Inspiration

  • Address student confusion about what counts as misconduct and provide supportive, educational nudges instead of punitive alerts.
  • Combine real coursework context (Canvas data) with micro‑lessons and actionable reminders to reduce risky behavior before submission.
  • Make integrity coaching practical, private, and accessible across common student personas (freshman, working students, international students, students with disabilities).

What it does

  • Integrates with Canvas via user-provided API token to surface assignments, deadlines, grades, and rubrics locally. (Currently parses the webpage to extract information)
  • Real-time nudges: copy/paste warnings, pre-submission self-check, honor-pledge prompts, and collaboration vs. individual hints.
  • Educational modules: micro-lessons, situational quizzes, and reflection prompts tied to course context.
  • Dashboards: integrity risk snapshot, progress metrics, private logs; links to campus resources and support.
  • Chatbot: Provides AI conversation with a vast knowledge base custom to the URLs mentioned and institute involved.
  • AI: Support for webllm, gemini and regex. Also provides various tools like planner and feedback generator to ease the process.

How we built it

  • Browser extension that stores the Canvas access token locally after validation; uses Canvas REST API scopes for assignments, courses, calendar, and enrollments.
  • Modular front-end components for dashboard, nudges, and educational modules (reusable UI + state management).
  • Lightweight background service to refresh Canvas data on demand and fire context-aware prompts.
  • Security-first choices: local token storage, explicit consent flows, minimal scopes by default, and clear privacy UI.
  • Selected libraries and patterns for reliability and testability (modular services, well-scoped API wrappers, robust error handling).

Challenges we ran into

  • Canvas API variability across institutions and changing LMS behavior after updates. Another method (webpage parsing via contentscripts) was adopted temporarily to extract data.
  • Balancing helpful nudges with non-intrusive UX, and ensuring accessibility (keyboard navigation, dyslexia-friendly fonts).
  • Handling token validation, refresh, and secure storage while keeping the flow simple for students.
  • Planning based on user capabilities, i.e., user customization.
  • Understanding huge webpages via AI for creating knowledge base. Sent separate chunks iteratively and combined them all in end to achieve final results.

Accomplishments we're proud of

  • Working Canvas token flow and reliable data pulls across common course setups.
  • A functional pre-submission self-check and embedded draft feedback that students can use instantly.
  • Clear privacy model (FERPA-aware design) and accessible UI features.
  • Scalable architecture that separates LMS adapters, UI, and rules so we can expand quickly.

What we learned

  • Privacy-first, opt-in design increases trust and adoption.
  • Small, contextual interventions are more effective than broad monitoring.
  • Modular code + clear API wrappers reduces friction when supporting other LMSes.
  • Real-world testing with actual student workflows uncovers edge cases faster than synthetic tests.

What's next for AIM

  • Expand LMS support beyond Canvas and add institution-specific policy templates.
  • Improve detection and guidance with explainable models (research-backed, low false-positive thresholds).
  • More educator-facing opt-ins and aggregated, anonymized analytics for institutional use.
  • Offline/low-bandwidth mode, deeper accessibility refinements, and longitudinal studies measuring behavior change.

Built With

Share this project:

Updates