Team Dynamics in Long-Standing Technology-Supported Virtual Teams

Abstract

This paper examines team dynamics in a virtual team that has been developing and issuing updates to open source software code for a period of at least eight years. Our study seeks to extend the applicability of behavioral leadership theory to the broader examination of team dynamics in virtual teams. We developed and use a content analysis framework deductivelyderived from the literature on behavioral leadership to examine discourse, task-oriented, relationship-oriented, change, and network/boundary spanning communication behavior at two points in time in the ongoing interaction of the virtual team. Our findings support the notion the derived behavioral framework can be appreciably used to study emerging team dynamics over the lifecycle of virtual team interaction. Shifts in team dynamics not found in previous studies of virtual teams were found. In particular, relationship-oriented behavior, which was not found to be common in prior studies, was the most prevalent form of communication with in the team at the two points in time at which interaction was examined. Additional shifts were observed in the proportion of process and substance communication related to the team's task that were not found in earlier studies. Although preliminary, these findings suggest that team dynamics in long-standing technology-supported virtual teams differ from those found in virtual teams that focus on the completion of a single task within a relatively short-timeframe, which have been the focus of the majority of studies of virtual teams. Directions and implications for future research are discussed.

Exploring the influence of technological support, cultural constructs, and social networks on online cross-cultural learning

“…In Group II, every participant interacted with any member of the group, and the interaction was not limited to a tradition and culture a participant chose to learn and experience. Thus, communication style was different for each group -focused communication for Group I and social communication for Group II (Iosub et al, 2014;Matveev & Milter, 2004;Misiolek et al, 2012). This division was planned in advance of the experiment, and the participants were allocated to a group.…”

“…The latter is affiliative and includes expressions of support, agreement, and acknowledgment of the contributions of others (Iosub, Laniado, Castillo, Morell, & Kaltenbrunner, 2014). Misiolek, Crowston, and Seymour (2012) suggested that focused communication relates more to decision-making and problem-solving. That is, participants actively discuss and critically examine the contributions of others.…”

Optimizing Features in Active Machine Learning for Complex Qualitative Content Analysis

“…To optimize the initial machine learning model, we systematically ran multiple experiments using a gold standard corpus of emails from a free/libre/open-source software (FLOSS) development project coded for leadership behaviors (Misiolek et al, 2012). The coding scheme contained six dimensions: 1) social/relationship, 2) task process, 3) task substance, 4) dual process and substance, 5) change behaviors, and 6) networking.…”

“…To effectively implement the active learning approach for coding qualitative data, we have to first understand the nature and complexity of content analysis projects in social science research. Our pilot case study, an investigation of leadership behaviors exhibited in emails from a FLOSS development project (Misiolek, Crowston, & Seymour, 2012), reveals that it is common for researchers to use a multidimensional coding scheme consisting of a significant number of codes in their research inquiry. Previous work has shown that not all dimensions in a multidimensional coding scheme could be applied fully automatically with acceptable level of accuracy (Dönmez et al, 2005) but little is known if it is possible at all to train an optimal model for all codes using the same combination of features.…”

Semi-Automatic Content Analysis of Qualitative Data

“…We report preliminary findings from our efforts to optimize the configuration of ML models to perform the first pass of machine annotation. For these tests, we use a gold standard corpus created in a study of leadership behaviors exhibited in emails from a FLOSS development project (Misiolek et al, 2012). This gold standard corpus consisted of 408 email messages.…”