Immersive technologies like virtual reality (VR) are more than just outlets to play games, socialize and increase productivity at work. They also offer new ways to explore human social dynamics in a controlled environment where every movement, gesture and action can be recorded and analyzed. This kind of work is the crux of Assistant Professor of Media Production, Management, and Technology Eugy Han’s research, and in a new study, she teamed up with researchers at Stanford and the Illinois Institute of Technology to examine how social groups form and change over time in VR.
Every week for eight weeks, eighty-one participants from a university course underwent sessions in VR. During these sessions, participants were tasked with completing various individual, small-group and large-group activities as researchers monitored their progress and interactions with others. Their nonverbal behaviors were also recorded and measured. Participants answered questionnaires at the beginning of the experience, after each session and at the end of the study. These questionnaires gathered data that couldn’t be measured objectively, like how close they felt to other participants or how they related to each other.
“In addition to being the third author, my role in this project was the creation and maintenance of this dataset,” Han explained. “This dataset is one of the largest and longest — i.e., longitudinal — VR datasets out there.”
All this data painted a detailed picture of how participants in the study interacted with one another and how they felt about those interactions. The research team used this data and the stochastic actor-oriented model, (SAOM) a framework used to analyze social networks and how they change over time, to map out participants’ social structures.
“Because this dataset was very unique and rich, we are able to ask a lot of interesting research questions related to how people’s experiences evolve over time and understand these outcomes from multiple perspectives, including what people said they perceived, and what people actually did on a nonverbal level,” Han said.
This analysis yielded three key trends. The first was that as the weeks went on, participants gradually disengaged from the activities and from each other. This pattern may have been caused by a variety of reasons, like the VR setting becoming less novel and exciting over time or the lack of incentives for engaging with their peers, but the exact reason is unknown and could be the catalyst for further research. The second pattern was that despite participants’ disengagement with the tasks and with the group at large, their smaller social groups persisted. Finally, participants were more likely to connect with others in the group that they had things in common with, a phenomenon known as homophily. Interestingly, among this group, participants felt closer with others based on shared interests and familiarity rather than on the basis of demographics, like gender or ethnicity.
“We found that class participants were more likely to connect with their existing peers or those similar to them, rather than to form new connections,” said Rinseo Park, a Stanford University Ph.D. candidate and the study’s lead author. “This finding calls for more open and inclusive learning environments that can naturally encourage students to bridge with unfamiliar others.”
The other big takeaway from their study was the effectiveness of the model they used in their data analysis. This study was the first of its kind to use the SAOM for this type of data, and their success with it proves that it can be applied to future studies. “We hope that this first application and demonstration of SAOMs to VR motion data will drive more discussion and empirical analysis of social networks in computer-mediated environments,” the team wrote.
