisms for the exchange of information within the communities. Thus, the "announcements" category of the typology should be sub-divided to contain the following two types of announcements:
  1. Pointers to outside information resources or events of potential interest to the community. These posts function as the virtual equivalent of flyers posted on a community bulletin board, or other types of "FYI" announcements.
  2. Personal updates. In the two newsgroups we examined, these types of posts typically included updates on the health of the poster, and were made to the community as a whole, on the assumption that participants would be interested.
However, a more important and challenging issue related to announcements involves their unexpectedly ambiguous relationship to spam. As the typology suggests, announcements and spam may be distinguished from each other by the degree to which they are on-topic. Spam, in the typology, is defined as being unrelated to the nominal subject interests of the virtual community, while announcements are defined as being "on-topic," or related to those subject interests. However, as the typology further notes, both types of posts expressly carry information, but, while spam is always unsolicited, announcements may be either solicited or unsolicited; because of this, announcements are, in the two newsgroups, a gray area for the typology.

Members of the two virtual communities typically perceived the posts that we have classified as "spam" as unwelcome intrusions from outside. In general, this is because it, like junk-mail, arrives unsought, and is often clearly off-topic. Thus, spam itself rarely causes ambiguity for the typology. Both informational and personal announcements, however, although they are made by community members, may be either sought or unsought by other participants. As a result, announcements, even though they were generated from within the community and were almost always clearly on-topic for the newsgroup, were often treated by participants in the newsgroups as if they were spam. Thus, it was difficult to determine whether certain posts should be classified as "spam" or as "announcements" in the typology; although the typology does make a clear distinction between the two (spam comes from outside the community, and is often off-topic, while announcements are generated by community members, and are largely on-topic) they often, in terms both of the perceptions of the community itself, and of their impact on the community, overlap, and are, thus, difficult to tell apart. That is, simple formal characteristics of posts (who presents them to the community and the degree to which they appear to be "on task") may not suffice for placing them within the typology. In such cases, ultimately, the meaning and social role of such posts may be a function of the ways in which the community as a whole responds in subsequent posts. (It should be noted that our interest here is not in how participants feel about posts outside of their public interactions, but rather in the characteristics and implications of the public interactions themselves; thus, our analyses in all cases limit themselves to the evidence available in the public texts of the communities themselves.)

To deal with this ambiguity in the current study, it was decided to simply classify posts as spam or announcements, depending on whether they were made by outsiders or by members of the community, regardless of either the response they received, or the degree to which they were on-topic. While such an arbitrary distinction does clarify the difference between two quite different types of post and between two quite different sources of information, further revision to the typology is also needed to deal with the ambiguous role of announcements in virtual communities. However, because the ambiguity of announcements appears to be related to the reception of such announcements by the community rather than something inherent (and easily identifiable) in the announcements themselves, it seems unlikely that simply revising the categories of the typology could reflect the differing roles of announcements adequately. Thus, the typology should be augmented with further descriptive analysis, which can be used not only to explain the choices made during the classification process, but also to enhance the power of the typology to describe the precise nature of information exchange within specific communities.

Flames

As a category in-and-of-themselves, flames present few problems, although the kind of post that is perceived as a flame can vary from community to community. However, in both newsgroups (and in other virtual communities we have observed), flames almost inevitably lead to numerous responses, some of which are, themselves, flames, some of which are exhortations to avoid a flame war by ignoring the flamer, and some of which answer the flame in some other way altogether. While many responses to a flame can be coded appropriately, according to their actual content (for example, a flaming response is, itself, a flame), many others do not fit easily within the typology as it is currently constituted. What, for instance, is to be done with a response that, avoiding the kinds of inflammatory and ad-hominem language typical of a flame, suggest that the flamer be ignored? Or, similarly, what is to be done with a response that, while clearly not a flame, addresses whatever substantive issues may have been raised in the flame? In order to deal with such posts during our analysis, we created a new category for responses to flames that did not, themselves, either become flames or fit easily into another category. As with announcements, the addition of this category can be enhanced by further descriptive analysis, which can give some additional context to aid in understanding the character of the interaction.


