Assessment Ideas by AI Role
Once you know what role you want AI to play in a step of your assignment, use this page to browse concrete assessment formats built for that role. Filter to one role, or browse all. Each entry includes a short description and a design note on how to make the role work in practice. For a version tailored to your own learning outcome, use the Assessment Design Tool.
Role 01Forbidden Tool
AI use is not permitted. The prohibition only works if the design makes it meaningful, live, oral, or observed formats are what make AI assistance actually impossible in the moment, not just against the rules.
Forbidden ToolLive / In-Class
Oral Exam
A structured oral examination where students respond in real time to a small set of core questions and follow-ups.
Design Note
Oral examination is the gold standard for Forbidden Tool assessments. Design 3–4 core questions and 1–2 probing follow-ups that shift based on the student's responses. Grade on reasoning quality and use of evidence, not polish or fluency.Forbidden ToolLive / In-Class
Presentation with Live Q&A
A short presentation followed by a required live Q&A, so unscripted follow-up questions can surface genuine understanding.
Design Note
Prepare 2–3 follow-up questions in advance that go well beyond any prepared script. Grade the Q&A response as heavily as the presentation itself, that is where genuine understanding, or its absence, becomes visible.Forbidden ToolLive / In-Class
Case Simulation
A realistic scenario requiring students to reason under structured conditions, ideally with no single right answer and a brief defense of their decisions.
Design Note
An in-class case simulation requires students to reason and respond in real time. A brief verbal debrief, asking each group to explain their key decision, adds accountability and depth.Forbidden ToolLive / In-Class
In-Class Exam
An in-class, closed-note written exam, ideally built around a scenario or case rather than pure recall.
Design Note
Make the prompt scenario-based rather than recall-based: give students a situation they could not have anticipated, and ask them to reason through it using course frameworks.Forbidden ToolLive / In-Class
Structured Discussion
A facilitated in-class discussion requiring students to engage in real time and build on each other's contributions.
Design Note
Structure it to require students to reference specific evidence, take positions, and respond to challenge. Consider brief post-discussion written exit reflections as evidence of individual engagement.Forbidden ToolLive / In-Class
Assessed Participation
Structured, assessed participation in class activities, defined by quality of contribution rather than frequency alone.
Design Note
Make participation expectations discipline-specific: what does it mean to analyze, evaluate, or contribute as someone in your field? Make those expectations visible to students in advance.Forbidden ToolLive / In-Class
Viva Voce (Flash Defense)
A brief live speaking examination, 5 to 10 minutes, in which students explain their argument or a specific choice aloud without referring to their paper.
Design Note
A two-minute "flash interview" at the instructor's desk, or a "concept justification" variant asking why a specific word or structural choice was made, both reveal instantly whether a student understands their own work.Forbidden ToolLive / In-Class
Blind Syllabus Challenge
Students receive a syllabus with reading descriptions blanked out and must justify, in a timed in-class response with no computer access, which course concepts solve a given problem.
Design Note
This tests the exact skill AI is most tempted to outsource: identifying which framework is significant and why. Grade on precision and justification of the concept selection, not on whether students chose the "right" answer.Forbidden ToolLive / In-Class
In-Class Localized Writing
In-class writing under conditions that remove access to external AI tools, using hyper-local prompts tied to specific class material.
Design Note
Reference specific class discussions, primary sources, or an obscure detail from course materials with no AI training-data footprint. Consider a "Sudden Pivot" prompt where a core constraint changes live during the session.Forbidden ToolLive / In-Class
Proctored Concept Defense
A short live or recorded defense in which a student explains how their understanding of one named concept evolved across the semester, with the concept revealed only at the start.
Design Note
A student who let AI weave the concepts together will have no personal narrative of learning to tell. Grade on intellectual development, not product quality.Forbidden ToolLive / In-Class
Raw Data Wrangling Exam
A proctored, internet-blocked session where students manually isolate variables and sketch a visualization from a raw, messy dataset, no automated tools or AI assistance available.
