We followed an agile methodology across multiple sprints. Key phases included:

• • Initial stakeholder interviews

  • Creating personas and mockups

  • Defining MVP through user stories

  • Building a MERN stack application with Docker

  • Conducting accessibility and ethics reviews

  • Iterating based on feedback

We continually refined CAVIAIR to meet usability, privacy, and performance needs for real Canvas integration.

• MongoDB, Express, React, Node (MERN Stack): Core framework for backend, frontend, and data management.

• Canvas API: Used to fetch courses, assignments, and submissions; also to post comments.

• AI Detection APIs: GPTZero, Copyleaks, Sapling for multi-source AI assessment.

• Docker & Docker Compose: Streamlined deployment into microservices (frontend, backend, detection queue, database).

• Figma: Designed low-, mid-, and high-fidelity UI prototypes.

• WAVE, NVDA, VoiceOver: Accessibility testing tools.

• AWS EC2: Canvas LMS test instance hosted on the cloud for integration testing.

User stories formed the foundation for our functional requirements and guided sprint priorities. 

Examples:

• “As a professor, I want to see sentence-by-sentence analysis so I can verify AI use.”

• “As a professor, I want to schedule detection to run automatically at due date.”

• “As a dean, I want data transparency to ensure policy compliance.”

Each story includes acceptance criteria and a priority level (Must Have, Nice to Have, Stretch Goal).

📄 User Stories 

We developed rich, research-backed personas to understand our stakeholders' needs. 

These included:

• Prof. Emma Lee (Humanities instructor focused on ethical AI use)

• Prof. John Kim (CS professor prioritizing automation and efficiency)

• Maya Thompson (Ethics-minded student)

• Lucas Chen (Tech-savvy student advocating for fair policy)

• Dr. Helen Ramirez (UCI Board Member ensuring institutional integrity)

These personas shaped our UX and accessibility priorities.

👤 Personas 

We identified the primary actors—Professors and Students—and mapped out how they interact with the system:

• Professors configure detection, run analysis, and view reports.

• Students submit assignments through Canvas.

The diagram captures authentication, parameter-setting, detection flow, and report delivery.

📊 Use Case Diagram 

The core flowchart illustrates how a professor enables detection:

1. Logs into CAVIAIR

2. Selects a course and assignment

3. Configures thresholds and model preferences

4. Triggers detection (or waits for automatic run at due date)

5. Reviews line-by-line results via SpeedGrader

This visualization clarified system behavior and shaped implementation sequencing.

🔁 Activity Diagram

We conducted needfinding interviews with professors at UCI and RCC, identifying concerns like:

• False positives from AI detection

• Disruption to existing grading workflows

• Ambiguity in academic policy
  
  

Findings directly informed system design, UX, and our ethics documentation.

📘 [Link to User Research Summary]

Interviewees included instructors from multiple disciplines and students across tech fluency levels. 

Themes we uncovered:

• Professors valued clarity and report explainability

• Students demanded fairness and transparency

• Everyone stressed accessibility and integration with Canvas

Insights shaped both technical and design trade-offs.

🎙️ [Link to Interview Findings]

Our final designs incorporated:

• Accessible color palettes

• Seamless Canvas-style UI

• Simplified detection dashboards

• Sentence-level color-coded AI breakdowns

All feedback from user testing was implemented in the final UI.

🎨 Figma Mockups 

CAVIAIR is a deployed, modular system with:

• Canvas LMS integration

• Real-time AI detection using 3+ services

• Sentence-by-sentence analysis

• REST APIs with authentication and session flow

• Docker-based deployment

• Accessibility-enhanced interface

Watch a walkthrough of the system from login to final report viewing.

🎥 Demo Video [3:14] 

Our documentation includes:

• Deployment instructions

• Canvas API setup

• Detection API usage

• Accessibility & ethics reports

• Architecture & database diagrams

📚 Documentation 

CAVIAIR seeks to modernize academic integrity workflows while respecting student rights and UX principles. Its implications:

• Helps instructors make informed decisions—not automatic accusations

• Supports students by being fair, transparent, and accessible

• Contributes to future-proofed learning environments where AI is acknowledged, not ignored

We’re building not just a tool, but a conversation around ethical AI use in education.