Student Engagement Tracker (SET)
Overview
The Student Engagement Tracker (SET) is a human-centered AI-powered learning analytics platform designed to help instructors better understand student engagement in synchronous online learning. During development, I focused on improving the instructor experience by redesigning complex workflows, simplifying system interactions, and creating an analytics dashboard that transforms Zoom meeting data into actionable insights for instructional decision-making.
The Challenge
Learning Problem
Online instructors often struggle to identify disengaged students because meaningful engagement cues are distributed across video, chat, transcripts, and participation data.
User Experience Problem
The original prototype required instructors to perform numerous setup steps before each class, including manually starting data collection and coordinating with a teaching assistant to operate the system. These workflow barriers significantly limited usability and adoption.
Phase 1 Empathize
Primary Users
University and K-12 instructors teaching synchronous online courses.
Understanding the Original Design Process
Identifying Opportunities
After analyzing the instructor workflow and the existing system, I identified several opportunities to improve both usability and instructional value.
1: High Setup Complexity
The original workflow required instructors to complete multiple setup tasks before each class, including creating classes, scheduling meetings through the portal, and coordinating with a teaching assistant. These extra steps increased the effort required to use the system and created barriers to adoption.
2: Dependence on Manual Operation
The system relied on a teaching assistant to manually launch the Zoom bot before every meeting. This dependency limited scalability and made the tool impractical for instructors teaching independently.
3: Analytics Did Not Match Instructor Needs
The most important insight came from examining what instructors actually wanted to know.
The original system primarily tracked real-time emoji reactions, assuming they represented student engagement. However, emoji reactions provided only a small part of the classroom picture and were rarely used by instructors to guide instructional decisions.
Instead, instructors were much more interested in understanding meaningful classroom interactions, such as:
-
Who actively participated in class discussions?
-
How often did each student contribute through voice or chat?
-
How many questions did students ask?
-
How often did students answer questions?
-
Which students participated consistently, and which remained silent?
-
How did participation patterns change across class sessions?
These insights are far more actionable than emoji counts because they help instructors identify engagement patterns and make informed instructional decisions.
Phase 2 Design
These findings shifted the focus of the project from simply collecting engagement data to supporting instructor decision-making.
The redesign focused on three goals:
-
Reduce instructor workload by simplifying the setup process.
-
Eliminate unnecessary dependencies by supporting asynchronous analysis through recorded Zoom uploads.
-
Provide meaningful learning analytics by emphasizing participation, questioning behavior, and discussion patterns rather than relying primarily on emoji reactions.
Key Design Insight: The challenge was not collecting more data—it was identifying and presenting the data that instructors could actually use to improve teaching.
Phase 3 Build
New Upload Workflow
Instead of requiring live monitoring, instructors can simply upload a recorded Zoom session directly through the portal.
This redesign:
-
removes dependence on teaching assistants
-
supports asynchronous analysis
-
simplifies onboarding
-
reduces instructor workload
Transcript Analysis Dashboard
I designed a new analytics module that extends the original chat and emoji analytics with transcript-based participation metrics.
This provides instructors with more meaningful insights into classroom interaction patterns.
Results Dashboard
Rather than presenting raw transcripts, the new dashboard summarizes classroom interaction into interpretable metrics including:
-
speaker turns
-
word counts
-
questions
-
responses
This allows instructors to quickly identify participation patterns.
Outcomes
Although, due to the funding issues, the project concluded before full deployment, several key improvements were successfully designed and implemented:
Workflow Improvements
-
Redesigned the instructor workflow to reduce setup complexity.
-
Added an asynchronous Zoom meeting upload feature, allowing instructors to analyze recorded sessions without relying on live system operation.
-
Eliminated the need for a teaching assistant to manually launch the Zoom bot, making the platform more practical and scalable for independent instructors.
Learning Analytics Enhancements
-
Expanded the analytics capabilities by introducing Transcript Analysis alongside the existing chat message and emoji reaction analytics.
-
Designed a dashboard that summarizes classroom discourse using metrics such as speaker turns, word counts, questions asked, and questions answered.
-
Shifted the focus from isolated engagement indicators to more meaningful measures of classroom participation that better support instructional decision-making.
Design Impact
-
This project transformed the platform from a technically functional data collection tool into a more instructor-centered learning analytics experience by reducing operational barriers and presenting information that aligns more closely with instructors' instructional needs.
Technology Skills
C++, JavaScript, AI-assisted coding, Virtual Private Server (VPS), User-centered design, UX workflow redesign, Learning analytics dashboard design, Information architecture, and Requirements analysis
What I Learned
At the beginning of this project, I assumed the primary challenge was building an AI system capable of collecting and analyzing student engagement data. However, working closely with the existing workflow shifted my perspective.
I realized that the success of an educational technology depends not only on the quality of its analytics but also on how easily instructors can integrate it into their everyday teaching practices. Even a sophisticated AI system provides limited value if it requires complex setup procedures, additional personnel, or presents information that does not directly support instructional decisions.
This realization changed my design approach. Instead of asking, "What data can the system collect?" I began asking, "What information do instructors actually need, and how can it be delivered with the least amount of effort?"
That shift led to redesigning the workflow, supporting asynchronous meeting analysis, and expanding the analytics to emphasize meaningful classroom interactions rather than simply reporting available engagement signals.
This experience reinforced a design principle that continues to guide my work:
Designing educational technology is not just about building powerful AI models—it is about creating tools that fit naturally into instructors' workflows and provide actionable insights that support better teaching decisions.