AI for Students: Ethical Boundaries

Is AI a friend or foe to students

The question is misleading. The correct answer: it depends on how you use it. Used well, AI accelerates learning; used poorly, it hinders learning and puts your academic career at risk.

In this article I lay out a practical roadmap for students through a four-level use framework, university policies, AI detection tools, and ethical use patterns.

The four-level use framework

Level 1: Learning/understanding (GREEN)

AI acts like a teacher for you. Have a topic explained, ask about a concept, request an alternative example, ask again about the point that confuses you.

Examples:

• "What is Bayes' theorem, explain it at an undergraduate level, with an example"
• "What does this piece of code do, explain it line by line"
• "What is the difference between the Copenhagen interpretation and the many-worlds interpretation of quantum mechanics"
• "How should I interpret this multiple regression result from this paper"

This use is 100% ethical. It is no different from taking a lesson from a live teacher. The only difference: AI is available 24/7, does not tire, and is patient.

Level 2: Study aid (YELLOW)

AI provides you with efficient study tools.

Examples:

• Summarize a long paper, check your understanding
• Generate flashcards from notes
• Have a quiz prepared that tests concepts
• Translate a foreign-language source, confirm your understanding
• Extract the outline of a topic, then explain it yourself

This use is also ethical in most cases. But be careful: thinking you understood something from the summary and not reading the original makes your learning shallow. An AI summary is a starting point, not an endpoint.

Level 3: Production aid (ORANGE)

AI produces a draft or skeleton for you, then you revise it and build structure on top.

Examples:

• Draw up an outline for homework, then write it yourself
• A draft of your literature summary for a thesis, then rewrite it with your own citations
• A slide skeleton for a presentation, your own content
• Boilerplate for code, write your own logic

This level is a gray area. Whether it is ethical or not depends on three things:

1. What the university/course policy says (is disclosure of AI use required, is it prohibited)
2. How much of the output is your own original contribution (is the AI draft 20% or 80% of the final text)
3. Whether you disclose it (do you state that you used AI)

Practical rule: If production starts with AI, your own original contribution in the result should predominate (>70% rewriting, additional sources, additional logic).

Level 4: Direct submission (RED)

Having AI do the entire homework/thesis/project and submitting it as-is.

Examples:

• "Write a 10-page paper on this topic" → copy → submit
• Have ChatGPT write a thesis chapter → change name/date → send to advisor
• Generate answers with AI from your phone during an exam

This is academic fraud. If caught:

• Failing grade in the course, repeat of the semester
• Disciplinary penalty, note in the record
• Revocation of the thesis/degree (if discovered later)
• Erasure from academic career

Do AI detection tools really work

Universities invested in AI detection tools throughout 2024-2026:

• Turnitin (default AI detector active)
• GPTZero
• Originality.ai
• Copyleaks

Accuracy: around 80-90%. That means there are 10-20% false positives and false negatives. False positives especially with:

• Non-native English writers (when someone whose native language is Turkish writes in English, the formal tone resembles AI)
• Scientific/technical texts (formulaic language)
• Very cleanly written paragraphs

For this reason an AI detection tool result alone is not a verdict. Generally a violation is decided when it combines with other evidence (plagiarism, an unexpected jump in quality, failure in oral defense).

Techniques for evading detection (not recommended, risky):

• Running AI output through a paraphrasing tool (Turnitin catches this too)
• Breaking the AI style with manual editing (takes time, insufficient)
• Mixing output from multiple AIs (still gets caught)

Right now "perfectly evading an AI detection tool" is difficult, and the technology on the defense side improves every month. It is not worth the risk.

University policies (2026 panorama)

Turkey:

• As of 2024 YOK published "Ethical Principles in AI-Assisted Content Production"
• Bogazici, ODTU, Koc, Sabanci, ITU published course-based policies
• Typical approach: disclosure of AI use required, direct copying prohibited

International:

• Harvard, MIT, Oxford: course-based policy, most with "ethical use + disclosure"
• Some universities banned it entirely (Sciences Po Paris did so for one term, then loosened it)
• Others left it completely open (some Stanford courses say "use any tool you want, just understand the answer")

Course-based variation: At the same university, course A may leave AI open while course B prohibits it. Ask your instructor directly, make sure of the permissions.