Design Note
Confirms data competence before students are allowed to use any automated design tool. Grade on variable selection, visualization logic, and caption reasoning, not polish.Forbidden ToolStructured / Staged
Staged Writing Process
A large writing assignment broken into multiple mandatory, separately graded stages, so a polished final product cannot be reverse-engineered from nothing.
Design Note
Include a handwritten brainstorm, a library-sourced annotated bibliography, a defended thesis and outline, and a final draft with a revision memo. Each stage must be completed before the next begins.Forbidden ToolLive / In-Class
Closed-Note Quiz
A short, targeted quiz assessing specific knowledge, administered in class under closed-note, closed-device conditions.
Design Note
Most effective as low-stakes retrieval practice given early and often rather than one high-stakes event. In-class administration removes the opportunity to look up or generate answers. Consider allowing one review-and-correct attempt to deepen encoding.Forbidden ToolLive / In-Class
Lab / Practical Task
A timed, observed practical task using course tools or methods, graded on process and decision-making as much as the final result.
Design Note
Build in a brief verbal debrief immediately after the task: "Walk me through your reasoning at this decision point." The debrief is often more revealing than the artifact itself.Forbidden ToolAsync-Friendly
Recorded Video Walkthrough
A short, 2 to 3 minute recorded video in which a student explains their own submitted written work aloud, required before the piece is graded.
Design Note
This catches the same thing a live oral defense catches, whether a student can talk through their own reasoning, without needing class time or an in-person room. A student who wrote the piece will explain it naturally; a student who did not will reveal the disconnect within the first thirty seconds. Particularly useful for online or asynchronous courses where a live defense is not practical.Role 02Transparent Partner
AI is allowed, but students must document how they used it, cite it, and show their thinking process. The documentation itself becomes assessable, not just the final product.
Transparent Partner
Essay with AI-Use Appendix
A focused analytical essay requiring students to construct an evidence-based argument using disciplinary frameworks.
Design Note
Require an AI Use Appendix: tool used, one or more example prompts, what the AI produced, and a paragraph on how the student's own judgment shaped what they kept, revised, or discarded.Transparent Partner
Individual Project with AI Documentation
An individual project unfolding in stages, proposal, draft, peer feedback, final submission, so the student's thinking stays visible throughout.
Design Note
Build an AI Documentation component into the project rubric. Students should be able to explain every AI-assisted decision in a brief debrief or written log, graded with real weight.Transparent Partner
Reflection on AI Use
A structured reflection anchored to specific prompts, examining not just what was learned but how the student's thinking changed.
Design Note
Where did the tool help? Where did it miss disciplinary context? What did the student have to already know in order to evaluate its output? This reflection often reveals expertise more than the work product itself.Transparent Partner
Portfolio with Revision Evidence
A portfolio of several artifacts, drafts, feedback received, and revision evidence, framed by a reflection tracing growth across the semester.
Design Note
Weight the framing reflection and evidence of growth heavily alongside the artifacts themselves, not just the polish of the final pieces.Transparent Partner
Peer Review with Revision Memo
Students give and receive structured feedback on sufficiently developed work, closed by a memo showing what changed in response and why.
Design Note
Require specific, evidence-based feedback rather than general impressions, and require students to act on the feedback they receive, not just log it.Transparent Partner
Version History Audit
Students submit work through a platform that preserves version history, so the instructor can confirm the document grew incrementally rather than appearing all at once.
Design Note
This does not eliminate AI use, it makes the documentation of a genuine writing process itself a graded deliverable. A Draftback-style playback of typing cadence works as a stronger variant.Transparent Partner
Dataset Interrogation
Before an analysis or visualization is due, students submit a one-page memo naming their dataset's publisher, year, and scope, confirming they can independently retrieve it.
