Classinsight
Fig. 1: Example of Unit View
ClassInSight is a computer-vision based, real-time analytics platform for classrooms.
The goal is to support teachers in making more informed decisions about how to improve their own teaching practice, by visualizing data from their instruction. Attending to every feature of the classroom is difficult for beginning teachers, and such a system can help provide the type of data that a human observer might be able to provide at rare intervals and with much greater cost.
This training system uses classroom sensor data to assist instructors in learning best practices and goal-setting, leading to overall improvement of teaching performance.
My research experience through the Summer REU (Research Experience for Undergraduates) Program equipped me with valuable knowledge of how to work in a lab setting within several interdisciplinary teams. Conducting research, from designing an experiment to communicating ideas with peers and faculty, is the most challenging yet fulfilling work I’ve been a part of. I continued my research in the fall by expanding on the study design, improving the existing sensor-based training system, and presented my work at Carnegie Mellon’s research symposium, Meeting of the Minds.
Fig. 2: Testing the facial recognition algorithm using Kinect sensor system
Fig. 3: Presenting our work at Meeting of the Minds in May 2018
Observational Study
I helped to design, conduct, and improve upon the observational study for implementing a dashboard in STEM courses for higher education teaching assistants. Using design-based research methods and qualitative data analysis, we introduced a sensor-based feedback and training system to guide higher education instructors to reflect on their teaching practices.
My personal contribution to the project, after helping design our study, was to collect, interpret, analyze, and visualize data from our classroom sensor system, then develop a curriculum for our participants, the CMU instructors during the summer session, and discuss our initial findings from the study to produce rich insights.
My contribution in the fall semester included developing the curriculum through creating the content of the instruction, as well as designing and implementing the delivery of the curriculum. The online instructional platform targeted goal-setting for professional development skills, such as asking deeper questions and increasing student participation. The process involved multiple iterations on designs for visualizing feedback and providing instruction to teachers, and background research in tailored instruction, adaptive control, and other instructional design principles.
Data Visualizations
For the teacher dashboard, we created meaningful and understandable visualizations of teacher data to prompt reflection.
Fig. 4.1: Question Count Visualization
Fig. 4.2: Question Count Visualization
Research Methods
Personal Informatics and Designing Technology for Behavior Change
Design-Based Research
Learning Science Principles
Fig. 5: Final Poster for Meeting of the Minds in May 2018