Academic Integrity and Verification: Student Use of Artificial Intelligence

Gains:

• Being able to distinguish the concept of academic honesty and in which cases the use of artificial intelligence by students is conducive to learning and in which case it is plagiarism/cheating.
• Ability to design process-based evaluation and transparent usage rules, knowing that artificial intelligence detectors are unreliable
• Ability to position AI as a tool to teach ethical use, source verification, and critical thinking with students

AI tools are also in the hands of students. Tools that can print an assignment in minutes, solve a problem, or reproduce a composition from scratch have raised a new and inevitable question for teachers: When students use AI, when does this help learning and when does it deceive? And how do we manage this fairly? This unit revisits academic integrity (the principle of representing knowledge and effort honestly) in the age of AI. The most important message: AI detectors are unreliable; Honesty is protected not by the technology of punishment but by process-based assessment, transparent rules, and teacher judgment.

Use of AI: aid or deception?

The distinction becomes clear with the question of whether AI replaces learning or serves learning. If a student has AI write the composition from scratch and submits it as his own work, he has skipped the skill that was intended to be measured (writing), that is, he has cheated. But if a student brainstorms ideas, has the AI ​​check his own writing for grammar, or has the AI ​​explain a concept he doesn't understand, it's a learning tool — just like getting help from a dictionary or a teacher.

Two things are critical: transparency (did the student clearly state where and how they used the AI) and skill measured (what was the purpose of the task). The same use of AI may be legitimate in one mission and deceptive in another; Therefore, the teacher should set the rule in advance according to the task.

Attention: Saying "using AI = copy" is as wrong as saying "everything is free". If the student himself needs to demonstrate the skill you are measuring, AI cannot do that skill; but its use as a productivity and learning support is legitimate. You define the rule.

Why are AI detectors unreliable?

There are AI detectors on the market that claim "was this text written by AI?" These are dangerously unreliable. They make two types of errors: false positive (mistaking the real student text for an AI) and false negative (mistaking the AI ​​text for a human). In particular, the actual texts of students who write in a different, plain or formulaic way from their native language are often marked as "AI" — this unfairly penalizes disadvantaged students. Giving a student a zero based on a single detector score is both unfair and untenable.

The right approach is process-based evidence: the student's drafts, his notes, his process log, the knowledge he can demonstrate in class. If a student is unable to verbally explain the text he or she has handed in and does not recognize the source in it, this is a much stronger sign than a single detector. However, the final honesty decision rests with the teacher who weighs the evidence and the interview with the student.

Design that reduces deception

Design works better than punishment. Process evaluation: see the journey, not the product, by asking for drafts, plans, interim deliveries. In-class product: having critical writing done in class. Personal/local task: tasks that require personal experience that the AI ​​cannot general produce, such as "find a problem in your own neighborhood." High-level task: tasks that require analysis, defense, application rather than memorization. Transparency statement: asking the student to declare “here is how I used AI in this task” — this normalizes honesty.

Step by step: establishing an honesty policy

1. Define rules according to the task. In this task, where AI is allowed and where it is prohibited.
2. Ask for transparency. Let the student declare the use of AI.
3. Collect process evidence. Draft, plan, interim delivery.
4. Trust the process, not the detector. When in doubt, meet with the student and have the product explained.
5. Teach before you punish. The first goal is to teach ethical use.
6. Take charge of the decision. Judgment of integrity violations is evidence-based and rests with the teacher.

three mini cases

Case 1 — False positive disaster. A teacher gave a zero to the homework of a student who wrote plainly when the detector said "92% AI". The student and parent objected; The student explained the text fully orally and showed his drafts. The assignment was his own; The detector had produced a false positive. The teacher reversed his decision. Lesson: single-score punishment destroys justice.

Case 2 — Solution through process. A teacher divided the essay assignment into three intermediate submissions: topic selection, outline, final text. A student who rewrote with AI was inconsistent in mid-term submissions and could not explain his text. The teacher spoke to the student with process evidence, without ever needing a detector; The student made a new submission of his own work. The design came out strong through the detector.

