# Workshop Participants

## Group 1: Using sequence analysis and optimal matching to analyze classroom-based video data

Elizabeth McEneaney is a quantitative sociologist working in education who has substantial training in statistical modeling of change over time, including econometric methods, survival analysis and event history analysis. With Martina Nieswandt she is investigating methods to explore large amounts of video data, using data from ~400 hours of small group interaction on science problems, to track temporal temporal sequence in two ways: 1) the trajectory of dynamics in these fixed-membership groups over all 6 tasks over the course of the academic year, and 2) the trajectory of dynamics within a single task. Principles of sequence analysis and optimal matching (Abbott & Hyrcak 1990; MacIndoe & Abbott 2004; Gabadinho, et al. 2011) are applied to the video data using R.

- Ugochi Acholonu
- Alejandro Andrade
- Anne Boyer
- Tsan-Kuang Lee
- Elizabeth McCearney

## Group 2: Using Hidden Markov Models to explore reasoning and behaviour in log data

One of the central theoretical issues that must be addressed in learning analytics, and in particular when conducting and interpreting temporal analyses, is the association of sequences of micro-events or actions with aggregate constructs (e.g., strategies or behaviors) that are relevant to learning, instruction or behavioral frameworks. One approach that can help researchers explore those associations is Hidden Markov Modeling (HMM), which represents probabilistic models of sequences of events as potential ‘states’ of reasoning or behavior. Britte Cheng will discuss her use of HMM in exploring group problem solving in a virtual environment, work with group members to consider how HMM might be useful in their analytic work, and where possible, support modeling of group members’ data during the workshop.

- Sherry Davis
- Phil Winne
- Scott Harrison
- Andrew Gibson
- Britte Cheng

## Group 3: Statistical Discourse Analysis applied to f2f turn-taking data

Ming Ming CHIU (Purdue University) invented statistical discourse analysis to analyze (a) sequences of group or individual events/processes, (b) relationships between processes and outcomes and (c) how these processes, outcomes and relationships evolve over time. SDA can be applied to any time-series data at the macro- or micro-levels simultaneously (e.g., chat logs, teacher diaries, eye-tracking, etc.); for example, teacher reflections, tutoring processes, group problem solving, teachers analyzing classroom videos, pilots’ flight simulations, etc.

- Ming Chiu
- Pablo Garaizar
- Arnon Hershkovitz
- Agathe Merceron
- Inge Molenaar
- Wanli Xing

## Group 4: Temporal patterns in assessing collaborative learning on wikis in secondary and primary schools

Xiao Hu’s research explores the temporal dimension of Wikis, which store page revision histories and student discussions alongside the final page versions. Social network analysis is used to measure interactions (co-comments or co-edits on the same page) between group members over time. Sequential patterns are mined from revision histories of each page, including the type of edit (addition, deletion), size (n of words), edit frequency, etc. Methods used to explore temporal analysis of online discussion can be applied to this data.

- Armelle Brun
- Xiao Hu
- Sherwyn Saul
- Shady Shehata
- Anna Smith
- Jennifer Pei-Ling Tan

## Group 5: Epistemic Network Analysis to understand trajectories of development

Golnaz Arastoopour and Wesley Collier work has involved development of the ENA tool. ENA is a network analysis tool that constructs comparative models of weighted networks, including specifically analyzing how a network or networks change over time. ENA characterizes networks in terms of the structure of the weighted connections and represents that structure in three coordinated representations: (1) a network graph showing the strength of connections for any single network; (2) a set of summary statistics through which change in the structure of connections over time can be modeled; and (3) a trajectory model that compares the temporal progression of one network to the progression of other networks, including analyses that compare the temporal progression of one network to the expected trajectory of change over time for some group of networks. ENA can model any networks with a fixed set of nodes (including nodes that have no connections), which might include sociograms of interactions in a classroom that change over time, or conceptual (ie, epistemic) networks of individuals that change during learning or other activities. Moving forward we are particularly interested in understanding how to construct visualizations of ENA models that are easier to interpret, for researchers but especially models that teachers or other practitioners could use to interpret trajectories of development among students over time.

- Golnaz Arastoopour
- Wesley Collier
- Jan Arild Dolonen
- Leah Macfadyen
- Eni Mustafaraj
- Tim Vogelsang

## CANCELLED: Contingency, uptake and temporal-sessions - analytics & representational tools

Dan Suthers is currently exploring a “Traces” approach to analysis of event data from socio-technical systems. Traces comprises a way of conceptualizing interaction, mediation and ties and associated analytic representations (Suthers, Dwyer, Medina & Vatrapu, 2010; Suthers & Rosen, 2011), and an evolving software suite for constructing higher level analytic structures from streams of event data (Suthers, 2015; Suthers & Dwyer, 2015). Logs of events are abstracted and merged into a single abstract transcript of events, and this is used to derive a series of representations that support levels of analysis of interaction and of ties. Three kinds of graphs model interaction: Contingency graphs record how events such as chatting or posting a message are observably related to prior events by temporal and spatial proximity and by content. Uptake graphs aggregate the multiple contingencies between each pair of events to model how each given act may be “taking up” prior acts. Session graphs are abstractions of uptake graphs: they cluster events into spatiotemporal sessions with uptake relationships between sessions. Relationships between actors and artifacts are abstracted from interaction graphs to obtain associograms, which can be folded into traditional sociograms. This is being applied to the Tapped In data corpus (Dan is focussing on two years of data in which 20,000 active user accounts for educational professionals interacted via chats, threaded discussion, and file sharing in both formal and informal settings.)