The document describes a study that examined how emergent leaders establish themselves as leaders in computer-mediated communication compared to assigned leaders. 92 Cornell students participated in an online murder mystery task in groups of 4, with some groups having an assigned leader and others allowing leadership to emerge. Language use, word count, and leadership ratings were analyzed. Preliminary results found emergent leaders were rated similarly to assigned leaders by group members and used language similarly, with leaders using less first-person pronouns and more plural pronouns than non-leaders. The study aims to help understand and design tools to support emergent leadership online.
As the structure of today’s organizations become more complex, dynamic, and global, virtual teams are becoming popular to adapt to these changes. These groups can be coworkers who need to complete a task or students who are getting an online education. Advancements in information and communication technology allows virtual teams to be flexible in their working environments. This is especially true for groups that are geographically and/or organizationally dispersed. Traditional face-to-face interaction does not exist in virtual teams. There is no eye contact, hand gestures, or body language. On top of completing the task and efficiently communicating with others, it may not always be the case that a leader is assigned to lead the team. This makes virtual leadership even more challenging. Team Now and Later is interested in the language of leaders when groups utilize CMC instead of F2F.
Organizational and group research has shown that effective leaders can enhance performance. This occurs through planning, organizing, and delegating. Leaders therefore increase likelihood of success. In a traditional F2F setting, leadership can be studied with behavioral theory – their social and task-related interactions. Misiolek and Heckman’s study supports that behavioral theory can be applied to studying virtual leadership. In our study, we hope to determine if these behaviors are transpired through the language of emergent leaders.
With our team’s interest in language of emergent leaders, we ask ourselves: From here, let’s discuss our methods used to approach these questions.
So far, we have had 92 Cornell students participate in our study. These students were recruited through SUSAN, and were given 2 credits for participating. Also, thanks to Professor Hancock, we were able to offer $10 to participants for a period of time. This greatly increased the number of participants who signed up. We are continuing to run experiments to gain more data. So far, we have run 23 groups of 4. We have had 11 assigned groups and 12 emerging.
We used the Murder Mystery Task which was created by Balakrishnan et al and given to us by Cornell professor, Sue Fussell, who worked on this paper. The task has 8 documents, and is meant to be solved by 4 people. Each participant is given 2 case documents, and they must solve the case solely through an online chat system. This makes the task more difficult. Our study included one manipulation. We had 2 randomly assigned groups. In the assigned leader group, 1 person was selected to be the leader, and the 3 other group members were told that this person was the leader. In the emerging leader group, 1 person was selected to be the leader, but the other group members did not know that this person was assigned to be the leader, and this leader could not tell the group members that they were a leader. Participants had 30 minutes to complete this task, which involved reading their individual case documents and communicating with their team through instant messenger. After the task, they filled out a post-experiment study which asked questions about who they wanted to arrest, how their team members behaved, and included a leadership inventory.
In our analysis we are looking at 2 independent variables. First, we are comparing assigned leader groups to emerging leader groups. For example, we are looking at how successful assigned groups were in arresting a suspect, vs. emering groups. Second, we are comparing information about leaders v. non-leaders in groups. For example, we are looking into leadership ratings of each group member.
In order to investigate our research questions we chose to initially look at certain measures. This first measure is leadership ratings of emergent leaders, emergent non-leaders, assigned leaders and assigned non-leaders. Literature has shown that in order to have effective leadership, group members much recognize that there is a leader. By looking at leadership ratings we can determine whether assigned leaders were seen as such and whether the pre-determined emergent leaders emerged. This will help us interpret the remainder of our analysis
We also looked at the group consensus and success rate of assigned leader groups and emerging leader groups. This question has implications for designing systems to facilitate leadership because if emerging leader groups are consistently performing and coming to consensus at rates much lower than assigned leader groups, it may be more beneficial to support leadership in a way that makes the leader known to all group members. If there is not a difference in success and consensus rates between the two groups that facilitating emerging leaders without calling them out may be a viable system feature.
Lastly we looked at the language differences between leaders and non-leaders as well as emerging and assigned groups. We used the LIWC tool for this analysis and at this point in time have looked at word count and use of pronouns. Previous work has shown that people of high status and low status use pronouns at different frequencies so we are using this finding and hypothesizing that leaders will speak as high status individuals and non-leaders as low status individuals. Both of these features, Word count and pronoun use, would be straightforward to incorporate into a model to predict leadership within a CMC system.