Case 3 — Legitimate use. One student declared that he gave his own essay to AI for grammar checking, and he chose the corrections himself. The teacher deemed this a transparent, legitimate use; because the skill (idea development) being measured belonged to the student and the use was clearly declared. Transparency made usage honest.

Copiable templates

1) Task-specific AI usage rule:

Your role: assistant to the teacher. Write a clear AI usage rule for students for the following task: "[describe the task and the skill it measures]". Itemize what is allowed (e.g. coming up with ideas, checking grammar) and what is prohibited (e.g. rewriting the text). Include how the student will declare their use of AI.

2) Deception-resistant mission design:

For "[topic/outcome]", design an assignment that makes it difficult for the AI to produce from scratch: require personal experience/local context, add process deliverables (plan, outline), include an in-class component, and require higher-order thinking (analysis/defense). [X. be appropriate to the grade level.

3) Transparency declaration form:

Draft a short "AI use statement" for students to fill out with their assignments: which tool was used, at what stage, for what; which part was entirely the student's own work. In age-appropriate, non-accusatory language that normalizes honesty.

4) Critical AI literacy mini-lesson for students:

[X. Write a 20-minute mini-lesson plan for grade 1: "AI makes mistakes and can fabricate sources." It should include an activity that allows students to verify information produced by AI through an example. Purpose: to teach critical evaluation rather than blind trust.

Weak prompt / Strong prompt

Weak prompt:

Tell me, was this assignment written by an AI?

AI (and detectors) cannot tell this reliably; Score-based imputation is unfair and carries the risk of false positives.

Powerful prompt:

Your role: assistant to the teacher. Prepare a checklist of 6 questions to help me evaluate an essay assignment as one's own work with PROCESS evidence: is there an outline, does one recognize sources, can one explain it orally, is it consistent with in-class performance. Do not rely on detector score; Focus on meeting with the student.

Student's use of AI

Is it legitimate?

Why

Idea/brainstorming

Generally yes

He does the writing himself.

Grammar/spell check

If declared yes

support tool

Concept explanation

Yes

service to learning

Reprint the text and submit it

no

Skips the measured skill

Presenting the source as if you found it yourself

no

Transparency violation

Common mistakes

• Penalizing the detector score. Unfair and indefensible due to false positive; The real victim is the disadvantaged student.
• Banning AI altogether. Banning the legitimate use of learning is both impractical and hinders development.
• Leaving the rule vague. It is not enough to say "don't use AI"; Define what is free by task.
• Not evaluating the process at all. A task that demands only the finished product invites deception.
• Moving from teaching to punishment. The first goal should be to teach ethical use; Punishment is the last resort.

Tip: Make integrity a teaching topic, not a game of catch. Show students in person that the AI ​​makes mistakes and fabricates sources; When they understand the value of their own work, the need for copies decreases.

In summary

Academic integrity in the age of AI looks at whether the student uses AI as a substitute for learning or in service of learning; The distinction is determined by transparency and measured skill. AI detectors are unreliable and alone cannot be a basis for punishment; Integrity is maintained through process-based assessment, clear task-specific rules, deception-resistant design, and teacher judgment. The most powerful strategy is to teach ethical use before punishing.

Application task

For an upcoming assignment, use the "Task-specific AI usage rule" and "Transparency declaration form" templates to prepare a clear rule set and declaration form for students. Then reframe the same assignment with the “deception-resistant task design” template (add process delivery, personal context). Write in 4 items how you ensure honesty without resorting to the detector.

checklist

• [ ] Have I clearly defined the allowed and prohibited use of AI in this mission?
• [ ] Did I ask for a transparency statement from the student?
• [ ] Do I collect evidence of process (draft, interim submission)?
• [ ] Do I rely on interviewing the student instead of the detector in cases of honesty doubt?
• [ ] Did I aim to teach ethical use before punishment?