|
The following article originally appeared in the 2020 edition of the School of Education magazine, advancing. For a PDF, please click on the magazine to the right. |
|
Role Model
Assistant Professor Nia Dowell created a software that analyzes conversations and identifies the different emergent roles individuals are playing in a group. By doing so, users can improve group and individual performance, address inequities, and more.
|
Different people assume different roles when working in a group. There are leaders and followers; both loquacious and soft-spoken individuals; and those whose words spark inspiration and those whose words are quickly forgotten.
What if there was a way to quickly and automatically identify the roles of different group members, so that individual and group performance could be improved? Better yet – what if it could also be applied to online settings, where body language is non-existent, and interactions are typically more strained? School of Education Assistant Professor Nia Dowell developed a software that does just that. “Roles are one of the most important concepts in understanding human socio-cognitive behavior,” Dowell said. “The software allows us to understand the underlying human socio-cognitive process. We can also understand and differentiate why certain groups and teams perform better than others and facilitate ways to improve performance.” |
Assistant Professor Nia Dowell
|
HOW IT WORKS
|
The software, Group Communication Analysis (GCA), reads transcripts of conversations – be it a Zoom call, an in-person meeting, an online chat room or Twitter feed. It then begins to code not only user contributions, but also how the rest of the group replies to a user’s contributions.
The GCA analyzes conversations on six metrics, or dimensions – general participation; consistency or novelty in contributions; responsivity (or not) to their peers contributions; ability to evoke replies from peers; originality in their statements; and the level of conciseness in their contributions (see Figure 1). From these, the GCA identifies individuals as performing one of six roles: Chatterers, Drivers, Followers, Lurkers, Socially Detached and Influential Actors (see Figure 2). For example, the socially detached individual contributes frequently to a group dynamic, but the social impact of their comments are low. These individuals are usually dominating the conversation, but not in a productive way, nor are they attending to their partners or teammates. Meanwhile, a driver is responding to individuals while moving the discussion forward and having his or her words well-received by others. |
Figure 1: The six metrics of the GCA
|
“You identify the role a person is playing not only by their engagement, but by how others are responding to their engagement,” Dowell said. “Through all this, we start to capture the basic fundamentals of human interaction and can create profiles and analyses to understand what role a person is playing, and what impact they will ultimately have on the group.”
USING THE DATA
|
There are myriad ways that the software, which is patent pending, can be used, and how the data can be interpreted.
Dowell currently uses GCA on collaborative problem solving and collaborative learning environments in academia, and recently began using GCA on chat forums for both Massive Open Online Courses and learning management systems, such as Canvas. The findings from a given analysis can lead teams to implement minor tweaks. For example, creative solutions and ideas come from groups of individuals who share novel information, not simply what is already known to the group. If a group struggles with “newness,” then they can work to implement procedures so that its members share novel thoughts more frequently. |
Figure 2: Centroids for the six-cluster solution across the GCA variables.
|
The GCA can also identify larger societal issues in group dynamics, such as threats to inclusivity and equity. Dowell has used GCA to gain a deeper understanding of the communication dynamics in online team interactions across gender and ethnic lines.
For instance, Dowell uncovered differences in learners’ interpersonal interaction patterns across ethnic populations, between male and female students, and the influence of gender group composition on equitable interpersonal discourse during STEM interactions. Across these studies, Dowell discovered substantial intra- and interpersonal differences in women and underrepresented minorities’ engagement, which could influence their sense of belonging in online STEM environments.
As depicted in Figure 3, Dowell and others discovered large differences in the sociocognitive interaction dynamics between male and female learners in STEM online team interactions.
Interestingly, the observed differences between males and females is not in their degree of participation, but in three other areas – the extent to which they: engage in productive discourse that is responsive to other learners (i.e., overall responsivity), provide meaningful contributions that warrant follow-up by peers (i.e., social impact), and monitor and build on their own previous contributions over the course of interaction.
Interactions such as this can have a detrimental impact on a female learners’ sense of belonging within STEM fields, and consequently contribute to the existing retention issues.
“The GCA allowed us to examine a group setting broadly, from which we found a threat to female students in STEM,” Dowell said. “Now, we can proceed to devise an inclusion-focused intervention.”
For instance, Dowell uncovered differences in learners’ interpersonal interaction patterns across ethnic populations, between male and female students, and the influence of gender group composition on equitable interpersonal discourse during STEM interactions. Across these studies, Dowell discovered substantial intra- and interpersonal differences in women and underrepresented minorities’ engagement, which could influence their sense of belonging in online STEM environments.
As depicted in Figure 3, Dowell and others discovered large differences in the sociocognitive interaction dynamics between male and female learners in STEM online team interactions.
Interestingly, the observed differences between males and females is not in their degree of participation, but in three other areas – the extent to which they: engage in productive discourse that is responsive to other learners (i.e., overall responsivity), provide meaningful contributions that warrant follow-up by peers (i.e., social impact), and monitor and build on their own previous contributions over the course of interaction.
Interactions such as this can have a detrimental impact on a female learners’ sense of belonging within STEM fields, and consequently contribute to the existing retention issues.
“The GCA allowed us to examine a group setting broadly, from which we found a threat to female students in STEM,” Dowell said. “Now, we can proceed to devise an inclusion-focused intervention.”
|
GCA has drawn interest from several companies: everything from a call center who would like to improve customer interactions to the U.S. Military, who are looking to analyze operational conversations in the field.