After participants completed the murder mystery task, they were asked to rate the leadership ability of themselves and their groupmates on a 5 point (Liquor) Scale A : We predicted that “assigned and emerging leaders…” because the B : We also predicted that “assigned leaders would be…” because each groupmember was aware that one member was the Chief detective and would rate them as such KYLE H1a – Although we predicted that the leaders, regardless of condition, would be rated higher than non-leaders, our data does not confirm our hypothesis. Even though this difference is significant in the assigned condition, this was not the case in the emerging condition. One possible explanation for this is that there may have been non-pre-determined emerging leaders in the emerging condition. H1b – Our data also suggests that our second hypothesis is true; assigned leaders did receive a higher rating than emergent leaders, however it only approaches significance. We believe that further data collection and analysis would make these results more significant and confirm our hypothesis.
In the post study survey, participants were asked to choose a suspect to arrest. If all four groupmembers chose the same suspect, we indicated that the group reached a consensus. A: We predicted that “assigned leader groups will have a higher group consensus rate” because the assigned leader groups have a defined leader If all four groupmembers chose the same suspect AND chose the correct suspect, we indicated that the group was successful B: We predicted that “assigned leader groups will have a higher success rate” because prior research has shown that groups are more successful when they have effective leadership KYLE H2a – Groups in the emerging condition came to a consensus more often than those in the assigned condition, contrary to our hypothesis. One explanation is that non-leaders in the assigned condition may have came to a consensus in the chat, but chose another suspect in the post-experiment survey. In the emerging condition, the group comes to a consensus together as opposed to having a designated leader choose a suspect. H2b – Literature suggests that groups with leadership perform better than those without any leadership. While assigned groups have a slightly higher success rate, this difference is not significant. This shows a promising path for future study in this area of emergent leadership leading us to believe that it is worth designing systems for emergent leaders, not just trying to enforce assigned leaders.
Along with the success rates of groups, we were also interested in the language patterns of participants A: We predicted that an analysis of the participants’ transcripts would show that “participants in leader roles will use more…” because leaders may need to use more words to help coordinate group members. Previous research has also shown that leaders typically support both social and task-oriented conversations, more words may need to be used for that B: predicted that “emerging leaders use more words…” Since everyone is not aware that they are the leader, the emerging leader may need to use more words to establish themselves KYLE H3a –We did not find support for our hypothesis that leaders used more words than non-leaders as these two averages were not statistically significant. In the assigned condition leaders did use more words, but this difference was not significant. In the emerging condition non-leaders used more words, but this was also not a significant difference. H3b- We also did not find support for our hypothesis that leaders in the emerging condition used more words than leaders in the assigned condition. We actually saw the opposite, but again this was not a significant difference.
Since we know that in emerging groups, someone else may have emerged as the leader, we wanted to look at word count vs. average leadership rating. To determine if word count was related to a groupmember’s leadership rating C: We predicted that there would be a positive correlation between the…” People who talk more and give more input, would be seen as leaders KYLE H3c – From the data we have collected and analyzed so far, there is not a positive correlation between leadership and word count, in fact there is a weak trend in the opposite direction. Further analysis and data collection will help us form a more solid conclusion.
In the murder mystery task, we expected the leaders to be of higher status than the nonleaders. Expected their language to reflect that. Our predictions are very similar to previous studies, as we believe “leaders will use more…” KYLE H4a – Data for both the assigned and emerging condition indicate that the leaders use first-person singular pronouns more than non-leaders. This suggests that our hypothesis is not supported because our results are significant in the opposite direction. Here is an example, you can see how the leader, Chief Skylar, uses I twice and the two detectives in this snippet only use it once in total.
We also used previous research to guide our prediction of “leaders will use more…” KYLE H4b – For both the assigned and emerging condition, the data suggest that leaders tend to use first-person plural pronouns more than non-leaders, however this result was only approaching significance with a p-value of 0.2. Here’s an example, you can see how the leader says “I don’t think we can assume that” and then all of the other team members continue on using “I”, not “we”
In addition, previous research suggests that “more first person…” Possibly because leaders make more recommendations and delegate more tasks KYLE H4c – Our data suggest that leaders use second person pronouns more often than non-leaders however this is not statistically significant. This finding is stronger within the assigned condition, where it is approaching significance with a p value of .19. In the emerging condition we see the opposite result with non-leaders using more second-person pronouns, but again there is no significant difference between the averages. One plausible explanation is that assigned leaders are comfortable delegating tasks, however emerging leaders feel the need to be more delicate in this manner. Here’s an example where the assigned leader uses “you” twice, and the other detectives do not use “you”. In this example the pre-determined emerging leader, Detective Alex, does not use you, however the other member do.
Even though we have a good start, there’s more work to be done in our research. The next steps would be a “deeper quantitative” There are aspects of our data that we have not looked into yet, such as leadership inventory, gender differences, etc. which could further explain our findings We would also like to perform some qualitative coding of transcripts. Finally we would hope to design tools to better facilitate cues and support emergent leaders, and test these designs.