The Temporal Dimension

As with the questions raised above concerning the role of announcements and responses to flames, an analysis of individual posts removed from the context of the ongoing discussions of which they are a part tells only part of the story of a virtual community. Indeed, all posts made to a community are subject to interpretation by that community, and the social meaning of any post is often at least as much a matter of its reception and the details of the interpretations given it as it is of any meaning inherent in its text. As any online discussion continues through time, the meaning of the posts that began it are irreducibly mediated both by the memories of the participants and, more importantly, by the posts that have followed, which subtly but powerfully alter and revise its place and meaning in the social world of the community. As Rafaeli and Sudweeks (1998) note, the interactivity of an ongoing discussion is what gives online communities their particular robustness and flavor.

All of the conversations within a virtual community are made up out of the building blocks of individual posts; from this point of view, individual posts are logical units of analysis for an investigation into virtual communities. However, as our discussion of announcements and responses to flames makes clear, focusing only on individual posts can not only lead to ambiguities, but can also miss the give-and-take of a community's interactions over time, and can miss what might be called the "arc" of their conversation–the sometimes subtle and sometimes extreme shifts in tone of voice and focus of attention as discussion unfolds.

For example, the appearance of one or more flames at a given point of time in the conversations of a virtual community tells us nothing about the context in which they occur. Such posts may be "drive-by flames," virtually unconnected to anything that precedes or follows them. They may mark the beginning of a protracted flame war. Or, they may occur in medias res, in the heart of an ongoing flame war, may be responses to a single earlier non-flaming post, may either lead to further flaming, or be a small flurry of flames that are actually an anomaly within the community. Even full flame wars, as they unfold, may be poles apart in both tone and content, perhaps leading to deep rifts in the community, or, conversely, ultimately leading to some kind of agreement or shared understanding. Similarly, other types of conversations, whether they are predominantly social or whether they focus directly on the exchange of information, may have very different "arcs," and can differ radically in their depth and level of detail, and the degree to which any given post plays an active role in furthering the discussion. While these different conversational "arcs" clearly hold very different meanings for the community itself, they may, when analyzed and broken down for categorization according to the typology, appear to be very similar in terms of the numbers of posts of specific types they contain.

Further, the typology is designed, as in the current study, to be used across a fairly brief "slice of time," and to reflect the content of a virtual community's discussions at that particular moment in its history. As the above discussion suggests, such a moment, removed from what precedes and follows it, may or may not be representative of the activities of the community over time. However, unlike many forums for virtual communities, Usenet has been in existence since the early 1980s (Bumgarner, 1997). A full longitudanal analysis would require the classification of many millions of individual posts. Also, while the typology, applied narrowly (as in the current study) can provide little more than a snapshot of the state of informational and other interactions within a virtual community at a specific moment of time, a series of such snapshots, taken over time, can give a much richer and fuller picture of a community. Further, if the typology is applied simultaneously to a sampling of Usenet newsgroups, it should be able to provide a clear picture of a particular moment in Usenet history. Or, if such a study is undertaken longitudinally, it should add greatly to our understanding of information exchange and social activities on Usenet generally.

In other words, even though the typology is, as noted above, limited in terms of its ability to deal adequately with the temporal dimension of virtual community interactions, it can be repeatedly applied over time and, thus, provide solid longitudinal data. Unfortunately, such longitudinal data will still be unable to account for what we have called the "arc" of specific conversations. Analysis of such "arcs," while clearly beyond the scope of the typology, would require separate, supplementary studies, using extensive and detailed content analysis techniques. Thus, a minor revision should be made to the typology to reflect the fact that the temporality of online conversations is, explicitly, beyond its scope.