Design Note
A student whose numbers came entirely from an AI tool's internal claims, with no real dataset behind them, will not be able to produce this memo. Require it before any analysis proceeds.Transparent Partner
Annotated Bibliography with AI-Use Statement
An annotated bibliography where each entry summarizes, evaluates, and explains the intended use of a source, making research process visible.
Design Note
Require an AI Use Statement documenting which tools were used, what prompts were entered, and how the student evaluated or modified any AI-suggested sources or summaries. The documentation itself is assessable.Transparent Partner
Concept Integration Map
A sentence-by-sentence index showing exactly where and how each course concept changes the argument at that specific point in the paper.
Design Note
A concept dropped into the text without genuinely changing the reasoning will have no sentence to point to, because it never actually changed anything. Grade the map alongside the paper.Role 03Required Collaborator
AI use is mandatory, built into the workflow. Students must use its output, evaluate it, and make explicit decisions about what to keep, revise, or discard, mirroring real professional practice.
Required Collaborator
Essay from AI First Draft
Students use AI to generate a first draft or outline, then revise it into their final essay.
Design Note
Require a documented revision process showing where and why the student departed from the AI output. Grade the revision and its rationale at least as heavily as the final essay.Required CollaboratorLive / In-Class
AI-Assisted Case Simulation
Students use AI to generate an initial case analysis or recommendation, then critique and revise it using course frameworks.
Design Note
The gap between the AI's output and the student's expert judgment becomes the central learning artifact. Add a "pitch" or defense component where the revision rationale is explained aloud.Required Collaborator
Individual Project, Defined AI Phase
An individual project with AI built into one defined workflow phase, ideation, prototype feedback, or literature synthesis.
Design Note
Students submit both their AI interaction log and their rationale for what they kept, changed, and rejected, making disciplinary judgment visible and assessable.Required Collaborator
Group Project with Divided Roles
A collaborative project for a small group, with individual contribution logs and peer assessment to keep each member's thinking visible.
Design Note
Design the deliverable so the whole is achievable only through genuine collaboration, not parallel individual work pasted together.Required Collaborator
AI First Draft Revision
The instructor distributes a generic AI draft on the assignment topic to the whole class; the assignment is to take it apart and rebuild it correctly.
Design Note
Students submit both the marked-up original and their revised version, with a memo explaining every significant change. The AI draft's weaknesses become the curriculum.Required Collaborator
Prompt Engineering Log
When AI is part of the assignment workflow, students submit every prompt they used as a graded appendix, justified and tied to verification.
Design Note
This changes what is graded: not just the final product, but the quality of human judgment exercised at every step of the AI-assisted process.Role 04Subject of Critique
Students evaluate, challenge, or improve AI-generated output. The critique itself is the evidence of learning, developing the disciplinary judgment to catch what AI gets wrong.
Subject of Critique
Hallucination Hunt
Students receive an AI-generated brief on their own course topic containing at least one fabricated statistic, misattributed study, or nonexistent citation, with no warning.
Design Note
Grade on the quality of the verification process, which lookup strategies were used, what evidence confirmed or disproved each claim, and what the student learned about what makes false claims look credible.Subject of Critique
Bait-and-Switch Source Check
Students receive a real article alongside a polished AI summary of it that has been altered to include a significant contextual error.
Design Note
Grade on whether students identified the planted error and on the reasoning they used, not just the correct answer. Award extra credit for catching errors the instructor did not deliberately plant.Subject of Critique
AI Output Deconstruction
The instructor runs the actual assignment prompt through AI and distributes the result; students annotate where it hallucinated connections or applied a concept superficially.
Design Note
Grade the annotation on the depth of disciplinary reasoning demonstrated, not on how many errors were found. This forces the cognitive work of editing and correcting a machine.Subject of Critique
Citation Autopsy
Students use AI to generate a short literature review, then audit every citation for whether the source actually said what the AI claimed.
Design Note
The three-question audit per source, does it say what was claimed, was it used appropriately, did the AI capture the researcher's tone of certainty, is assessable and cannot be faked.Subject of Critique
Essay Fact-Check
Students fact-check, annotate, or formally critique an AI-generated essay on the course topic.
