(PDF) The Rules of Virtual Groups

Proceedings of the 38th Hawaii International Conference on System Sciences - 2005 The Rules of Virtual Groups Joseph B. Walther Ulla Bunz Natalia N. Bazarova Cornell University Rutgers University Cornell University

Abstract leading to subjectively-experienced trust and/or liking, or trusting behaviors that are associated with Concerns about virtual groups include how group productivity. A second concern in our members develop trust and liking for partners. understanding of distributed groups is to identify the Previous studies have explored behavioral variables kinds of communication behaviors that facilitate leading to subjectively-experienced trust and/or effective interaction, positive affect, and group liking, or trusting behaviors that are associated with performance. A number of researchers have group productivity. Drawing on practices described examined ad hoc behaviors [e.g. 3] or examined elsewhere and deriving principles from social behaviors post hoc [e.g. 4]; fewer have deliberately information processing theory of computer-mediated instigated communication behaviors in virtual group communication, this research identified a set of settings in order to evaluate their impact. Drawing on communication rules for virtual groups. A quasi- the best practices of virtual groups described in experimental procedure promoted variance in rule several studies, and deriving additional rules from following behavior, allowing assessment of rules in social information processing theory of computer- an inter-university course. Six rules were assessed mediated communication (CMC) [5], this research through self-reported measures, and results reveal employed a quasi-experimental procedure to increase correlations between each rule and trust and liking. the variance in rule following behavior, so that the Less consistent are the relationships between rule- utility of some of these best practices could be following and actual performance. Results suggest assessed. that either a powerful set of rules has been identified, This research analyzes several processes or that rule-following per se reduces uncertainty and identified by previous research as important for enhances trust in distributed work teams. successful virtual teamwork despite the challenges proffered by mediated communication and geographic dispersion of team members, and their impacts on trust, relationships, and performance. 1. Introduction 1.1. Virtual teams A surge of research interest in, and increasingly widespread adoption of distributed work in various Virtual teams are comprised of group members settings indicates a growing popularity of virtual who collaborate from different locations using teams supported by modern communication communication technology. Virtual teams can offer technology. Understanding the promise and the flexibility, responsiveness, and diversity of problems of these sociotechical systems can inform perspectives in ways that differ from traditional contemporary theory in the areas of new technology, groups. Despite these benefits, however, virtual group communication, and social psychology, while teams encounter numerous challenges due to their informing management and shaping applications such dispersion and communication limitations, which can as distributed learning groups and virtual impede their effectiveness, or at least require great organizations. Even though the concept of distributed efforts to accommodate to the virtual environment work is not new [see for review 1, 2], new and virtual partners. perspectives on enduring social dynamics come into view from the vantage point of electronically- 1.1.1. Impacts of virtuality. Geographic dispersion connected teams. among group members incurs a number of disruptive A longstanding concern in understanding the effects. These include incongruities in work prospects for virtual groups is how their members environments and social structures, dissimilar develop trust and liking for their partners. Several organizational cultures, and temporal differences previous studies have explored behavioral variables associated with different locations. Any or all of 0-7695-2268-8/05/$20.00 (C) 2005 IEEE 1 Proceedings of the 38th Hawaii International Conference on System Sciences - 2005 these can create disparity in working contexts for Finally, the quality of CMC groups’ output may team members and disrupt the work-flow and shared be suboptimal compared to their face-to-face (FtF) experience of such groups [6]. counterparts’. This may be due to dynamics related to Communication media, and the limitations of the the relational aspects discussed above, or due to CMC used by virtual teams, are also characterized as discretionary participation with asynchronous problematic. Nardi and Whittaker [7], for example, participation, free-riding, difficulty integrating argue that proximity and face-to-face interaction are information, or other information-processing aspects critical for establishing collaborative interpersonal of virtual work [see e.g. 17]. A meta-analysis of such relationships built on non-task communication early results, however, was inconclusive [18]. in working partners’ interactions. In subsequent In sum, a number of problems associated with interactions, Nardi and Whittaker argue, collaborators distance and restricted communication media have need full-cue communication in order to detect when been alleged to impact the ability of distributed partners are paying attention, listening, and groups to function as effectively as FtF groups. From exhibiting backchanneling behaviors, which CMC one perspective these disparities may seem does not readily provide [see also 8]. insurmountable, if the constraints of distance and The potential problems from dispersion and media are relatively impervious. However, other media may affect a variety of outcomes, such as research has indicated that the bases for the relational communication and trust, as well as the presumably deleterious effects of distance and media productivity and quality of the group’s work. are more permeable. Accommodations may emerge Relational communication pertains to the reciprocal or be applied by which participants adjust to the processes of how group members regard one another alternative environment, and the bases of the and how they express that regard. The dimensions of relational dynamics themselves may change. this regard include their affection, cohesion, and task- versus social orientation, among other themes. 