AI as TA
Teaching With AI in Intro to Machine Learning
Shen Shen
January 30, 2026
Expanding Horizons in Computing
an updated talk on this topic can be found here
What kind of talk this is
Not a survey. Not a prescription.
A reflection on what emerged from experiments in my own teaching,
offered as a local view, and as a lens you might test against your own context.
Where did AI leverage actually show up?
Two axes for reasoning about AI use
Student-facing
Staff-facing
Routine
Creative
syntax drills
grading, logistics
open-ended projects
course design, explanation, narrative
The arc of this talk
Start with a clear win
↓
Then look at where pressure appeared — and how we adapted
↓
End with a small surprise that emerged
I:
Staff-facing AI amplified judgment and creativity
Brainstorm for analogies
strategize for design
critique and improve (iteratively)
Generate script
to smooth out segues
Vibe-code
slides animations
inspiration from podcast
\(\theta^*=\left({X}^{\top} {X}\right)^{-1} {X}^{\top} {Y}\)
A small moment from office hours
II:
AI shifted where learning signals became visible
AI changes which learning signals we can rely on, and when
Signal inflation — advanced courses
Master's class project execution quality now ≈ earlier PhD-level work before GenAI.
To assess depth → shift toward reasoning, design choices, failure recovery.
Signal delay — introductory courses
Students more fluent (on the high-level), but foundational gaps less visible early on.
To assess gap → shift how we train our staff.
Our legacy lab checkoff
(like structured OHs)
Our legacy lab checkoff
(like structured OHs)
how we change our staff training for this?
III:
Moving judgment upstream changed who could deeply participate in teaching
Registrar Registration Data
Student Course Materials Interaction Data
Internal Who-Has-Felt-What data
Exam Stats
Exam Source
| Branches | Contributors | |
|---|---|---|
| Spring25 | 45 | 21 |
| Fall24 | 1 | 7 |
| Spring24 | 1 | 11 |
lowered barrier for contribution
- All TAs are contributing
- Lots of LAs (sophomores) are contributing
- Workflow was improved by AI
- Content edit aided by AI
What actually carried the leverage
- The biggest gains came from where judgment lived, not which tools we used
- Making execution cheap didn't remove judgment, it made its placement visible
- Once judgment was placed deliberately, small adaptations had outsized effects
"Education is what remains after one has forgotten everything one learned in school."
— often attributed to Einstein
If AI accelerates the forgetting, our job is to be more deliberate about what we want to remain.
For me, what I want to remain is not knowledge, or skills per se, but the capacity to exercise judgment, and the motivation to want to.
expanding-horizons-talk
By Shen Shen
