AI assignment exception decision tree

For evaluating a student's request not to use a required AI component of an assignment.

Decision tree for AI assignment exception requests Flowchart from a student declining required AI use, through a protected-class accommodation check, an objective-necessity check, a rigor check, and a feasibility check, ending in one of five outcomes. Student declines required AI use Protected-class accommodation request? disability or religious exemption requests go to Accommodations or Complaince offices. No Yes Accommodations/Compliance handled outside this tree (denied → returns below) if not accommodated Is AI itself the learning objective? e.g., a course on prompting, ML methods, or the tool itself No Yes AI required the objective is about AI itself Achievable without AI, at equal rigor? pedagogical/validity test. Does a non-AI task measure the same outcome? Yes No AI required no non-AI task tests the same outcome Deliverable without disproportionate cost or labor? resource/undue-hardship test based on documented demand reviewed by Chair and CSEA Rep. Yes No AI required (for now) not feasible at current scale/ staffing. Revisit if that changes Build an equivalent alternative assessment same objective, rigor, and weight
Routed to another office AI required (pedagogical or objective-based) AI required/contingent (resource-based, revisit later) Alternative assessment built

Guidelines and Definitions 

1. Protected-class accommodation request?
This is not an instructor judgment call. The tree's only job here is to recognize the trigger and route it out. The Compliance Office assesses sincerity (per EEOC guidance) and, for disability, Accommodation Services determines the specifics of the functional limitation. If the request is denied there, it returns and re-enters the tree at gate 2, evaluated on pedagogical grounds like any other objection.
2. Is AI itself the learning objective?
Applies to courses where the objective is about AI (e.g., prompting, a specific AI-driven tool like Firefly or an ML pipeline), or AI literacy as a stated outcome. If yes, AI use is non-negotiable and no alternative is provided.
3. Achievable without AI, at equal rigor?
Does a non-AI task exist that measures the same outcome at the same level of difficulty? This is decided by the instructor or department based on what the assignment tests.
4. Deliverable without disproportionate cost or labor?
This is a resource/feasibility question that should be evaluated against real, documented demand (how many students per term actually request this), and reviewed above the individual instructor so two sections of the same course don't reach opposite outcomes on one person's sense of workload. Because this gate is contingent on current resources, a "no" here should be revisited if staffing, budget, or request volume changes.
5. Build an equivalent alternative assessment
The alternative should be documented with instructions, rubrics, grade weights, etc. indicating equivalent rigor and outcome measurement, not just "an oral exam."

Open items — not yet resolved in this tree