Eclatech Solutions

AI LEARNING FOR ALL

Faculty Resources

AI in the Classroom:
Detection, Dialogue, and Academic Integrity

Practical moves for community college instructors. Copy a prompt. Redesign one assignment. Use it this week.

Start with the Four Moves Print this page
Start Here

Why Redesign Beats Detection

Three things worth knowing before you rely on a detection score.

Detectors produce false positives

Every AI detection tool flags some human writing as machine-generated. There is no reliable way to know, from the score alone, which flags are wrong. Published research on this is linked in Sources.

A score is not evidence

Most academic integrity processes require evidence that can be explained and defended. A percentage from a third-party tool, with no visibility into how it was calculated, rarely meets that bar on its own.

The reliable alternative is design

You do not need to catch AI use. You need assignments where understanding is visible and its absence shows on its own. The four moves below do that with the prep you are already doing.

The Framework

Four Redesign Moves

You do not redesign your course. You redesign one assignment, using one of these.

1

Make the process visible

Staged submissions: outline, draft, revision, each with a checkpoint. A polished product can be generated in seconds. A messy, evolving process across three weeks cannot.

2

Make it local

Anchor the task to something AI cannot access: a campus event, a business near campus, a workplace observation, an interview. Harder to outsource and a better assignment anyway.

3

Make AI the object

Students generate an AI answer, find three errors or omissions, and fix them with reasoning. Auditing AI output is a skill employers now ask for directly.

4

The two-minute conversation

"Walk me through your main argument." Thirty seconds tells you whether the understanding is theirs. It is show-your-work, and it works in every discipline.

Three Actions

Three Things You Can Do This Week

One for you, one for your students, one for your department.

1 · For you: Test one assignment

Paste one assignment prompt into any AI tool (ChatGPT, Copilot, Gemini, Claude). If the AI completes it cleanly, that is your redesign candidate. Then apply one of the four moves.

Here is an assignment prompt from my course: [PASTE YOUR ASSIGNMENT]. Complete it exactly as a student would.
Rewrite this assignment so it requires [staged submissions with checkpoints / a local interview or observation component / a section where students critique an AI-generated answer]. Keep these learning objectives intact: [PASTE OBJECTIVES].

2 · For your students: Have the conversation

Before mid-semester, spend one class period on the four questions in the next section. Non-punitive, no grades attached. Most students have never been shown how to use AI well, and a classroom is the best place for that to happen.

3 · For your department: One question, one meeting

Bring one question to one meeting this semester. Two you can use as written:

What is our department's current position on AI detection tools, and what evidence standard do we apply before a score is treated as a finding?
Could we agree on a shared set of assignment-level AI-use labels so students see consistent expectations across sections?
One Class Period

The Class Conversation

Project these four questions. Make it a discussion, not a lecture, and keep it non-punitive.

  1. What AI tools are you using?
  2. What are you using them for?
  3. When has AI given you something wrong or misleading?
  4. What do you think you lose when AI does the thinking for you?
Every Discipline

The Litmus Test

After a student uses AI, ask them to explain their work without looking at the screen.
If they can, AI scaffolded them.
If they cannot, all they have is polish.

Adapted from Cheryl Strauss Einhorn, Harvard Business Publishing (see Sources)

Copy and Adapt

Syllabus AI Statements

Three starting points. Pick one, edit the brackets, check it against your college's policy before you publish it.

Level A · Not permitted

In this course, the use of generative AI tools (such as ChatGPT, Copilot, Gemini, or Claude) is not permitted on graded work unless I state otherwise for a specific assignment. The purpose is for you to build the underlying skills yourself. If you are unsure whether a tool is allowed, ask before you use it.

Level B · Permitted with disclosure

You may use generative AI tools in this course for [brainstorming, outlining, checking grammar, explaining concepts]. If you do, add a short note at the end of your submission stating which tool you used and how. You are responsible for the accuracy of anything you submit, including anything an AI tool produced. Undisclosed use will be treated as a violation of course policy.

Level C · Required and assessed

Some assignments in this course require you to use generative AI and then evaluate, correct, or improve what it produces. On those assignments, your grade is based on your analysis of the AI output, not on the output itself. Instructions for each assignment will state exactly what is expected.

Whichever level you choose, state it once in the syllabus and again on each assignment. Consistency is what students remember.

Paste on Any Assignment

Assignment-Level AI Labels

One line at the top of each assignment removes most confusion. Use the same four labels all semester.

AI-0

No AI

Complete this entirely on your own. No AI tools at any stage.

AI-1

AI for planning only

You may use AI to brainstorm or outline. Everything you submit must be written by you.

AI-2

AI with disclosure

You may use AI at any stage. Disclose which tool you used and how at the end of your submission.

AI-3

AI required

Use AI as instructed, then critique or improve its output. Your analysis is what is graded.

Instead of a Detector

If You Suspect AI Use

A short, non-accusatory script. It works whether the suspicion turns out to be right or wrong.

