Trust
How we use AI, and what we do not claim
Plain words for deans, assessors and students.
What the AI does
- It reads a student’s work three separate times and drafts a score for each criterion.
- The program keeps the middle draft, the median, so one odd run does not decide.
- Every quote in the feedback is checked word for word against what the student wrote. A line that is not in the student’s work is dropped.
- When the drafts disagree by a wide gap, the work is flagged for the assessor.
What people do
- An assessor reads the work and confirms every score. Nothing is confirmed automatically.
- Each confirmation carries a statement that the assessor read the work and judged each criterion personally.
- A sample of work is scored blind, before the assessor sees the AI draft. That is how the program checks the AI against people.
- A mentor writes a short note between attempts.
What we do not claim
- We do not claim that the AI matches faculty. The program measures it, and shows the measurements with the number of items behind them.
- We make no claim about how much students learn.
- A figure here is shown only when real records produced it. Demo figures say “Synthetic demo data”.
Your data
- Retention policy. What is kept, for how long, and how to ask for deletion. Mentoring-session transcripts are deleted after a set period.
- Access control. Who can see what. A student sees their own work, a mentor sees their own students, and staff see only what their role needs.
- Subprocessors. The outside services that handle data for the program, and what each one does.
Ask the program office for the full documents.