1.1.2. Accommodations to virtuality. Alternative Several prominent approaches to the effects of CMC approaches and several specific findings challenge have argued that since relational cues are normally the contention that virtual groups are impotent with conveyed nonverbally in traditional communication, regard to relational dynamics, trust, and performance. the relative absence of nonverbal cues from email These alternatives also suggest possible strategies for and computer conferencing occlude the expression of the deliberate remediation of virtual groups’ potential interpersonal dynamics. As a result, it has been problems. argued, CMC may make it difficult, if not impossible, for the development and detection of relational Walther’s social information processing theory aspects critical to the social dimensions of group [5] predicts that CMC users adapt to the medium’s work [see for review 9]. restriction of nonverbal cues by imbuing their text- Trust is another relational dimension that has based messages with both task and social significant import for virtual teams, and one that has information. Due to the real decrements in carrying received particular attention in the virtual teams capacity of the CMC medium, computer-mediated literature. Handy [10] asserted that trust cannot be exchanges require more frequent interactions and/or maintained in virtual teams. According to Jarvenpaa more time in order for users to accrue the task and and Leidner [11], trust is traditionally considered to social information about partners to reach the level of be based on “personal relationships and past or future relational development that FtF partners accomplish memberships in common social networks that define more quickly. Given enough time, however, long the shared norms of obligation and responsibility,” term CMC groups can and do achieve liking, trust, which may not fit the conditions of virtual teamwork. and sociable states [20; see for review 9]. Indeed, Other conceptualizations of trust fit CMC better; trust whereas short-term CMC group members benefit in CMC “refers to an expectancy held by an from seeing pictures of one another in order to hasten individual or a group that the word, promise, or their social attraction, group members who exchange verbal or written statement of another individual or a number of messages over time, with no photos to group can be relied upon” [12, p. 718]. In the CMC help them, like each other more than short term literature trust has been found to be positively related groups, and more than they do if they eventually see to performance [13, 4], problem solving and each others’ pictures [21]. uncertainty resolution, as well as social information In line with this notion of accrual, Weisband and exchange [11], and liking [14]. As with other Atwater [19] compared experimental CMC and FtF relational dynamics, trust has been asserted and groups on the members’ degree of liking for one found to diminish with the relative lack of visual and another. They found a significant correlation in CMC vocal cues that text-based CMC implies [e.g. 15, 16]. groups between the frequency of a member’s task- 0-7695-2268-8/05/$20.00 (C) 2005 IEEE 2 Proceedings of the 38th Hawaii International Conference on System Sciences - 2005 related message postings and the degree to which of trust had been sociable, exchanged intensely other members liked the contributor. In FtF groups frequent messages, showed interest in other there was no such relationship. The authors members’ responses, showed initiative, provided concluded that liking in off-line groups is based on substantive feedback to one another, and notified non-rational or non-task bases of attraction, whereas others of their expected participation periods or in CMC task-oriented groups, people like those more absences. Those with the lowest levels of trust who contribute more to the group’s endeavor. One exhibited little initiative and had little social content appliction of this finding may be to encourage virtual in their messages. Groups with moderate trust levels group members to communicate task contributions had predictable but infrequent communication, very frequently. focused their messages on tasks only, and devoted a Both the Weisband and Atwater [19] results and disproportionate level of messages to establishing those of Walther and colleagues suggest that rules and procedures. accumulated messaging leads to greater liking. It is Further analyses led Jarvenpaa et al. [4] to unclear whether messages must be task-related, specify several antecedents of enduring trust in online socially-related, or whether either will do. Both have groups. Their findings confirm that in online been implicated in the literature [e.g. 22]. Thus, relationships trust is maintained through performance another implication may be to encourage frequent consistency rather than cognitive or affective exchange of social and sociable messages, in addition perceptions alone. to task messages. A research question addresses this In another investigation of trust, Iacono and issue: Weisband [25] argued that “action forms” promote What are the relative effects of socially- and trust in virtual groups. These researchers focused task-related messages on affective and substantive specifically on the exchange of messages initiating outcomes in virtual groups? work processes (asking a specific question or As suggested above, trust is also an important proposing action, implicating a response from variable in virtual groups. Like the effects of time others), and messages responding directly to and and message accrual on group liking, trust develops thereby confirming such initiations, as behaviors that over time in longer-term virtual groups consistent equate to trust. Iacono and Weisband facilitated 14 with social information processing theory [20, 23]. self-selected virtual teams among students at several In other approaches, Jarvenpaa and Leidner [11] universities for three weeks, and coded message distinguished between swift trust and alternative, archives from these groups for initiations and developmental trust that may develop in online responses in several categories (getting together, groups. Swift trust is conceived