AI use in the thesis process

AI use in undergraduate theses, master's theses, and doctoral theses is read through the ICMJE/Vancouver framework.

Language/format (green):

• English translation (for writing in Turkish and producing an English version)
• Academic tone correction
• Citation format conversion (APA → Vancouver)

Structural help (yellow):

• Creating chapter drafts
• Summarizing literature (papers you read yourself)
• Writing code (R, Python statistics)

Declare it in the Methods or Acknowledgements section. Which model, which version, at which step.

Data analysis (orange):

• Was the analysis code written with AI or manually
• Is the table AI-generated or manual
• With what interrogation did you verify the AI suggestion

Clear disclosure + raw output in the supplementary + human revision are mandatory.

Content production (red):

• Having AI write the entire thesis and submitting it as-is
• Fictional references (AI made them up, you did not verify)
• Fabricating data results

If caught, degree revocation + academic retraction.

Practical recommendation:

• A clear paragraph in the Methods section:

"We used [Model name vX] for English language editing and to assist with code generation for statistical analysis. All AI-generated content was reviewed, revised, and verified by the authors. Citations were manually verified against PubMed. AI was not used to generate study results, data interpretation, or original arguments."

This sentence is sufficient for most Turkish and international journals.

Use in exams: strictly prohibited

Using AI is not permitted in any exam format; using AI in an exam is outright academic fraud.

Online exam:

• Webcam-monitoring systems (Proctorio, Respondus) track the screen and eye movement
• Browser lockdown
• Phone use not permitted
• Detection methods even so: keystroke pattern, response time analysis, unexpected quality

In-person exam:

• Phones, smartwatches prohibited
• Smart glasses are entering monitoring as a new risk

Consequence if caught: exam score zero, failing course grade, disciplinary investigation, in severe cases suspension from school.

Practical recommendation: Use AI to the fullest in your pre-exam study (Levels 1-2), do not even consider using it during the exam.

An efficient AI workflow for students

A suggested daily routine for maximum efficiency within ethical boundaries:

Morning (60 min):

• Request a conceptual summary from AI for the day's topic
• Have 3-5 check questions created, answer them yourself, compare with AI

After class (30 min):

• Paste your class notes into AI, ask it to "guess what I did not understand"
• Close the gaps the AI identified with reading

Evening (60 min):

• Have a quiz created, test yourself
• Return to the previous topic, spaced repetition

Weekend (2 hours):

• Produce a weekly summary and mind map with AI, revise it yourself
• A pre-reading list for the coming week

Thesis/project (5 hours weekly):

• Literature search (AI search + manual reading combo)
• Writing process (AI skeleton + your own writing)
• Data analysis (AI code + your own interpretation)

KVKK and privacy, the student side

Points to watch in student use:

• Do not paste thesis data (if it contains patient data) into free AI, this is a KVKK violation, an ethics committee violation
• Pasting classmates' personal information
• If the university has licensed institutional subscriptions, prefer them (ChatGPT EDU, Claude EDU)

Three common mistakes

Mistake 1: Submitting homework entirely written by AI. The assumption "my instructor is very busy, it won't be noticed" is dangerous. The Turnitin AI detector is active. Your classmates' level is known, and a jump in quality gets noticed. The cost of getting caught is high.

Mistake 2: Having AI do the literature search and not checking the citations. AI can fabricate fake papers. In a thesis with 50 references, 5 are fictional. When the reviewer / committee catches it: thesis withdrawal.

Mistake 3: Hours of AI chatting under the guise of "I'm using it to learn." Thinking you learned something by chatting with AI is a delusion. Without active testing, explaining to yourself, and writing, no lasting learning forms. AI is a tool; the one who actually learns must be you.

Serteser Consulting for students

Artificial intelligence literacy and ethical use guidance for undergraduate, master's, and doctoral students. Serteser Consulting:

• Hourly mentorship (B1 service): personal workflow design
• AI use in the thesis process and ICMJE Vancouver compliant declaration
• Systematic review or meta-analysis thesis design (PROSPERO compliant)
• Statistical analysis consulting (SPSS, R, Python)
• Small group workshop for 3-8 people (B3 service): university club, class event

The student discount is clarified in the first meeting. In a 15-minute free introductory meeting, we listen to your situation and draw up a practical roadmap. Focused on transferring know-how, not on selling.

To fit AI into your own workflow, you can look into the individual mentorship option.