AI Governance

How MyTA uses AI responsibly in higher education

1. Our Approach to AI

MyTA uses AI to assist instructors — never to replace them. Every AI-generated output (grades, feedback, draft replies) is a draft that requires instructor review and approval before reaching students. No AI decision is final without a human in the loop.

2. AI Models We Use

MyTA uses Google’s Gemini models, accessed via their commercial API:

Gemini Flash: Essay grading, draft reply generation, document classification, coordinator responses
Gemini Flash-Lite: Intent classification and fast planning tasks
Gemini Pro: Complex grading, voice profile analysis

All models are provided by Google. We do not build, train, or fine-tune custom AI models.

3. Student Data Never Trains AI Models

Google’s Gemini commercial API has a zero-retention policy — inputs and outputs are not stored, cached, or used for model training.
Before any text is sent to the AI, student personally identifiable information (names, IDs, emails) is deterministically strippedand replaced with pseudonyms like “[Student A]”.
The stripping pipeline is not AI-powered — it uses roster-based matching and regex patterns. If stripping fails, the AI request is rejected. Data is never sent unstripped.

4. Human Oversight

MyTA follows a draft → review → approve → publish model:

AI generates a draft (grade, feedback, reply, evaluation)
The instructor reviews the draft in the web app or via Telegram notification
The instructor can approve, edit, or reject the draft
Approved content is saved as an invisible draft in the LMS — students cannot see it yet
The instructor explicitly publishes when ready — only then can students see the content

High-risk actions (grading, messaging students, posting to the LMS) always require explicit instructor approval.

5. Fairness & Bias Mitigation

AI grading is anchored to rubrics or assignment instructions provided by the instructor — not to general AI knowledge or training data biases.
Instructor voice profiles ensure feedback matches the instructor’s own communication style, not a generic AI tone.
Course-specific policy settings (word count minimums, grading intensity, feedback level) are configurable to reflect institutional standards.
All AI grading is reviewed by the instructor before release to students.

6. What AI Cannot Do in MyTA

AI grading is advisory — final grades are always set by the instructor
The system cannot contact students directly
The system cannot modify LMS settings or administrative functions
Content involving self-harm, harassment, accommodations, or academic integrity is triaged to the instructor without AI processing

7. AI Usage Monitoring

Every AI API call is logged with the action type, model used, and token counts. This data is available to institutional administrators through the admin dashboard, supporting transparency and cost tracking.

8. Questions & Documentation

For questions about MyTA’s AI practices, contact privacy@courseops.ai. Detailed technical documentation, including our AI System Card and data flow diagrams, is available to institutional IT teams upon request.