as a depersonalized work process, work content, technical aspects, action based on categorically-derived information contact regulation, and fun). They also evaluated the and stereotypical assumptions about virtual partners quality of the groups’ work on papers that the groups and their behavior. It resembles enduring trust, but it collaboratively wrote as a measure of performance. is interpersonally untested [24]. In other words, No perceptual measures of trust were involved. before having interpersonal knowledge with which to Results showed that initiations and responses in both make a decision about the trustworthiness of specific the categories of work content (substantive colleagues, partners act as though they trust one contributions to the project) and work process (how another based on a presumption of trustworthiness, at to work as a group) were significantly associated least until shown they should not. with the quality of team performance. Additionally, Jarvenpaa and Leidner [11] argue that online while fun messages were few and not related to trust is not as depersonalized or categorically quality, the majority of fun messages occurred in stereotypical as swift trust. Rather, it may be more high performing teams. Additional analyses revealed behaviorally based, inferred from observations about that high performing teams formed quickly and other members’ electronic communication. These handled several activities at once: “If groups focus researchers facilitated 29 global, virtual student teams exclusively or primarily on work process issues, they comprised of 6 to 8 members over six weeks’ time. push the hard work related to work content to the last Participants completed self-reported measures of trust minute. (I)n electronic communications, people can after a median time period and at the end of the send multi-layered messages with a variety of types teams’ work. On the basis of these measures the of interactions (e.g. fun, procedure, contact, technical researchers classified teams as high or low in trust, information), but if they ignore a focus on work and then analyzed the transcripts of teams’ content, trust development and performance may interactions post hoc in order to identify behaviors suffer” (p. 8). Low performing teams struggled to common to trusting and untrusting teams. Analyses meet deadlines and worked most intensely suggested that team members with the highest levels immediately prior to deadlines, interacted less 0-7695-2268-8/05/$20.00 (C) 2005 IEEE 3 Proceedings of the 38th Hawaii International Conference on System Sciences - 2005 frequently overall, and had members who went takes longer for users to achieve satisfactory absent without explanation. development and accomplishment over time via their bandwidth-restricted exchanges. Traditionally, 1.2. Virtual team behavior control, and the groups tend to procrastinate the production phase of rules of virtual groups their work until half-way through their temporal existence [33]. Delays such as this which may be Given the theoretical connection between the normal in FtF groups can be much more deleterious frequency of specific behaviors and their positive in CMC as a group’s remaining calendar time may impacts on virtual groups’ relational and instrumental not provide sufficient interaction opportunity processes, then such behaviors, if encouraged or considering the retardation of information flow via mandated, might contribute to these groups’ CMC. This is especially true when using optimization. The question becomes what specific asynchronous CMC, where delays between message behaviors can be identified, and whether and how exchanges further strain temporal resources. Thus, they might be encouraged. According to Piccoli and CMC groups should start early in order to avoid Ives [26, p. 575], “Previous studies seem to implicitly running out of time even more severely than we assume that virtual teams will be self-directed—i.e. would see in procrastinating FtF groups. Iacono and that managerial control mechanisms are not required Weisband’s [25] results also indicate that more in this setting.” Even when virtual groups are successful groups began their work quickly. specifically instructed about useful practices or rules, Rule 2: Communicate frequently. As with the but left to their own decision to adopt them or not, first rule, social information processing theory [5] such groups tend to ignore these rules until direct indicates the importance that ample messaging interventions and social arrangements draw explicit portends for group development. The value of attention to their rule-following failures [27]. communicating frequently is also identified as a However, the managerial manipulation or critical behavior in Iacono and Weisband’s [25] and incentivization of such behaviors may provide a Jarvenpaa and Leidner’s [11] research, where trusting worthwhile experimental and practical approach to behaviors, or trust perceptions, were associated with understanding and improving virtual teams, and the frequent exchanges. Communicating frequently also deliberate management of virtual teams is a topic that allows the dispersion of activity over time, avoiding is receiving growing attention [e.g. 28]. the pile-up of activity toward the end of groups’ work In sociological and psychological literatures, the time periods. The limited nature of electronic and identification of communication rules and norms asynchronous communication may otherwise impede provides understanding of how groups and adequate rates of information-per-message-per-time relationships work, and how they might be affected exchanges. by deviations from normative or preferred behaviors Rule 3: Multitask getting organized and doing [e.g. 29, 30]. Norms and rules may be particularly substantive work simultaneously. This principle is potent in CMC. According to Lea and Spears [31], also derived from issues about temporal pace in CMC CMC offers little apparent information about and the normal sequencing of events in FtF groups. individual characteristics, making attributions based Traditionally, groups first set about organizing, on the conformity to or deviance from group norms defining, and allocating tasks before members begin more salient and important. Whereas Mark [32] has executing them. In CMC groups, especially those argued that implicit norms, or “conventions,” are with limited time, this sequence can be counter- more difficult to appropriate and observe in functional for two reasons. First, if approached distributed as opposed to long-term co-located linearly and sequentially, organizing activities may groups, and that explicit conventions invented by consume so much time that there is insufficient groups themselves are violated, her analysis did not opportunity for substantive exchange given the pace examine the deliberate introduction of behavioral of CMC. Secondly, the perception that it is rules as a managed or incentivized system. Based on undesireable to begin working on collaborative tasks the social information processing theory [5] and the before organizing them is fallacious in many cases. In findings reviewed above, a number of coordination tasks where duplication of effort might lead to rules were devised for empirical testing of their multiple perspectives about the same input effects on the relational and instrumental dynamics of information, organizing in order to minimize virtual teams in a system where the rules were duplication is not only unnecessary but explicitly encouraged and in some cases incentivized. disadvantageous. Not all tasks are of the magnitude Rule 1: Get started right away. Social or complexity to benefit from some duplication of information processing theory [5] describes the effort, but neither is it the case that all tasks must be temporal distortion that CMC incurs because CMC organized and allocated prior to first steps being 0-7695-2268-8/05/$20.00 (C) 2005 IEEE 4 Proceedings of the 38th Hawaii International Conference on System Sciences - 2005 taken. For the most efficient use of time in virtual Rule 6: Set deadlines and stick to them. There groups it may be better to begin substantive efforts on are several benefits to this suggestion. Most simply, it group tasks immediately rather than to wait until is important for work to be on time so that it can be every aspect has been negotiated. Iacono and legitimately accepted. More importantly, Weisband’s [25] findings on the simultaneous value accomplishing sub-tasks in a timely way reduces the of work process and work content exchanges lend uncertainty that participants may experience in an credence to this rule. interaction environment where there is less likely Rule 4: Overtly acknowledge that you have perceived accountability [34]. Social information read one another’s messages. In asynchronous processing theory [5] recognizes the communication CMC it is difficult to tell whether others have read of vulnerability patterns [35] as a means of one’s postings. Failure to do so may be due to developing trust in CMC. That is, actors must put senders’ errors (e.g. emailing to an individual rather themselves in a position in which another partner can than the group), system errors (e.g. network, server, provide an outcome which will, even if met, not be or software malfunctions), or receiver behavior greater in benefit to the actor than the potential costs (failing to check). Moreover, unless the email to the actor should the partner fail. If a member could software used has an automatic notification option, do the job herself but chooses to wait for a partner to once a receiver has gotten a message, it requires overt do it, there is vulnerability due to the greater costs if acknowledgement in order for the sender to know. In the member must share in the failure or do the work FtF settings message reception may be more herself if the partner reneges. Vulnerability and passively inferred through the attention monitoring validation lead to trust over repetitions. In a sense the that co-presence provides [7]. As discussed above, greatest demonstration of trustworthiness may be the time, channels, and member distribution make it honoring of agreements [12]. difficult for virtual team members to know if they are In order to test the utility of these rules and the experiencing common knowledge, and as Cramton principles on which they are based, the following [3] has detailed it is easy to assume that common general hypothesis was generated: knowledge is operating in distributed groups when in The greater the adherence to each of the rules fact it is not. Explicit acknowledgements should of virtual groups, the more that virtual team members combat this problem. Iacono and Weisband’s [25] trust the group, like each other, and perform better findings on the value of responses to proposals, not work. just the proposals themselves, supports this suggestion. 2. Methods Rule 5: Be explicit about what you are thinking and doing. Social information processing In order to ascertain whether any of these rules theory [5] holds that CMC can translate into verbal had impacts if they were managed, i.e. if variation in behavior that which is traditionally done nonverbally, their adherence could be deliberately manipulated, a and the articulation of feedback messages is one quasi-experimental procedure was employed. Such application of this principle. Given that there is no an approach would have great implications for the nonverbal backchanneling in CMC, message senders practical management of virtual groups, a notion cannot take advantage of partners’ nods of assent or deserving applied research attention. However, given head-shaking disagreement in order to know whether the multiplicity of rules and the limited field proposals have been accepted. Overt response experiment conditions for their evaluation, a single messages are beneficial. Such benefits were seen in research study could not manipulate all of the Iacono and Weisband’s [25] empirical results. conditions necessary for a purely experimental study Related to the common knowledge problems of all rules in a setting active enough for their described above [3], when one does not explicitly dynamics to accrue. In the following we describe a state agreement or disagreement with a proposal in a field experiment design that allowed us to stimulate virtual group, other members do not know, but may variance in adherence to various rules, and the falsely assume, that the group is in agreement. This methods by which we assessed their effects. can lead to conflict and misdirection about task allocations or decisions. Explicit verbal feedback about suggestions and proposals, while easy to 2.1. Participants neglect and potentially face-threatening if offered, may allow virtual groups to cohere on decisions and Using an inter-university course, groups were action plans more effectively than when suggestions composed of members from two major research go unconfirmed or unchallenged. universities in the northeastern United States. Participants (N = 44) were students enrolled in two 0-7695-2268-8/05/$20.00 (C) 2005 IEEE 5 Proceedings of the 38th Hawaii International Conference on System Sciences - 2005 senior-level courses focusing on the topic of CMC, second third of groups was assigned to follow rule 2, with 28 participants from one university and 16 from requiring them to communicate frequently. Groups the other. The courses were electives which did not with this rule were told that each member had to post require particular expertise in technological a message to the group on at least five out of seven applications. The students were informed at the consecutive days (and that posting more than one courses’ inception about the simultaneous message per day did not release a subsequent daily instructional and research objectives involved in requirement). A formula was presented by which course participation, for which continuation in the decrements would be penalized. The final third of courses implied consent. The courses involved groups was required to post at least as many several bi-weekly assignments, and three major group substantive messages as organizing messages (rule papers of which the first two provided the test-bed for 3). Substantive messages discuss the content of the the research design. As the results will show, assigned readings or reflect conclusions based on the assignment to experimental conditions harmed no readings or their synthesis. In contrast, organizing one’s grades on these assignments, and actually messages suggest or discuss procedures related to improved them. Participants were assigned to groups allocating, writing, and organizing the work leading of 3 or 4 members using a randomized/blocking up to the completion of the paper. In the first round strategy, so that each group had at least one member of papers, this latter rule was implemented as a from a different university, and that four-member simple ratio of substantive to organizing comments. groups were evenly distributed. A total of 28 unique This version of the rule did not facilitate frequent groups were formed, 14 for each of the two paper early posting of substantive messages to the extent projects. that it had been intended. Organizing messages dominated the early postings, although substantive 2.2. Treatments and channels message postings “caught up” in later exchanges. For the second round of group work, the rule was Groups had 18 days in which to develop the first modified to require that on any day that an organizing paper, and 20 days for the second, including two days message was posted, at least one substantive message of extension due to campus weather-related closings must also be posted. Penalty formulas were also and electrical outages. The papers consisted of articulated for this rule. Finally, it is very important collaborative research reviews of six primary to note that all groups were strongly and repeatedly research articles which were provided to the groups, encouraged to follow all the rules whether or not they on the topics of virtual community for the first were graded on adherence. project and on media selection for the second project. Communication was restricted to the use of 2.3 Measurement asynchronous bulletin-board discussions with file sharing capability, and synchronous chat, both of Dependent variables included several self- which were embedded in a course support system. administered measures and one observed variable. At The system recorded all postings and chats for later the completion of each paper, participants completed analysis (the results of which to be reported a web-administered questionnaire. The first measure elsewhere). Participants were admonished not to use included a single-item, five-interval scale measuring e-mail or Instant Messenger: Although such channels the participants’ assessment of the quality of their are often appealing to group members, previous group’s paper. Using original measures, participants research [3] has shown how email exchanges often also rated the degree to which they believed their develop among sub-sets of group members, groups adhered to each of the first six rules (from sometimes unknowingly, causing groups to lose “not at all” to “completely”) with respect to getting common knowledge about ideas. Consistent group started right away, communicating frequently, conferencing avoids this. acknowledging others’ messages, being explicit Treatments encouraging rule-related behavior about expectations and progress, multi-tasking were implemented in the following way. One third of content and organizing, and sticking to deadlines. the groups were instructed that they would be Next, participants completed subscales assessing the evaluated entirely on the quality of their group relational communication of another member of the papers. The other two-thirds of the groups were group who was a geographically-distributed, virtual instructed that 60% of their grades would be partner [20], for immediacy/affection (13 items, determined by the paper quality, and the other 40% Cronbach Į = .93), task-social orientation (4 items, Į would be determined on their adherence to one of = .67), and dominance (4 items, Į = .86). two specific rules their groups were assigned. The Participants also evaluated the same partner on liking. Weisband and Atwater’s [19] two liking items 0-7695-2268-8/05/$20.00 (C) 2005 IEEE 6 Proceedings of the 38th Hawaii International Conference on System Sciences - 2005 appeared to relate to task- or working-related indeed affected groups’ rule-following. As a form of attraction (“how much participants enjoyed working manipulation check, therefore, an analysis of with a particular member,” and “their desire to work variance was conducted to examine the effects of rule with the other member again,” p. 5), similar to the assignments on self-reported rule-following, which task attraction construct identified by McCrosky and involved collapsing groups into three categories: McCain [36]. For comparative purposes, scales from those assigned to communicate frequently, those