Design Note
Where does the AI get it right? Where does it fail, factually, interpretively, or disciplinarily? The critique itself demonstrates what the student actually knows.Subject of CritiqueLive / In-Class
Presentation, Response to AI Argument
Students present a "response" to an AI-generated argument or analysis: where they agree, where they correct, and what the AI's reasoning lacked.
Design Note
The presentation format forces students to make the critique legible to an audience, not just note it privately.Subject of CritiqueLive / In-Class
Oral Exam, Critique Format
A live oral exam built around evaluating AI-generated content rather than recalling course material.
Design Note
Provide the AI output at the start of the exam and grade the student's real-time evaluation, not a rehearsed take.Subject of Critique
Annotated Bibliography, AI-Sourced
Students evaluate an AI-compiled bibliography, checking whether each source is real, relevant, and accurately characterized.
Design Note
Require students to independently verify and re-characterize each entry, not just format citations AI already found.Subject of Critique
Reflection on AI's Blind Spots
Students prompt an AI tool with the course's learning outcome, then reflect on where its response was credible and where their own disciplinary knowledge produced something AI could not.
Design Note
This reflection makes a student's own expertise concrete and visible, in contrast to what the AI could generate.Subject of Critique
Peer Authenticity Review
Students evaluate a peer's (or a provided) piece of writing for signs it was AI-generated rather than authentically the student's own, and must point to specific evidence for their judgment.
Design Note
Require concrete markers, not a gut feeling: generic phrasing, a voice inconsistent with the student's earlier work, or an absence of specific references to material only covered in class. This trains a skill several universities now favor over AI-detection software, which faculty at Vanderbilt, UT Austin, Northwestern, and Maryland have all been advised not to rely on.Design Add-Ons
These are not standalone assessment types, they are rubric additions and grading-policy tweaks that strengthen any option above. Add one or more to close the most common AI-use loopholes in your design.
Universal
Two-Stage Submission
Require a rough draft for low-stakes feedback before the final version is due. Grade the revision process, not just the final product.
Universal
The "So What?" Requirement
Require every claim to be followed by an explicit statement of its significance to the learning outcome. Reduces AI content that is correct but analytically hollow.
Universal
Metacognitive Reflection Add-On
Attach a short paragraph on the student's own decision-making process, graded independently. A student who outsourced the thinking cannot write this with any specificity.
Universal
Randomized Prompt Variation
Each student receives a slightly different version of the prompt, a different case, dataset, or constraint. Shared AI outputs become immediately detectable.
Forbidden Tool
In-Class Baseline Writing
Early in the semester, collect a short unassisted in-class writing sample as a reference point for later comparison against take-home work.
Forbidden Tool
Restructuring the Grading Matrix
Shift 70 percent or more of the grade to visible, unlaunderable steps, handwritten outlines, in-class drafting. The polished final product is worth the least.
Transparent Partner
The Prompt Log
Require every prompt given to AI as an appendix, with a sentence justifying each one. Graded on the quality of the justification, not the AI's output.
Transparent Partner
The "What I Changed" Annotation
Students mark up the AI output they received, highlighting every edit made and why. A student with no meaningful changes has not exercised real judgment.
Transparent Partner
The Audit Trail
Grade a portion of the assignment on the process itself: AI output received, handwritten corrections, and cross-references to sources used to verify or replace AI claims.
Required Collaborator
AI Limitations Statement
Students must identify at least two specific places where the AI output was inadequate for the task, and explain how they compensated.
Subject of Critique
Counter-Prompt Challenge
After finding AI's errors, students write a better prompt that would have avoided them, and explain why. Teaches AI literacy as a discipline-specific skill.
Subject of Critique / Transparent Partner
Closing the Citation Format Loophole
Separate citation formatting from citation verification. A correctly formatted citation earns zero credit until the student confirms the source exists and says what it is claimed to say.