Definitional Limits

From the point of view of the information needs and behaviors of an individual participant, the typology is limited by its focus on the openly public interactions of the community. That is, the typology acknowledges only those exchanges that are made to the full community, explicitly disregarding any interactions that take place between individual community participants via other means, such as email, instant messaging, telephone calls, and even face-to-face encounters, even though such private exchanges may, in many cases, be perceived by the participants as of-a-piece with the more public interactions that take place within full view of all community participants. Further, such private exchanges, when they take place between community members, can have a powerful if hidden impact on both the "feel" and the activities of a community, because of their influence on personal relationships of individuals within the community.

However, virtual communities depend upon the public interactions of their members, and those public interactions, in the form of posts made to the group as a whole, form the very basis of whatever community is to be found within them. Thus, by viewing virtual communities as "information neighborhoods" (Burnett, 2000; Marchionini, 1995), while acknowledging that the information lives of their participants are broader than what can be seen in public, the typology explicitly treats only the public aspects of such neighborhoods. Those elements of the full information environment of the community–private interactions, interactions between individual community members and other communities or other information sources, etc.–are, by definition, beyond the scope of the typology.


Revisions to the Typology

With these issues in mind, it is clear that minor revisions must be made to the typology if it is to reflect more accurately the range of behaviors to be found within virtual communities. As noted above, certain limitations of the typology–its inability to account for the temporal dimension of virtual community interactions, the role played by a community's perceptions of the meaning of certain types of posts, and non-public interactions between community members–cannot be ameliorated by simple revisions, but require additional descriptive analysis. Thus, the typology should not stand by itself as the sole tool for analyzing virtual communities, but should always be accompanied by interpretive and descriptive commentary. For the current study, our analysis of the typology itself–its limitations as well as its overall adequacy–must stand in the place of such additional commentary. Further, we propose that the current study be repeated, using the same two Usenet newsgroups, at least twice at intervals of one year, to test the usefulness (and accuracy) of the typology's "slice of time" approach to virtual communities. Such longitudinal study is, clearly, beyond the scope of the current study.

Minor revisions to the typology itself, apart from any added commentary, will increase its capacity to measure the degree to which virtual communities engage in the exchange of information, compared to other activities, and will enhance its ability to provide an accurate reflection of the types of information-related activities to be found in virtual communities. Thus, for future studies, the typology will be revised in the following ways: These revisions will allow the typology to provide a more finely-grained analysis of interactions within virtual communities.

Table III summarizes the results of the current study, with these revisions added.

Message Type
Postings Group A
%
Postings Group B
%
Hostile Interactive Behaviors
Flaming
3
0.26
61
9.53
Trolling
8
0.70
73
11.41
Spamming
3
0.26
35
5.47
Cyber-rape
0
0.00
1
0.16
Responses to Flaming
0
0.00%
96
15.00
 
Collaborative Interactive Behaviors: Non-Informational
Neutral
330
28.87
68
10.63
Humorous
40
3.50
37
5.78
Emotional
85
7.44
36
5.63
 
Collaborative Interactive Behaviors: Informational
Announcements (Pointers to Information Sources)
82
7.17
50
7.81
Announcements (Personal Updates)
27
2.36
19
2.97
Queries to the Group
93
8.14
24
3.75
Group Projects
20
1.75
0
0.00
Responses to Queries
404
35.35
124
19.38
Queries to Individuals
48
4.20
16
2.50
 
TOTAL
1143
100.00
640
100.00

Table 3. Compiled postings from Group A & B


A full discussion of the two newsgroups in light of these changes to the typology is beyond the scope of this paper. However, the additions of the new categories–especially the categories that reflect queries directed to specific individuals and responses to flames–clarify further the differences between the two groups we discussed earlier, and show that, while one group actively engages in information-oriented behaviors both as a group and between individuals within the group, the other much more strikingly orients itself toward activities related to flaming and hostilities.