McCroskey and McCain’s multidimensional measure assigned to multi-task organizing and substantive of attraction were employed measuring task attraction messages, and those with no specific rule assignment. and social attraction. Items from all sources were Consistent with intentions about the rule for subjected to factor analysis, and the two items from communicating frequently, there was a significant Weisband and Atwater loaded together with task difference in the extent to which groups reported attraction items, while the social attraction items frequent messaging, F (2, 83) = 13.62, p < .001. formed a distinct second factor. In subsequent Group members assigned to this rule reported a analyses the Weisband and Atwater [19] scales were significantly higher mean of rule adherence (M = combined with the task attraction scales for a 4.17, SD = .99), according to post hoc Scheffe composite measure. The final measures, with items analysis, than did members of the multitask rule dropped for low reliability, consisted of 8 items for groups (M = 3.39, SD = 1.15), which in turn task attraction (Į = .85) and 9 items for social communicated more frequently than groups with no attraction (Į = .94). rule (M = 2.64, SD = 1.11). Participants repeated these measures for a co- In terms of differences in multi-tasking, a located group partner if they had one, and another significant difference in self-reported rule-following virtual partner if they did not. Finally, participants also obtained, F (2, 83) = 5.59, p = .005. Group completed scales reflecting the trustworthiness of members assigned to the multitasking rule reported their group, using the five item scale employed by more multitasking (M = 3.71, SD = .99) than those Jarvenpaa and Leidner [11] who in turn developed it with no rule (M = 2.80, SD = 1.56), according to the from several related sources. Alpha reliability was Scheffe test. Those who had the rule to communicate .88. frequently also exhibited a moderate degree of All scales were analyzed at the individual level, multitasking (M = 3.23, SD = .97) and were not since individual scores varied more systematically by different from the multitasking rule condition or the conditions than by groups. Although each individual no rule condition. operated in two groups sequentially, repeated There were also significant differences in measures analyses were not used since individuals adherence to other rules, according to rule condition worked with different partners and different rules in assignments: getting started right away, F (2, 83) = the first and second tasks, mitigating the comparison 3.79, p = .027; acknowledging others, F (2, 83) = within-subjects analysis would otherwise provide. 5.73, p = .005; and explicitness, F (2, 83) = 4.95, p = The quality of the groups’ output was .009. No effect obtained on self reports of sticking to determined by using the average grade from both deadlines, F (2, 83) = 2.47, p = .091. For five of the instructors’ assessments of the group papers (using a six rules tested, however, there did seem to be a good 0-100 grading scale). Overall alpha for inter-rater degree of variance in rule-following behavior, at least agreement was .75 before any adjustment. On the to the extent reported by the participants themselves, first assignment three papers’ grades were discrepant which had been the purpose of the rule assignment between instructors. After the disagreement was manipulation. articulated and adjusted, final inter-rater reliability achieved .94. On the second round of papers 3.2. Hypothesis tests reliability was .88 prior to discussion. The next analyses examined the general 3. Results hypothesis about the impacts of rules, by computing correlations between the degree to which participants 3.1. Manipulation check reported that their groups followed each rule and the various outcomes to which they were predicted to Although the purpose of the assignment of relate. The most striking findings were the relatively certain rules to specific groups was not to create a strong and consistent correlations between the degree factorial design, but rather, to encourage greater to which participants reported following each rule variance in at least rule-following behavior, it was and the level of trust that they experienced, with each nevertheless worthwhile to see if the rule assignments 1-tailed correlation p < .001: getting started right 0-7695-2268-8/05/$20.00 (C) 2005 IEEE 7 Proceedings of the 38th Hawaii International Conference on System Sciences - 2005 away, r (86) = .43; communicating frequently, r = adherence to the rule for frequent communication, .65; acknowledging others, r = .57; being explicit, r = adj. R2 = .16, F (1, 84) = 16.68, p < .001. .67; multitasking, r = .45; and sticking to deadlines, r The research question about the relative impacts = .65. of socially-oriented and task-oriented messaging was Rule following was also associated with task assessed by examining the relationships of attraction, p < .001 in each case: getting started right participants’ assessments of partners’ affectionate away, r = .41; communicating frequently, r = .53; communication and task-oriented communication on acknowledging others, r = .53; being explicit, r = .60; the liking, trust, and performance quality measures. multitasking, r = .41; and sticking to deadlines, r = Consistent with the findings of Weisband and .60. Social attraction was also predicted by the extent Atwater’s [19] correlations between observed task each rule was followed, p < .021 or lower: getting messages and work-related liking, participants’ started right away, r = .22; communicating assessments of partners’ task oriented communication frequently, r = .45; acknowledging others, r = .37; was strongly associated with task attraction, r (86) = being explicit, r = .40; multitasking, r = .29; and 57, p < .001, 2-tailed. However, the correlation sticking to deadlines, r = .43. between affectionate communication ratings and task Additionally, following each rule corresponded attraction was greater, r = .82, p < .001. Social with how well the participants thought they had done attraction was associated with affectionate on their collaborative tasks, with p < .025 for each communication, r = .82, p < .001, and with task- rule’s correlation with self-rated success: getting oriented communication to a lesser extent, r = .23, p