Clustering and Graphical Approaches to Examine Diversity

Return to search results | New search

Title of Abstract: Clustering and Graphical Approaches to Examine Diversity

Name of Author: Mark Grimes
Author Company or Institution: University of Montana
Author Title: Associate Professor
PULSE Fellow: No
Applicable Courses: Bioinformatics, Cell Biology
Course Levels: Introductory Course(s)
Approaches: Adding to the literature on how people learn, Assessment, Changes in Classroom Approach (flipped classroom, clickers, POGIL, etc.)
Keywords: learning exploratory data analysis clustering learning analytics education data mining

Goals and intended outcomes of the project or effort, in the context of the Vision and Change report and recommendations: In a classroom situation with one teacher and many students, it is widely recognized that a one-size-fits-all approach is not effective for the majority of students. The use of learning activities and supplemental online materials is an attempt to engage more students in different ways than is achieved by the standard lecture-and-regurgitation model. Yet we don’t know very much about how activities are used by different groups of students, or how different activities may lead to learning gains for students with different approaches to learning. We now have the ability to gather large amounts of data to ask questions about diversity in patterns of student learning. The ability to gather such data gives rise to two challenges: first, to detect patterns in the data; and second, to make the results accessible to instructors who wish to help students succeed using the approach to learning that works best for them.

Describe the methods and strategies that you are using: We address the first challenge using state-of-the-art pattern recognition methods applied to education data to uncover relationships between student learning, participation in activities, and demographic data that have not been previously resolved. Data-driven clustering by pattern recognition algorithms helps us understand diversity because patterns in all parameters are compared simultaneously, so links among them are revealed that would be less likely to emerge by manual sorting of data. We address the second challenge using novel graphical approaches to summarize and visualize the results of exploratory data analysis. The goal is to make the results accessible to a wide audience of teachers to guide further development of teaching methods, which will in turn reach a wider variety of students. We hypothesize that instructors in STEM disciplines will be more likely to appreciate a graphical approach even if they do not have previous experience with the underpinning pattern recognition techniques because data analysis and interpretation of graphs is part of all STEM disciplines.

Describe the evaluation methods that you used (or intended to use) to determine whether the project or effort achieved the desired goals and outcomes: There is a clear need to bring together the people who develop algorithms for data analysis and the people who are engaged in instruction. An important goal of the project is to produce informative graphs to help teachers make decisions for appropriate intervention strategies for particular groups of students. The graphical approach is likely to succeed with instructors in STEM disciplines who are trained in the interpretation of data. The evaluation of the graphical approach proposed in this study by biology instructors who are not well-versed in pattern recognition algorithms will guide development of software with a user-friendly interface that may be used by teachers in different settings. We hope to make the results of sophisticated pattern recognition algorithms accessible to teachers who will use the results to tailor teaching methods to different students with diverse learning approaches.

Impacts of project or effort on students, fellow faculty, department or institution. If no time to have an impact, anticipated impacts: The practice of using data, and the analysis of data, to motivate actions is a discipline that is essential in a democratic society. Therefore, skills in interpretation of graphs that are based on data are essential for teachers and students alike. Teachers who evaluate their own classroom data will be able to share the results with their students, which will help them learn, and also expose them to graphical displays of data. Promulgation of data-driven approaches to problem solving is necessary to address social issues, such as educational approaches to addressing the needs of underrepresented minorities in STEM disciplines, as well as scientific problems, such as how to glean useful information from very large data sets.

Describe any unexpected challenges you encountered and your methods for dealing with them: Integration of pattern recognition algorithms and graphical visualization augments the creativity of human brains to sort and filter data to gain insight in relationships between actions and outcomes. As stated above, the goal is to bring cutting-edge tools to instructors who can appreciate graphs but who may not wish to delve into the technical details behind the algorithms that generate the graphs. The challenge is to explain the underpinning techniques well enough so that the graphs can be interpreted.

Describe your completed dissemination activities and your plans for continuing dissemination: Students will be informed that there are a number of different ways to succeed. The data reinforce the notion that a wide variety of study strategies may be successfully used by students, and motivate us to provide a number of different resources to help students who learn using different strategies. The danger of providing too many resources is that students will be overwhelmed and lightly sample activities ineffectively. Thus, it will be important to inform students about individual strategies that work specifically for different groups, so that they may explore avenues that are likely to be right for them. Data from the previous class will be used to reinforce general suggestions that students read the book before class, and use online resources in a timely manner. For students who ask for ways to improve their performance, their current study habits may be assessed and alternative strategies suggested based on successful strategies that emerge from the analysis of clusters. Importantly, the detailed methods and conclusions will be described in a peer-reviewed publication so that it may be rigorously evaluated.

Acknowledgements: Collaborators on computational biology and bioinformatics projects include Gary Bader (University of Toronto); Paul Shannon (Fred Hutchison Cancer Research Institute); and in pattern recognition, Wan-Jui Lee and Laurens van der Maaten (Delft University of Technology). Collaborators on the biology education research project are Paula Lemons (University of Georgia, Athens); David Terry (Alfred University, New York); and Clyde F. Herreid (State University of New York, Buffalo).