Conclusions

While the current project has focused explicitly on the validity of the typology itself, and only secondarily aims to describe the information environments of two medically related Usenet newsgroups, future projects, using the revised typology, will attempt to account more fully for virtual communities themselves, with an emphasis on the ways in which their members engage with information as a part of the their ongoing activities. Such future projects will examine other types of virtual communities, both in terms of their technical characteristics and in terms of their subject interests, and will address the following questions: By providing a metric through which to address such questions, the typology will allow a richer understanding of virtual communities as social information environments.


Footnotes

1. For example, the Usenet newsgroup alt.flame is devoted solely to the exchange of insults and other hostilities.

2. Although, as Walther (2001) points out in a comment posted to the Association of Internet Researchers listserv, Usenet newsgroups are public forums, with all posts easily and openly accessible to anybody who wishes to see them, we have opted neither to identify the two newsgroups by name nor to quote directly from posts. Because these two groups both involve the health of participants, and because, for at least some of them, questions related to their health could have implications for other aspects of their lives, it seemed to us to be respectful, if not quite imperative, to discuss them here without direct identification.

3. In the discussion that follows, we refer to the group devoted to the physical illness as "Group A." We refer to the group focused on the emotional disorder as "Group B."

4. Usenet reader software is typically provided as part of the standard access package offered by Internet Service Providers. Some readers (such as TIN and TASS) are command-line programs accessed via UNIX-based servers. Other reader software packages are built into Web browsers such as Netscape and Internet Explorer. The functionality of these different newsreaders can vary; however, they all, at a minimum, provide enough text-based access to newsgroups to allow users to become full participants in Usenet-based communities.

5. In a 1999 study, Smith found that approximately 3500 posts were made by approximately 1200 people to one or more newsgroups per hour. An informal inventory of newsgroups made at the time of the current study showed that the number of posts available in newsgroups ranged from one to several thousand at any given time. However, this informal observation accounts only for the number of posts housed at one time on a specific news server; since servers may vary widely in terms of the length of time they house posts, it does not accurately reflect actual day-to-day traffic.


References

Bumgarner, L. S. (1997). Usenet history. Retrieved October 17, 2001 from http://www.vrx.net/usenet/history/.

Burnett, G. (2000). Information exchange in virtual communities: A typology. Information Research: An International Electronic Journal, 5(4). Retrieved October 17, 2001 from http://www.shef.ac.uk/~is/publications/infres/ircont.html.

Burnett, G., Besant, M., & Chatman, E. A. (2001). Small worlds: Normative behavior in virtual communities and feminist bookselling. Journal of the American Society for Information Science and Technology, 52(7), 517-535.

Culver, J. D., Gerr, F., & Frumkin, H. (1997). Medical information on the internet: A study of an electronic bulletin board. Journal of General Internal Medicine, 12(8), 466-470.

Feldman, M. D. (2000). Munchausen by Internet: Detecting factitious illness and crisis on the Internet. Southern Journal of Medicine, 93, 669-672.

Ferguson, T. (1997). Health care in cyberspace: Patients lead a revolution. The Futurist, 31, 29.

Finn, J. (1999). An exploration of helping processes in an online self-help group focusing on issues of disability. Health Social Work, 24(3), 220-231.

Kennedy, T. (2000). An exploratory study of feminist experiences in cyberspace. Cyberpsychology & Behavior, 3(5), 707-719.

Klemm, P., Hurst, M., Dearholt, S. L., & Trone, S. R. (1999). Gender differences on Internet cancer support groups. Computer Nursing, 17(2), 65-72.

Kowalczyk , L. (2001, August 16). Hospitals offering online 2d opinions. Boston Globe, p. A1.

Lea, M., O'Shea, T., Fung, P., & Spears, R. (1992). 'Flaming' in computer-mediated communication: Observations, explanations, implications. In M. Lea (Ed.), Contexts of computer-mediated communication (pp. 89-112). New York: Harvester Wheatsheaf.

Lebow, J. (1998). Not just talk, maybe some risk: The therapeutic potentials and pitfalls of computer-mediated conversation. Journal of Marital and Family Therapy, 24(2), 203-206.