started right away, r = .21; communicating < .03. Trust, too, was associated with affection, r = frequently, r = .37; acknowledging others, r = .38; .66, p < .001, and task-oriented communication, r = being explicit, r = .49; multitasking, r = .41; and .37, p < .001. Groups experiencing more affectionate sticking to deadlines, r = .42. communication also performed better on their work, Interestingly, the objective quality of the both according to their own performance participants’ end results, their collaborative papers, assessments, r = .57, p < .001, and the instructors’, r was not correlated with each and every reported level = .47, p < .001. of rule-following. Most rules, again, did relate to this outcome, but with a somewhat lower magnitude than 4. Discussion they did with participants’ perceptions. In the case of the instructors’ average assessment of the quality of groups’ papers, multi-tasking did not appear to help The present study examined the degree to which groups perform better, r = .12, p = .16. The other the potential problems of virtual teams could be rules did appear to enhance the objectively-rated ameliorated by specific behavioral guidance, quality of teams’ work, p ” .025: getting started right associated with several rules for virtual teams. away, r = .21; communicating frequently, r = .41; Virtual groups sometimes do overcome geographic acknowledging others, r = .28; being explicit, r = .29; dispersion and channel constraints, and/or they and sticking to deadlines, r = .31. possibly could do so when we recognize different In order to assess which of the rules were most behavioral bases for judgments important to group influential in terms of each of these outcomes, work, and foster them. This study attempted examine multiple regression analyses were conducted. The some of these dynamics through assignments that led level of trust in groups was predicted most strongly to variations in key behaviors, and assessing the by three rule-following variables, with a total impacts of those behaviors on perceived and adjusted R2 of .55, F (3, 82) = 35.02, p < .001: being objective evaluations of virtual groups’ work. Results explicit (b = .62), sticking to deadlines (b = .30), and indicated that there are several viable behavioral communicating frequently (b = .23). Task attraction routines available to virtual groups and that the more was influenced most by sticking to deadlines (b = these behavioral routines are adopted, the better the .37) and being explicit (b = .35), adjusted R2 = .42, F experience and the better the results of virtual teams (2, 83) = 31.50, p < .001, whereas social attraction efforts. was predicted by communicating frequently (b = .23) The results of this study supported the and sticking to deadlines (b = 19), R2 = .22, F (2, 83) effectiveness of each of the six hypothetical = 12.86, p < .001. Although the self-assessed quality behavioral rules, and did so surprisingly strongly. of the groups’ work was predicted most strongly Indeed, the consistent and high correlations between predicted by explicitness, adj. R2 = .23, F (1, 84) = self-reports of the groups’ adherence to rules, and 26.50, p < .001, the objective assessment of the perceived trust, task attraction, social attraction, and groups’ work quality was affected most strongly by perceived quality of work were somewhat remarkable. Further, it appears that the perceptual 0-7695-2268-8/05/$20.00 (C) 2005 IEEE 8 Proceedings of the 38th Hawaii International Conference on System Sciences - 2005 states corresponded relatively well with an objective [2] M. O’Leary, W. Orlikowski, and J. Yates, (2002). assessment of the quality of groups’ work. When it “Distributed work over the centuries: Trust and control in came to the instructors’ outside evaluation of the the Hudson's Bay Company, 1670-1826,” in Distributed groups’ projects, almost every one of the rule- Work: New Research on Working Across Distance Using Technology, P. Hinds and S. Kiesler, Eds. Cambridge, MA: following reports correlated with paper quality also, MIT Press, 2002, pp. 27-54. with multitasking substantive/organizing messages being the only rule that failed to predict actual grade. [3] C. D. Cramton, (2001). “The mutual knowledge Two possible explanations for the results as a set problem and its consequences for dispersed collaboration,” are available. One is that, as hypothesized, a Organization Science, vol. 12, pp. 346-371, 2001. parsimonious and powerful set of rules has been [4] S. L. Jarvenpaa, K. Knoll, and D. Leidner, “Is identified through this research which affects the anybody out there?: The implications of trust in global continuum of behavior, perception, and outcomes. A virtual teams,” Journal of Management Information more skeptical but no less important interpretation Systems, vol. 14, pp. 29-64, 1998. may be that the efforts of the present and previous [5] J. B. Walther, “Interpersonal effects in computer- research to identify specific rules or best practices mediated interaction: A relational perspective,” that facilitate virtual groups are misguided. It may be Communication Research, vol. 19, pp. 52-90, 1992. that the mere following of any rules reduces uncertainty and leads to trust and liking in virtual [6] Hinds, P. J., and S. Kiesler, Eds., Distributed Work: groups. There will be two ways to assess this rather New Research on Working Across Distance Using grand but compelling contention. One will be to Technology. Cambridge: MIT Press, 2002. assess the actual behaviors associated with rule [7] B. Nardi and S. Whittaker, “The place of face to face following, with perceptions, and with performance communication in distributed work,” in Distributed Work: assessment variables. If behaviors that should New Research on Working Across Distance Using logically align with rules do in fact correspond with Technology, P. Hinds and S. Kiesler, Eds. Cambridge, MA: self-reports, we can be more confident that MIT Press, 2002, pp. 83-110. perceptions mapped on to what participants actually [8] J. Galagher and R. E. Kraut, “Computer-mediated did. Even if so, such analyses will not untangle the communication for intellectual teamwork: An experiment possibility that following any behavioral regimen in group writing,” Information Systems Research, vol. 5, might not have been just as effective. pp. 110-138, 1994. On the other hand, if these