Faculty Examples: Anchor to Context

Connect the task to human experience, local context, personal interpretation, or fieldwork. That kind of grounding makes an assignment harder for AI to fully stand in for.

See the Course & Assignment Design page for more on how this move fits into the fuller design picture. 

Below are examples from UO faculty.

Professor McLees grades weekly research write-ups in GEO 201 with a rubric designed to reward high-quality, multi-modal, evidence-based work, no matter the tools used to complete it. This lets her and her class focus less on AI policing and more on the criteria she cares about. For her, the presence of GenAI, together with an increased focus on career readiness, highlights the importance of showing relevance to students, encouraging buy-in, and meeting them where they are. Her approach varies by level: she is largely AI-discouraging in lower-level courses, while in higher-level courses she models specific ways to use AI responsibly.

Design Moves

• The rubric names five criteria, each worth equal points: appropriate use of concepts from class, use of specific and appropriately scaled data, embedded visuals that support the argument, clear communication, and accurate, specific citation.
• Concepts must be used as they were presented in class and applied to the case the student is writing about, not defined in the abstract.
• Data must be specific and clearly support the discussion; the lowest tier names "stereotypes and unsupported generalizations."
• Each write-up includes at least three visuals (graphs, maps, pictures, or embedded videos) that are captioned, cited, and referenced in the text. Decorative images do not earn credit.
• Citations appear both in the text and in a full citation at the end.

In Professor Pashby's Linguistics 101 course, she revised an assignment around students collecting their own audio recordings of natural speech, which they eventually analyze. That personal, self-generated data is the assignment anchor. And it's not something AI can supply or substitute.

Design Moves

• Students collect their own audio data of natural speech as the first step. The raw material is inherently personal and non-substitutable.
• Both the no-AI and AI-integrated versions of the assignment keep this same anchor; what changes downstream is how the analysis is done, not where the data comes from.
• In the AI-integrated version, students are asked to check AI's analysis against their own recording, a check only possible because they hold the primary source themselves.

In Professor Heinz's HIST 308 course, students don't just pick a topic. Step One requires choosing one specific group (e.g., enslaved people, Indigenous communities) and one specific region (e.g., Deep South, Chesapeake), and that intersection has to shape the entire project. At least three of the 10 required timeline points must also cite a primary source from Through Women's Eyes, explained and connected back to the student's own argument.

Design Moves

• Required group and region intersection (e.g., "enslaved women in the Deep South") narrows the project to something genuinely specific, not a generic topic AI could fill in.
• At least three of 10 timeline points must cite a specific primary source from the course text, with an explanation of what it reveals and how it connects to the student's thesis.
• The project can't be completed without engaging specific, assigned source material.

An explicit "Authenticity Check" is built into the assignment design itself. Professor Raza answers her own design questions before sharing it with students, naming exactly what makes this resistant to generic AI output.

Design Moves

• Groups choose a current Oregon-specific criminal justice issue from a provided list (police use of force, sentencing disparities, immigration enforcement, etc.), not a generic or historical topic.
• Authenticity Check explicitly names "local knowledge" as something AI cannot replicate, and identifies the reflection component as the place students bring in their own perspective before turning to outside sources.
• Presentations must include eight sources at minimum, with three required to be directly tied to the local topic (e.g., Oregon Department of Corrections data), grounding the analysis in real, current, place-specific evidence.