A field of submission marks with one column being read
Day One

AI in this Course

We built something for you.
Tim Fraser
Tim Fraser, PhD
Assistant Teaching Professor, Systems Engineering
Cornell University
SYSEN 5300: Systems Engineering & Six Sigma
for Design and Operation of Reliable Systems

Everyone worries about bad AI. So what would good AI look like?

A pilot, built and tested by your teaching team. We have the methods — let's find out together.
AI in this Course
Where AI helps — and where it doesn't
What AI is genuinely good at
Reading code

AI evaluates what you submit.

  • Practice-activity submissions
  • Homework code
  • Feedback faster than any human team
  • An automated checker picks up new submissions every 30 minutes
  • We're aiming for ~1-hour turnarounds
What AI is bad at — humans keep it
Everything else

Our reclaimed grading hours go here.

  • One-on-one office hours
  • Team support
  • Project mentoring
  • Coaching you toward your black belts
AI in this Course
And when it gets one wrong?
1 Name the exact question
2 Name the exact problem
3 A human reviews it and fixes it
Every petition is logged. Your problem becomes our data.
New this term — when you need it, the regrade petition form is linked from Canvas.
We tested it like we teach you to test things
Homework: gap from a human grader

1.2 pts

Per sub-question — about 23% of a typical 5-point one. Most of that gap is on purpose: we hold a more rigorous line than past human grading.

Homework gradings measured

1,998

Sub-question gradings over 344 real submissions from previous terms, scored against what the human graders actually gave.

False "incomplete" marks

0

You do the work; the machine says you didn't. We refuse that failure mode.

Activity checks: agreement

98.6%

On our test corpus, against graders' work from previous terms.

Repeat confirmations

504

Same submissions, run again and again. Zero grade-flips.

Models benchmarked

3

Over 72 real-format submissions, screenshots included. The model we picked is cheap enough to run for everyone, all term.

AI in this Course
Your data stays home
Every call runs through Cornell's own AI gateway
Covered by Cornell's contracts — never a public chatbot
No grade changes without an audit trail
A human can always be summoned
This is a pilot. Tell us what's broken — measuring what fails is literally the discipline this course teaches.