rules are not a [9] J. B. Walther and M. R. Parks, (2002). “Cues filtered placebo, they may be a panacea. Perhaps these rules out, cues filtered in: Computer-mediated communication and are beneficial not only to virtual groups, but to relationships,” in Handbook of Interpersonal groups in general. There is little in the rules Communication, 3rd ed., M. L. Knapp and J. A. Daly, Eds. themselves that does not look like good advice for Thousand Oaks, CA: Sage, 2002, pp. 529-563. any or all groups, virtual or not. Trust issues in groups are nothing new, and as one experiment has [10] C. Handy, “Trust and the virtual organization,” Harvard Business Review, vol. 73, pp. 40-50, 1995. found, more delicate and problematic in comparable FtF than virtual teams [34]. At the same time, the [11] S. L. Jarvenpaa and D. E. Leidner, D. E. (1998). notion that virtual partners know little else about one “Communication and trust in global virtual teams,” Journal another besides their online behavior suggests that of Computer-Mediated Communication [Online]. 3 (4). behavioral rules may have special potency in the Available: electronic environment. Compliance to or deviation http://www.ascusc.org/jcmc/vol3/issue4/jarvenpaa. html from rules may have stronger impacts online than [12] J. Feng, J. Lazar, and J. Preece, “Interpersonal trust offline. Such dynamics would be consistent with and empathy online: A fragile relationship,” in Proc. previous research in which attraction was rationally SIGCHI, 2003, pp. 718-719. based online but irrationally offline [19], and in other [13] W. F. Cascio, “Managing a virtual workplace,” research where attributions are more extreme in CMC Academy of Management Executive, vol. 14, pp. 81-90, than in FtF conditions [see for review 9]. 2000. 5. References [14] M. C. Greene, “Development of trust in on-line social relationships,” presented at the Conference on Computer- [1] J. L. King and R. L. Frost, “Managing distance over Supported Social Interaction, Oxford, OH, April 2002. time: The evolution of technologies of dis/ambiguation,” in Distributed Work: New Research on Working Across [15] N. Bos, J. Olson, D. Gergle, G. Olson, and Z. Wright Distance Using Technology, P. Hinds and S. Kiesler, Eds. (2002, April). “Effects of four computer-mediated Cambridge, MA: MIT Press, 2002, pp. 3-26. communications channels on trust development,” in Proc. SIGCHI, 2002, pp. 135-140. 0-7695-2268-8/05/$20.00 (C) 2005 IEEE 9 Proceedings of the 38th Hawaii International Conference on System Sciences - 2005 [16] E. Rocco, (1998). “Trust breaks down in electronic [27] J. B. Walther, M. Boos, and K. J. Jonas. (2002, Jan.). contexts but can be repaired by some initial face-to-face Misattribution and attributional redirection in distributed contact,” in Proc. SIGCHI, 1998, pp. 496-502. virtual groups. Proc. HICSS 35 [Online]. Available http://www.hicss.hawaii.edu/HICSS_35/ [17] J. Y. Smith and M. T. Vanacek, “Dispersed group HICSSpapers/PDFdocuments/OSVWE02. pdf decision making using nonsimultaneous computer conferencing: A report of research,” Journal of [28] B. L. Kirkman, B. Rosen, C. B. Gibson, P. E. Tesluk, Management Information Systems, vol. 7, pp. 71-92, 1990. and S. O. McPherson, “Five challenges to virtual team success: Lessons from Sabre, Inc.,” Academy of [18] P. L. McLeod, “An assessment of the experimental Management Executive, vol. 16, pp. 67-79, 2002. literature on electronic support of group work: Results of a meta-analysis,” Human-Computer Interaction, vol. 7, pp. [29] M. Argyle and M. Henderson, “The rules of 257-280, 1992. relationships,” in Understanding Personal Relationships: An Interdisciplinary Approach, S. Duck and D. Perlman, [19] S. Weisband and L. Atwater, “Evaluating self and Eds. London: Sage, 1985, pp. 63-84. others in electronic and face-to-face groups,” Journal of Applied Psychology, vol. 84, 632-639, 1999. [30] S. B. Shimanoff, (1988). “Group interaction via communication rules,” in Small Group Communication: A [20] J. B. Walther and J. K. Burgoon, “Relational Reader, 5th ed., R. S. Cathcart and L. A. Samovar, Eds. communication in computer-mediated interaction,” Human Dubuque, IA: Wm. C. Brown, 1988, pp. 50-64. Communication Research, vol. 19, pp. 50-88, 1992. [31] M. Lea and R. Spears, “Paralanguage and social [21] J. B. Walther, C. Slovacek, and L. C. Tidwell, “Is a perception in computer-mediated communication,” Journal picture worth a thousand words? Photographic images in of Organizational Computing, vol. 2, pp. 321-341, 1992. long term and short term virtual teams,” Communication Research, vol. 28, pp. 105-134, 2001. [32] G. Mark, “Conventions for coordinating electronic distributed work: A longitudinal study of groupware use,” [22] L. C. Abrams, R. Cross, E. Lesser, and D. Z. Levin, in Distributed Work: New Research on Working Across “Nurturing interpersonal trust in knowledge-sharing Distance Using Technology, P. Hinds and S. Kiesler, Eds. networks,” Academy of Management Executive, vol. 17, pp. Cambridge, MA: MIT Press, 2002, pp. 259-282. 64-77, 2003. [33] C. J. G. Gersick, “Time and transition in work teams: [23] J. M. Wilson, S. G. Straus, and W. J. McEvily, “The Toward a new model of group development,” Academy of development of trust in distributed groups over time,” Management Journal, vol. 31, pp. 9-41, 1988. presented at the Academy of Management conference, Toronto, Canada, 2000. [34] D. S. Staples and P. Ratnasingham, “Trust: The panacea of virtual management?” in Proc. ICIS, 1998, pp. [24] D. Meyerson, K. E. Weick, and R. M. Kramer, “Swift 128-144. trust and temporary groups,” in Trust in Organizations, R. M. Kramer and T. R. Tyler, Eds. Thousand Oaks, CA: [35] F. E. Millar and L. E. Rogers, (1976). “A relational Sage, 1996, pp. 166-195. approach to interpersonal communication,” in Explorations in Interpersonal Communication, G. R. Miller, Ed. Beverly [25] C. S. Iacono and S. Weisband, “Developing trust in Hills, CA: Sage, 1976, pp. 87-104. virtual teams,” presented at the 30th Hawaii International Conference on System Sciences, Maui, HI, Jan. 1997. [36] J. C. McCroskey and T. A. McCain, “The measurement of interpersonal attraction,” Speech [26] G. Piccoli and B. Ives, (2000). “Virtual teams: Monographs, vol. 41, pp. 261-266, 1974. Managerial behavior control’s impact on team effectiveness,” Proc. ICIS, 2000, pp. 575-580. 0-7695-2268-8/05/$20.00 (C) 2005 IEEE 10