Use capable tools without guessing the rules.
Students need to know what assistance is allowed, what contribution must be disclosed, and what work they must still be able to defend themselves.
For universities
Students already use agents. Faculty still set the standard.
When every course invents its own AI boundary, students cannot build a coherent practice and faculty inherit the exceptions. We propose one faculty-controlled pilot inside an existing program.
The current collision
Two legitimate needs meet inside the assignment.
Protect learning without pretending the tools are absent.
Faculty need authority over sources, assistance, assessment, integrity, and exceptions without becoming support staff for a campus-wide platform.
The working object
A boundary sheet attached to one assessed task.
Not a universal AI policy. A rule faculty and students can actually apply and inspect.
Preparation
Summarize faculty-approved sources and surface questions.
Students disclose material agent contribution.
Practice
Challenge a draft or rehearse an argument against the rubric.
Faculty define what help is permitted for the assignment.
Assessment
Return feedback only where the assessment design allows it.
Faculty retain grading, integrity, and exception decisions.
One-program pilot
Learn before asking the institution to standardize.
- 01
Select one module
Choose an existing course, one assessed task, and the faculty owner.
- 02
Write the boundary
Specify approved sources, permitted agent help, disclosure, review, and prohibited use.
- 03
Run one practicum
Students and agents work on a bounded artifact while faculty observe where the rules hold or fail.
- 04
Decide from evidence
Review work samples, exceptions, and faculty effort before considering a wider rollout.
First conversation
Bring one module and one unresolved boundary.
We will test whether a bounded practicum is useful before discussing a larger program.