PERC 2016 Abstract Detail Page
Previous Page | New Search | Browse All
| Abstract Title: | Network Analysis in Physics Education Research |
|---|---|
| Abstract: | This poster symposium showcases recent applications of network analysis in physics and science education research. Approaches include a focus on individual students (ego networks), a time-evolving range of interaction types for an entire class (multiplex networks), and combinations with other tools such as text mining and discourse analysis. Social network analysis can complement qualitative methods in exploring the rich complexity of student participation, and is especially valuable for capturing the many student-student interactions that are an integral part of active learning environments. Other applications of network analysis, rather than approximating a social landscape, uncover hidden connections and themes in dense bodies of text such as interview transcripts. The research in this session highlights a variety of ways that the network data perspective extends the existing methodological toolbox of physics education research. |
| Abstract Type: | Poster Symposium |
Author/Organizer Information | |
| Primary Contact: |
Adrienne Traxler Wright State University 3640 Colonel Glenn Hwy Department of Physics Dayton, OH 45435 |
| Co-Author(s) and Co-Presenter(s) |
Jesper Bruun |
Parallel Session Information | |
| Activity Description: | Poster presenters will give short presentations about their work, followed by discussion, then moving to the next poster. An overall discussion of connections or common themes will follow after the set of presentations. |
| Discussion Panel Members: | Presenters (besides organizers): Madelen Bodin, Eric Brewe, Mats Lindahl |
| Proposed Discussion Questions: | Posters will show different applications of network analysis tools to physics education research problems. |
Symposium Specific Information | |
| Moderator: | Adrienne Traxler |
| Presentation 1 Title: | Students' progress during an assignment in computational physics: mental models and code development |
| Presentation 1 Authors: | Madelen Bodin |
| Presentation 1 Abstract: | Solving physics problems in university physics education using a numerical approach requires knowledge and skills in several domains, for example, physics, mathematics, programming, and modeling. In this study students' mental models are monitored using interviews at several occasions during an assignment in computational physics. The interview data was analysed using a network analysis approach. Interview transcripts were coded according to the context dependent concepts that were used to define the particular context and situation of this assignment. The adjacency of concepts in the transcripts was assumed to reflect the associations between them made by students, and thus representing students' mental models of the problem solving situation at the time of the interview. For each student a network was built where the concepts were nodes and their adjacency formed the links between them. The changes in students' mental models between the interview occasions gave important information about what the students were focusing on at different stages of the solution process. What students focused on at the different interview occasions was assumed to be an indication of what they believed was useful in solving the task. The visualization of the mental models showed that at the beginning students were concerned about how to deal with writing the Matlab code that was needed to model the problem. As students got more comfortable with the coding process, the physics needed to assure that their simulation was following physics principles, such as energy conservation, became more and more central in their narratives. This study gives important contribution to how networks can be used to model students' thinking in a particular context and provides important knowledge about students' progress in a task in computational physics. |
| Presentation 2 Title: | Ego network analysis of upper division physics student survey |
| Presentation 2 Authors: | Eric Brewe, Justyna Zwolak, Remy Dou, and Eric Williams |
| Presentation 2 Abstract: | We present the analysis of ego networks derived from a survey of upper division physics students. Analysis of ego networks are somewhat different than network analyses that are becoming more common in PER. Ego networks focus on the connections that center on one person (the ego). The ego networks in this poster come from a survey that is part of an overall project focused on understanding student retention and persistence. The theory underlying this work is that social and academic integration are essential components to supporting students continued enrollment and ultimately graduation. This work uses network analysis as a way to investigate the role of social and academic interactions in retention and persistence decisions. We focus on student interactions with peers, on mentoring interactions with physics department faculty, and on engagement in physics groups and how they influence persistence. Our results, which are preliminary, will help frame the ongoing research project and identify ways in which departments can support students. |
| Presentation 3 Title: | Investigating physics learning with layered student interaction networks: Combining time and modes of interaction |
| Presentation 3 Authors: | Jesper Bruun |
| Presentation 3 Abstract: | Network analysis has previously been employed to show that centrality in student interaction networks can predict future grades. However, previous analyses have only been employed on isolated categories of student interactions which have been summed over time. In this study we use multiplex (layered) networks to relax these constraints. Within a week, we now allow bidirectional links between the same student as represented in two different layers, for example a layer representing problem solving interactions and another representing conceptual discussion, to model how these layers may interact. Likewise, we allow links from a student as represented in a network at time, t, and the same student in the network for time, t+1. The nature of links between network layers is different from the original links and thus need a theoretically based model. These models may yield insights into the interplay between categories of interaction into the significance of time development. |
| Presentation 4 Title: | Integrating text-mining, network analysis and thematic discourse analysis to produce maps of student discussions about sustainability |
| Presentation 4 Authors: | Mats Lindahl, Jesper Bruun, and Cedric Linder |
| Presentation 4 Abstract: | We use a combination of network analysis (NA), text-mining (TM) techniques, and thematic discourse analysis (TDA) to characterise and compare student discussions about sustainable development. Three student groups at three different times were analysed. The analysis entails an iterative design where NA, TM, and TDA continuously inform each other to produce a rich and coherent picture of the discussions. The output of such an analysis is a set of maps of these discussions, which have both qualitative and quantitative uses. Qualitatively, the maps show how thematic patterns in the discussions are related for each group, and we can see how discourses differ between groups as well as over time. Quantitatively, we use network motif analysis, entropy based measures, and degree distributions to distinguish between discussions. |
| Presentation 5 Title: | CourseNetworking and community: Linking online discussion networks and course success |
| Presentation 5 Authors: | Adrienne L. Traxler, Andy Gavrin, and Rebecca Lindell |
| Presentation 5 Abstract: | We report on connections between student participation in an online course forum and their grades in an introductory physics class. Discussions started before the semester, ended after finals, and continued well beyond the level incentivized by extra credit. We collect all posts and replies and construct a bipartite network of actors (students) and events (threads) in the forum. The actor projection of this two-mode network is a weighted representation of students' mutual participation in discussion threads, and shows whether they are central or peripheral participants in the semester-long discussion community. We look for significant links between network centrality and final grades, as part of a larger investigation of the CourseNetworking software as an online community-building tool. These questions are especially relevant for institutions with many commuter and non-traditional students, where asynchronous forum talk is a key way for students to engage with their peers outside of class time. |