Madara, E. J. (1997). The mutual-aid self-help online revolution. Social Policy, 27, 20-26.

Marchionini, G. (1995). Information seeking in electronic environments. Cambridge: Cambridge University Press.

Preece, J. (1999). Empathic communities: Balancing emotional and factual communication. Interacting With Computers, 12(1), 63-77.

Rafaeli, S., Sudweeks, F., Mabry, E., & Konstan, J. (1998). ProjectH: A collaborative qualitative study of computer-mediated communication. In F. Sudweeks, M. McLaughlin, & S. Rafaeli (Editors), Network & netplay: Virtual groups on the Internet (pp. 265-282). Menlo Park, CA: AAAI Press/M.I.T. Press.

Rafaeli, S., & Sudweeks, F. (1998). Interactivity on the nets. in F. Sudweeks, M. McLaughlin, & S. Rafaeli (Editors), Network & netplay: Virtual groups on the Internet (pp. 173-189). Menlo Park, CA: AAAI Press/The MIT Press.

Riva, G., & Galimberti, C. (1998). Computer mediated communication: Identity and social interaction in an electronic environment. Genetic, Social, and General Psychology Monograph, 124(4), 434-464.

Savicki, V., Lingenfelter, D., & Kelley, D. (1996). Gender language style and group composition in internet discussion groups. Journal of Computer-Mediated Communication, 2(3). Retrieved October 17, 2001 from http://www.ascusc.org/jcmc/vol2/issue3/savicki.html.

Schmetzke, A. (1996). Disability-related resources on the internet. Intervention in School and Clinic, 32, 69-81.

Smith, C. B., McLaughlin, M. L., & Osborne, K. K. (1998). From terminal ineptitude to virtual sociopathy: How conduct is regulated on usenet. In F. Sudweeks, M. McLaughlin, & S. Rafaeli (Eds.), Network & netplay: Virtual groups on the internet (pp. 95-112). Menlo Park, CA: AAAI Press/MIT Press.

Smith, M. A. (1992) Voices from the WELL: The logic of the virtual commons. Retrieved October 17, 2001 from http://www.sscnet.ucla.edu/soc/csoc/papers/voices/Voices.htm.

Smith, M. A. (1999). Invisible crowds in cyberspace: Mapping the social structure of the usenet. In M. A. Smith & P. Kollock (Eds.), Communities in cyberspace (pp. 195-219). London & New York: Routledge.

Walther, J. (walthj@rpi.edu). (2001, May 8). Re: ethics of data use. E-mail to air-l ( air-l@info.comm.uic.edu).

Wellman, B., & Gulia, M. (1999). Virtual communities as communities: Net surfers don't ride alone. in M. Smith & P. Kollock (Editors), Communities in cyberspace (pp. 167-194). New York: Routledge.

Wellman, B., Salaff, J., & Dimitrova, D. (1996). Computer networks as social networks: Collaborative work, telework, and virtual community. Annual Review of Sociology, 22, 213-238.


About the Authors

Gary Burnett holds a Ph.D. in English from Princeton University and is an Assistant Professor at the School of Information Studies at Florida State University, where he teaches courses in the development and organization of online information resources. His research investigates the text-based information environments of virtual communities as well as the creation and maintenance of social norms within those communities. His work has been published in journals such as Information Research, The Journal of the American Society for Information Science and Technology, and Library Quarterly.
Address: School of Information Studies, Florida State University, Tallahassee, FL 32309.

Harry Buerkle is a doctoral candidate at FSU' School of Information Studies and currently works as an Information Technology Resource Specialist in the Leon County School's Exceptional Student Education Department in Tallahassee, Florida. His dissertation research focuses on the best practices of school/library media specialists and their integration of information and technology literacy in both the school and in the media center. He also has research interests in publishing via the WWW by children, web development and production, information literacy, computer-mediated communication, information typologies and distance learning.
Address: School of Information Studies, Florida State University, Tallahassee, FL 32309.


©Copyright 2004 Journal of Computer-Mediated Communication