“This can be used in many ways beyond what we’re already doing,” Dowell said. “The GCA can be applied to any formal or informal situation or conversation where individuals are coming together to solve a problem.” While it is necessary to have a text transcript of an interaction for GCA to work, technological advances are making this an impending reality. Today, transcripts of Zoom calls are readily available. Dowell also sees speech-to-text technology as a way to instantly deliver a written transcript of a meeting – physical or virtual – to the GCA. Over the past few months, conversations have moved to a virtual space due to the widespread adoption of remote instruction and telecommuting. As a result, Dowell said the software has taken on heightened importance. |
Figure 3: Gender differences across GCA dimensions.
Compared to females, male learners were less
responsive to their teammate’s contributions
(i.e., overall responsivity), provided less meaningful
contributions that warrant follow-up by peers
(i.e., social impact), and engaged in less monitoring
and building on their own previous contributions over
the course of the interaction (i.e., internal cohesion).
|
“The pandemic is highlighting what was already an important movement in our society,” Dowell said. “Every day, even before COVID-19, we engaged in online interactions. Optimizing those dynamics was already important, but COVID-19 underscores that.”
LEARNING THE LANGUAGE
Dowell attended the University of Memphis, where she earned a B.A. in Psychology and a Ph.D. in Cognitive Psychology with a Cognitive Science Certificate. In lab work at Memphis, she worked on affect detection and its implications on learning – how to tell through pupil dilation and physiological factors if a student is upset, uncomfortable, bored, or honed in.
Dowell soon became interested in natural language processing tools. At the time, these tools were limited to individual measures – there was nothing available that captured the temporal dynamics or back-and-forth nature of conversation, which Dowell wanted to quantify.
“I started thinking about what is important in conversation, and I reviewed all the literature out there, from cognitive and organizational psychology to social computing,” Dowell said. “I came away with a very interdisciplinary understanding of conversation and socio-cognitive roles individuals take on during human interactions.”
In addition to her research and scholarship, Dowell provided text analysis consulting to several organizations, including FedEx and the U.S. Department of Defense. Dowell found this work very exciting and these diverse experiences helped to broaden her interdisciplinary perspective on language and discourse. For instance, at the latter, Dowell used natural language processing tools to help predict and prepare for events during social movements, such as the Arab Spring and authoritarian regimes more broadly.
Dowell joined the UCI School of Education in summer 2019. She currently directs the Language and Learning Analytics Lab, an interdisciplinary group of students and faculty who explore the intersections of technology with teaching, learning, and education, with a particular focus on learning analytics, educational data mining, and collaborative engagement. The lab has several research projects underway: Implementing Agile Methodology to academic research; analyzing how group dynamics differ between cultures; and examining the sense of student belonging during team interactions within UCI STEM courses.
In spring, Dowell was elected to the executive committee for the Society of Learning Analytics Research (SoLAR). She is also serving as program chair for the Society’s annual Learning Analytics and Knowledge (LAK) conference, LAK21, scheduled to take place in Newport Beach, April 11-15, 2021.
Dowell teaches graduate and undergraduate courses at the School of Education. She hopes to follow the examples set by her Ph.D. Advisor, Dr. Arthur C. Graesser – whom Dowell calls “the best human being I ever met.”
“He never gave me the answers, and instead gave me the freedom to explore on my own. At the same time, he gave me all the resources I needed to succeed,” Dowell said. “I really hope that I can do that for my students in my career.”
Dowell soon became interested in natural language processing tools. At the time, these tools were limited to individual measures – there was nothing available that captured the temporal dynamics or back-and-forth nature of conversation, which Dowell wanted to quantify.
“I started thinking about what is important in conversation, and I reviewed all the literature out there, from cognitive and organizational psychology to social computing,” Dowell said. “I came away with a very interdisciplinary understanding of conversation and socio-cognitive roles individuals take on during human interactions.”
In addition to her research and scholarship, Dowell provided text analysis consulting to several organizations, including FedEx and the U.S. Department of Defense. Dowell found this work very exciting and these diverse experiences helped to broaden her interdisciplinary perspective on language and discourse. For instance, at the latter, Dowell used natural language processing tools to help predict and prepare for events during social movements, such as the Arab Spring and authoritarian regimes more broadly.
Dowell joined the UCI School of Education in summer 2019. She currently directs the Language and Learning Analytics Lab, an interdisciplinary group of students and faculty who explore the intersections of technology with teaching, learning, and education, with a particular focus on learning analytics, educational data mining, and collaborative engagement. The lab has several research projects underway: Implementing Agile Methodology to academic research; analyzing how group dynamics differ between cultures; and examining the sense of student belonging during team interactions within UCI STEM courses.
In spring, Dowell was elected to the executive committee for the Society of Learning Analytics Research (SoLAR). She is also serving as program chair for the Society’s annual Learning Analytics and Knowledge (LAK) conference, LAK21, scheduled to take place in Newport Beach, April 11-15, 2021.
Dowell teaches graduate and undergraduate courses at the School of Education. She hopes to follow the examples set by her Ph.D. Advisor, Dr. Arthur C. Graesser – whom Dowell calls “the best human being I ever met.”
“He never gave me the answers, and instead gave me the freedom to explore on my own. At the same time, he gave me all the resources I needed to succeed,” Dowell said. “I really hope that I can do that for my students in my career.”
