Research Overview

Watch a recent seminar by Dr. Teasley, this research team’s Principal Investigator.

With the advancement of data science, we are facing unprecedented opportunities and challenges in understanding the effects of the vast number of learning tools, strategies and electronic learning systems on learning behavior and academic performance.  The challenges come from the availability of the enormous amount and types of data on learning behavior, interactions between learners and learning systems, and grades, test scores, competency measures as learning outcomes   The opportunities come from our increasing sophistication in processing and interpreting the vast types and amount of data.  The research team aims to bring revolutionary changes in our understanding of education and learning through the holistic modelling of learners, impact the development and deployment of effective learning tools and methods, and greatly enhance UM’s research capability in learning analytics.

The holistic models examine the relationship of learner behavior, learner interaction with learning systems, and outcomes, the interconnected semantics as knowledge representation, and data on learner activities, artifacts and achievements.  The models are generated by integrating multiple types of learning data and using a variety of data science approaches including inferential statistics, machine learned predictive models, and probabilistic semantic graphs.  The learning data used in these holistic models come from different disciplinary traditions, including psychological constructs such as affect and cognition, linguistic features expressed through text, learner interactions with computational systems and artifacts as well as peers and instructional experts.  The research team’s extensive experience on processing educational data and research on the effect of learning approaches on academic performance, together with the existing momentum at Michigan in learning analytics, will help fundamentally shift the focus of learning analytics at Michigan away from iterative technique-driven research towards multi- and interdisciplinary connected research.

Recent presentations

Research Team

  • Stephanie Teasley, Principal Investigator, Research Professor, School of Information
  • Christopher Brooks, Research Assistant Professor, School of Information
  • August Evrard, Arthur F. Thurnau Professor of Physics and Astronomy, Department of Physics
  • Tim McKay, Arthur F. Thurnau Professor of Physics, Astronomy, and Education, Department of Physics
  • Perry Samson, Professor, School of Information
  • Kevyn Collins-Thompson, Associate Professor, School of Information
  • Anne Gere, Arthur F. Thurnau Professor, Department of English

Updates

February 2018

  • The team has expanded its study by including an understanding of how learner population changes with time in a large learning environment, and an additional research focus on how data science students learn.
  • The team is developing a model and running a field experiment to understand how students respond to feedback from reviewers during learning so as to determine the best approach to match students with reviewers.
  • The team has worked in a new subarea of research that explores how models of human learning can be integrated into search engine algorithms. The impact includes the introduction of a new type of search engine that can provide material that is optimized for highly personalized learning, accounting for both user expertise and the required learning effort.
  • The team has used the ECoach platform to conduct a large random controlled trial, in which student writings in a number of STEM courses is being analyzed to provide Natural Language Processing-based information. Such information is to be used in real-time to tailor responses to student submissions. The results of this text analysis will also be combined with student behavior data to predict student achievements.
  • ECoach, a personalized learning tool that provides individualized messages, information and coaching, as well as collects learning data, was developed by one of the team members, Tim McKay, and is now deployed at UC Santa Barbara.

Funding

  • McKay et al.  NSF grant (1625397) “Addressing Ubiquitous STEM Gender Performance Differences”
    This grant utilizes the holistic model of learning being developed by HOME, including efforts to incorporate Natural Language Processing.

Previous Updates

July 2017

The team has been experimenting with data virtualization infrastructure for merging datasets across disparate sources.

In the past year, various members of the team have given a total of 18 invited talks and conference presentations, including:

  • National Science Foundation EHR and CISE Director’s Distinguished Lecture
  • Dean’s Distinguished Lecture: University of Buffalo Graduate School of Education
  • Expert panelist at the Fourth Annual International Conference on Education & Global Cities: Smart Learning for World Universities.
  • Panel presentation at The White House Symposium on the Future of Education R&D and Digital Learning
  • 7th International Conference on Learning Analytics and Knowledge
  • 10th International Conference on Educational Data Mining