Unizin’s member institutions like Indiana University, the University of Iowa, and the University of Michigan use the Unizin Data Platform (UDP) to conduct research that allows them to better understand teaching and learning methodologies and practices at scale.
Unizin’s member institutions like Indiana University, the University of Iowa, and the University of Michigan use the Unizin Data Platform (UDP) to conduct research that allows them to better understand teaching and learning methodologies and practices at scale.
“Massive data in education will be game-changing, as it has been in other domains. The individual experience can be interrogated at scale.”
Ben Motz, Research Scientist, Department of Psychological and Brain Sciences and Director of the eLearning Research & Practice Laboratory, Indiana University
As Director of Indiana University’s eLearning Research & Practice Laboratory, Research Scientist Benjamin Motz pays special attention to how his students learn. But even when he experimented with different teaching methods he never had enough data to conclude that one strategy was uniformly better than another: his evidence was always limited to his own classes, in his own field, at his own university. Scientific validation requires showing that other researchers can reproduce the same results under the same conditions, with a robust sample size. That standard has often kept teachers from being able to generalize from their own experiences to establish pedagogical best practices.
Motz was enthusiastic when he heard that Indiana University was a founding member of Unizin, a non-profit consortium of higher education institutions dedicated to improving student success by focusing on data integration, learning analytics, digital content, research, and community at scale. He and his colleagues in the Department of Psychological and Brain Sciences set up a pilot project called ManyClasses, which aims to recruit teachers from across Unizin’s member institutions to collaborate on the same experimental protocol and test their teaching methods on a larger pool of student data. In a 2019 PsyArXiv preprint, Motz and co-authors Emily Fyfe, Joshua de Leeuw, Paulo Carvalho, and Robert Goldstone explain the ManyClasses approach: to “examine the same research question and measure the same experimental effect across many classes spanning a range of topics, institutions, teacher implementations, and student populations.”
To do this, the ManyClasses team leverages Unizin’s shared resources like the interoperable Unizin Common Data Model (UCDM) and the Unizin Data Platform (UDP). Built to be scalable and secure on Google Cloud, the UDP enables member institutions to easily share data and tools, while also managing technical support, security, regulatory compliance, and data transfer. Each member university maintains its own instance of UDP but by sharing a common data format researchers can collaborate more easily and test their hypotheses across larger sample sizes. Unizin’s member institutions together enroll nearly 800,000 students, making the UDP the largest, richest, and broadest collection of depersonalized learner data in higher education.
Motz and his colleagues enlisted 38 teachers in different fields from Indiana, University of Michigan, University of Nebraska at Lincoln, University of Minnesota, and Penn State University to participate in ManyClasses. Students volunteered their consent to have their depersonalized data included, and so far thousands of students enrolled in those courses have agreed to participate. The team started with a simple but important question about teacher behavior: what is the optimal time for students to get feedback on assignments? For example, standard practice encourages teachers to return test scores as soon as possible, but recent research suggests there might be benefits to delaying (e.g., Butler & Woodward, 2018).
For the pilot, “we used a crossover within-subjects randomized experimental design,” Motz explains. “We didn’t manipulate any course material or instructor behavior; all students got delayed and immediate feedback but in different orders so we could judge the effects on their final grades or learning.” The results of this first experiment are still being tabulated, but Motz is encouraged by the project’s potential and by the prospect for future work with Unizin. “It’s been spectacularly collaborative,” he reports. “The only way to proceed was through Unizin. They were especially accommodating and offered lots of solutions. It really helps that they have figured out many problems already, like complying with the Family Educational Rights and Privacy Act (FERPA). That makes security and privacy much easier.”
“Unizin will open a lot of possibilities with their de-identified dataset. We like to use data to make predictions, but our ultimate goal is to design interventions that change those predictions.”
Jae-Eun (Jane) Russell, Director of Office of Teaching, Learning, and Technology, University of Iowa
“The UDP and public cloud present interesting opportunities for us,” says Sean DeMonner, Executive Director of Teaching and Learning at U-M’s Information and Technology Services (ITS). “They offer foundational infrastructure, technical innovation, and data security at consortial scale.” Working with Google and Unizin, DeMonner gathered the team that developed My Learning Analytics (MyLA), a data analytics tool that visualizes course data and activities so students can monitor their own learning behaviors. John Johnston, the ITS Program Manager at U-M who acts as technical liaison with Unizin, explains that “it guides students to effectively budget their study time on the most impactful learning activities.” By visualizing resource utilization, grade distribution, and assignment planning, MyLA helps students develop what is called self-regulated learning: they become conscious of their behaviors and the outcomes that follow. Students can then use that knowledge to improve their habits to achieve better results in a virtuous cycle. To date, MyLA has been used by over 2,300 U-M students over three semesters in 32 courses across different disciplines—and it has been deployed at seven other Unizin institutions because they were able to leverage the common data backend.
For Jae-Eun (Jane) Russell, Director of Office of Teaching, Learning, and Technology at the University of Iowa, “one big benefit to our Unizin membership is that multi-institutional research is much easier. My job is to help faculty make data-informed decisions to support faculty teaching and research. To do that we want to consolidate and unify our online data and understand more about student engagement off line.” Russell’s own research focuses on student reading habits, which is hard to assess beyond self-report. By collecting de-identified student data on Engage, Unizin’s digital content platform, researchers can now measure students’ actual reading behaviors, like the number of pages students read and the reading tools they use, not just what they report afterwards to researchers. In fall 2019 forty faculty members participated and 3,800 students’ data were included in Russell’s study, which is still in process. Through the data collected from Engage on the Unizin Data Platform, professors can ask new questions about students’ online reading, like which reading strategies work best for which students and how often students use digital tools like bookmarking or key word searches.
These IT leaders and researchers all believe in the power of learning analytics to help teachers maximize student learning. “Massive data in education will be game-changing,” Motz says, “as it has been in other domains. The individual experience can be interrogated at scale.” Russell agrees that “Unizin will open a lot of possibilities with their de-identified dataset. We like to use data to make predictions, but our ultimate goal is to design interventions that change those predictions. That’s how we can support the faculty who support students.”
Learn more about how the Unizin Data Platform can support your institution.
For content more relevant to your region, we suggest: