Generative AI showed up on campus the way a marching band shows up in a library: loudly, suddenly, and with
everyone arguing about whether it’s “allowed.” For many faculty, the first reaction has been
understandablepanic-flavored vigilance. New tools appear, students experiment, and the academic integrity
alarm starts doing its best smoke-detector impression at 2 a.m.
But here’s the twist: treating AI like a contraband energy drink doesn’t make it disappear. It just makes it
sneakier. The more we center our response on “catching” students, the more teaching time gets spent on
surveillance instead of learning. A better move is the one Faculty Focus and many teaching centers are
pointing toward: shift from policing to pedagogydesign learning that stays meaningful even when AI exists,
and teach students to use these tools with judgment, transparency, and discipline-specific skill.
Why the Policing Mindset Breaks (Even If Your Intentions Are Pure)
Policing feels appealing because it’s familiar: define wrongdoing, detect it, punish it. The problem is that
generative AI doesn’t behave like a hidden cheat sheet. It behaves like a writing partner that never sleeps
and never asks for extra credit. Detection tools can be inconsistent, and many institutions have warned that
relying on AI-detector scores alone can lead to false accusations and inequitable outcomesespecially for
multilingual writers.
Even when detection “works,” it can quietly sabotage the class climate. Students feel suspected before they
feel supported. Faculty feel like bouncers instead of educators. And the real tragedy: the course becomes a
contest of evasion rather than a community of inquiry.
What students are actually doing with AI (hint: it’s not always villainy)
Students use AI for many reasons, and not all of them are “to avoid thinking.” Some are overwhelmed and
triaging time. Some have anxiety about writing. Some want quick feedback. Some are trying to translate ideas
into academic English. Some are curious. If we treat every use as misconduct, we miss the chance to teach the
hard part: how to evaluate, refine, cite, and take responsibility for work created with help.
Pedagogy-First Thinking: The Question That Changes Everything
A teaching-centered response doesn’t start with “How do I stop AI?” It starts with:
“What do I want students to learn, and what evidence would convince me they learned it?”
Once you answer that, AI becomes less of a threat and more of a design constraintlike calculators, Google,
or the fact that half your students will attempt the reading five minutes before class (bless them).
The U.S. Department of Education has emphasized that AI brings real opportunitieslike personalized feedback
and supportwhile also creating risks that require thoughtful policy and human oversight. In other words:
don’t ignore it, don’t worship it, and definitely don’t let it run your course like an untrained substitute.
Course Policies That Don’t Make Everyone Miserable
The most effective AI policy is the one students can actually understand and follow. Many universities now
recommend a “clarity over cleverness” approach: state what’s allowed, what isn’t, and what students must do
if they use AI. Keep it short enough that it doesn’t need its own final exam.
A practical “AI-use ladder” policy
-
Level 0 No AI for this task: e.g., in-class writing, timed quizzes, reflections based on
personal experience, oral exams. -
Level 1 AI allowed for brainstorming: topic ideas, outlines, generating questionsbut
students must write the final product themselves. -
Level 2 AI allowed with attribution: students may use AI for drafting or editing, but
must disclose how they used it and verify claims. -
Level 3 AI required: students must use AI as part of the learning objective (e.g.,
critique AI output, compare it to scholarly sources, or use it to generate test cases in a coding course).
Teaching centers often add two essential reminders: AI can “hallucinate” (confidently wrong answers), and
students are responsible for verifying outputs. Also: don’t paste sensitive student data or protected
materials into consumer tools unless your institution has approved safeguards.
Assessment Redesign: Make Learning Hard to Fake and Easy to See
Here’s the good news: you don’t have to throw out every essay forever. You just need to stop using a single,
generic product as the only proof of learning. In the AI era, strong assessment is
process-rich, context-specific, and multi-evidence.
1) Shift from product to process
Ask for artifacts that reveal thinking over time: drafts, annotations, decision logs, reflection memos, or
short “why I made these choices” notes. Faculty Focus highlights this shift because it changes the student’s
job from “submit something that looks right” to “show me how you got there.”
- Draft checkpoints: thesis statement + 2 sources by Week 3, outline by Week 4, revision plan by Week 5.
- Source trail: students highlight where each claim is supported (or admit it isn’t).
- Revision memo: “Here are the three biggest changes I made and why.”
2) Add authentic constraints (without turning your class into a security theater)
You don’t need medieval proctoring. You need thoughtful constraints that make learning visible:
short oral explanations, in-class synthesis, or task variations that require course-specific context.
- Micro-vivas: a 3–5 minute conversation where students explain one key choice in their work.
- In-class “bridge” writing: students bring notes, then write a short analysis connecting two course ideas.
- Personalized data sets / prompts: each student gets a slightly different case, scenario, or dataset.
3) Design tasks where AI is the “first draft,” not the final answer
One of the best uses of AI in higher education is as a foilsomething students must evaluate, correct,
and improve. This builds the exact skills employers want: judgment, domain knowledge, and critical thinking.
Example (History): Ask AI to summarize a primary source. Then students identify what it missed,
what it misread, and how context changes interpretation.
Example (Nursing): Give AI a patient scenario and ask for an assessment plan. Students critique it
for safety, bias, missing information, and alignment with clinical guidelines.
Example (Business): AI drafts a marketing pitch. Students must verify every factual claim, cite real
sources, and rewrite the pitch for a specific audience.
Example (Computer Science): AI proposes code. Students test it, document edge cases, explain why
it fails, and harden it.
Teaching AI Literacy: The New “Library Session” for 2026
If AI is going to be part of students’ lives, “Don’t use it” is not an education strategy. A pedagogy-forward
response makes AI literacy explicitjust like we teach citation, research methods, and how to tell the
difference between peer-reviewed evidence and someone’s very confident blog post.
AI literacy outcomes worth grading (yes, grading)
- Prompting with purpose: students can specify constraints, context, and success criteria.
- Verification: students can fact-check AI claims with credible sources.
- Bias awareness: students can identify how outputs may reflect skewed data or assumptions.
- Attribution: students disclose AI use and cite appropriately when required.
- Discipline norms: students know what “ethical use” means in your field (and why).
Stanford and other universities have developed “critical AI literacy” resources for instructors and students.
The theme is consistent: the goal isn’t to become a prompt wizard; it’s to become a critical thinker who can
use AI without outsourcing responsibility.
Ethics, Equity, and Trust: The Parts Everyone Forgets Until Something Goes Wrong
AI raises ethical issues that land directly in teaching: privacy, intellectual property, accessibility, and
fairness. NIST’s risk management guidance emphasizes trustworthiness characteristics like managing harmful
bias, protecting privacy, and maintaining accountability. Those aren’t abstract idealsthey map neatly to
classroom decisions.
Privacy: “Don’t paste student data into the robot” is a start, not a policy
If students are asked to use AI tools, be explicit about what they should never upload: grades, private
health details, confidential research data, or protected course materials. Offer alternatives when possible
(institution-approved tools, anonymized data, or non-AI options).
Equity: access and false accusations
Some students have paid AI tools; others don’t. Some students have quiet workspaces; others share bedrooms,
jobs, and responsibilities. And because AI-detection tools can be error-prone, a heavy policing strategy can
disproportionately harm students who already feel vulnerable in academic spaces. A pedagogy-forward course
reduces this by emphasizing transparency, process, and multiple forms of evidence.
Faculty Workflow: Let AI Save Your Time (So You Can Spend It on Teaching)
Here’s a little secret: the best “AI policy” is the one that also improves your life. Used responsibly, AI
can reduce busywork and expand feedbackwithout replacing your expertise.
High-value, low-risk faculty uses
- Rubric refinement: generate draft criteria, then adjust to match your standards.