Step 1 · Ask about the work, not the tool

Invite the student to a short conversation and open with the content: "Tell me how you arrived at this argument." "Which source was most useful and why?" "Walk me through this section." You are looking for whether the understanding is theirs.

Step 2 · Ask about the process

"How did you approach this assignment?" "What did you start with?" "Did you use any tools along the way?" Asked plainly, most students answer honestly, and disclosure under your course policy may resolve it right there.

Step 3 · Decide based on what you heard

If the student can explain and defend the work, treat it as theirs. If they cannot, follow your college's academic integrity process using the conversation as your evidence, not a detector score. Either way, document what was asked and what was said.

Do not paste a student's work into an AI tool to "check" it. That shares their writing with a third party and rarely produces anything you can act on.

Tuned to Your Area

Find Your Discipline

One copy-ready prompt, one redesign example, and one idea to try. Every prompt works in ChatGPT, Copilot, Gemini, or Claude.

Engineering, Mathematics & Sciences

Prompt

Create a set of 8 practice problems on [TOPIC] at three difficulty levels. For each problem, also generate one intentionally flawed worked solution containing a common student error for the class to diagnose.

Redesign Example

Replace one problem set with an error hunt: students receive AI-generated solutions, find the mistakes, and defend the corrections at the board. Move 3 plus Move 4 in one assignment.

Try This

  • NotebookLM: upload a textbook chapter, generate a study guide and an audio overview students can listen to on their commute.
  • Ask any AI tool to build a simple interactive simulation for one concept, then project it and change variables live in class.
Humanities, Communications & Fine Arts

Prompt

Generate a 400-word essay responding to this prompt: [PASTE YOUR ESSAY ASSIGNMENT]. Then list the five weakest moves in your own essay: vague claims, missing evidence, generic phrasing. Format as a class handout for critique.

Redesign Example

Students compare their own draft against an AI-generated version in class: which makes the stronger argument, which is more specific, and why. Their voice becomes the lesson.

Try This

  • Staged submissions in your LMS: outline, draft, revision, each with a timestamp. Process becomes visible without any new tool.
  • In-class writing one day per unit. Thinking in real time, no devices.
  • Process portfolios over final pieces: sketches, iterations, and a short statement on what the student made that a machine would not have.
Nursing & Health Sciences

Prompt

Create a patient scenario for [CONDITION] appropriate for [COURSE LEVEL]. Then generate a draft care plan containing two clinically unsafe or unsupported elements. Do not identify them. Students will verify every element against protocol.

Redesign Example

AI drafts, students verify: the same workflow entering clinical practice. Grade the verification, not the draft. Then the two-minute oral: "Why is this intervention safe for this patient?"

Try This

  • Skills-lab demonstration stays the anchor assessment. Competency-based verification is AI-proof by design.
  • AI-generated case studies as pre-lab prep: students arrive having already interrogated a scenario.
Business & Professional Programs

Prompt

Draft a one-page business memo recommending a decision on [SCENARIO]. Include two plausible-sounding but weak or unsupported claims. Do not label them. I will have students find and fix them.

Redesign Example

Swap the generic case analysis for a local one: interview the owner of a business near campus about their hardest decision this year, then analyze what you heard. AI cannot make the phone call.

Try This

  • Prompt, then audit: every deliverable includes a paragraph on what the AI got wrong and what the student corrected.
  • For criminal justice and paralegal: AI-generated legal summaries with planted errors are the discipline-native version of Move 3.
Skilled Trades & Technical Programs

Prompt

Generate a troubleshooting checklist for [SYSTEM/EQUIPMENT] that a student technician would follow. Include two steps that are out of order or that skip a safety check. Students will identify the errors and explain why the sequence matters.

Redesign Example

Your discipline already has the strongest assessment model on campus: hands-on competency demonstration. Use AI for what it does well here, better written study materials, pre-lab quizzes, and scenario-based prep so students arrive ready for the shop floor.

Try This

  • AI-generated pre-lab safety scenarios: students diagnose the hazard before they touch the equipment.
  • AI-generated equipment procedures with planted errors: the same critical-reading skill certification exams test.
Social & Behavioral Sciences

Prompt

Generate a summary of [THEORY/STUDY] with citations. I will have students verify whether each cited source exists and actually says what the summary claims.

Redesign Example

The source-verification exercise: some AI citations will not check out, and discovering that firsthand teaches more about evidence than any lecture. Pair with a local observation or mini-interview assignment.

Try This

  • NotebookLM for dense readings: students upload the article, interrogate it with questions, then write from their own notes, not from the AI's summary.
Read the Research

Sources

Direct links to the documents behind the ideas on this page.

GPT detectors are biased against non-native English writersLiang, Yuksekgonul, Mao, Wu, Zou · Patterns, 2023 5 Ways to Protect Your Teaching Expertise When Using AICheryl Strauss Einhorn · Harvard Business Publishing Education, 2025 The Impact of AI on Learning AssessmentEDUCAUSE, 2026 Work Trend IndexMicrosoft and LinkedIn · annual employer and workforce data