- Feedback starters: create comment banks for common issues, then personalize them.
- Practice questions: generate quiz items, then verify and calibrate difficulty.
- Examples and counterexamples: produce “almost-right” responses for critique activities.
The key is to keep humans in charge. AI can draft; you decide. AI can suggest; you verify. AI can remix; you
align to outcomes. Think of it as a power tool: useful in trained hands, terrifying in a toddler’s.
A Ready-to-Use Classroom Script (Because Students Need to Hear This Out Loud)
Try saying something like this in Week 1:
“AI tools are part of the world you’re graduating into. In this course, we’ll use them in ways that support
learningnot shortcuts that erase learning. If you use AI, you must disclose it and you are accountable for
accuracy, citations, and the quality of your reasoning. My job is to help you build skills that AI can’t
replace: judgment, domain understanding, and ethical decision-making.”
That one minute of clarity can prevent ten weeks of confusion (and seventeen emails titled “Quick Question”
that are not quick).
The Bottom Line: AI Doesn’t End EducationIt Exposes Weak Assignments
Generative AI is a stress test for our teaching. If an assignment can be completed with shallow pattern
matching, AI will do it faster. But if the assignment requires interpretation, evidence, lived experience,
disciplinary reasoning, and reflectionAI becomes a tool to interrogate, not a shortcut to hide behind.
Moving from policing to pedagogy isn’t “going soft.” It’s going smarter. It’s teaching students how to think
in a world where machines can draft words, generate images, and imitate competence. The real outcome is a
graduate who can use AI responsiblyand still produce work that proves they know what they’re doing.
Experiences from the Shift: from the Front of the Classroom
The first time I heard a colleague say, “I can just feel when it’s AI,” I nodded politely and kept my
face neutralbecause nothing says “objective assessment” like vibes. In early semesters of widespread GenAI,
many instructors tried to fight the tide with stricter rules and sharper suspicion. The result wasn’t more
learning. It was more awkward conversations, more student anxiety, and a weird new genre of office-hour
moment: “Professor, I swear I wrote it, but I also used Grammarly, and also I asked ChatGPT for synonyms, and
now I’m not sure if I exist.”
What actually improved things wasn’t a better detectorit was better design. In one writing-heavy course, a
faculty member replaced the single end-of-term essay with three smaller submissions: (1) a personal stake
statement (why this topic matters to you), (2) a source map (how each claim is supported), and (3) a
revision memo that explained what changed between drafts. Students still used AI sometimesmostly for
outlining or phrasingbut the center of gravity shifted. Instead of “Can I get away with this?”, students
asked “Is this claim real?” and “How do I make my argument clearer?” That’s not just compliance; that’s
learning.
In a STEM course, another instructor leaned into AI deliberately: students were required to prompt an AI
tool to explain a concept, then write a short critique identifying at least two errors, oversimplifications,
or missing assumptions. The surprise was how quickly students stopped treating AI like an oracle. They began
treating it like a fallible intern: helpful for speed, dangerous for accuracy, and always in need of
supervision. The instructor reported fewer integrity headaches because the assignment made AI use visible and
academically useful.
The most human moment came from a student who said, “I use AI because I’m scared I sound dumb.” That’s a
pedagogical signal, not a disciplinary one. When instructors responded with structureexamples of strong
writing, opportunities for feedback, and transparent rulesstudents didn’t become perfect angels. But the
relationship changed. AI use became something students could talk about honestly. And once it was talkable,
it became teachable.
The shift isn’t magical, and it isn’t instant. It’s a series of small moves: a clearer policy, an assignment
that requires decision-making, a reflection that captures thinking, a quick oral explanation, a reminder to
verify and cite. Over time, the classroom stops feeling like a courtroom. It starts feeling like what it was
supposed to be all along: a workshop for building minds that can work with powerful tools without surrendering
their integrity.
