Gains:
- Being able to distinguish where in the teaching workflow (plan, material, question, feedback draft) artificial intelligence saves real time and where decisions such as grading and pedagogical judgment are left to the teacher, depending on the task risk level
- Ability to apply a discipline that verifies each AI output through the steps of linking it to the curriculum/resource, checking its authenticity, and passing it through pedagogical filtering.
- Anonymizing student data within the scope of KVKK and professional confidentiality and gaining the habit of choosing a safe vehicle
A teacher's week is filled with more writing tasks than it seems: lesson plans, worksheets, test questions, feedback notes, parent messages, group minutes, activity instructions. Most of this work is repetitive and time consuming; It steals the moments spent with the child that are truly precious. Artificial intelligence (AI) — computer programs that can understand human language and produce output such as text, plans, questions, etc. — is a powerful aid to speeding up this repetitive writing load. Throughout this module, we will learn to use AI not as a magic wand, but as a teaching assistant that produces drafts, gives ideas, and organizes. But let's put the most important sentence from the beginning: AI does not decide for the teacher; The teacher produces the draft and carries the decision and responsibility.
This first unit establishes three foundations: distinguishing where AI works in the teaching workflow and where it should stop; verifying each output; protecting student data. These three are the underlying security basis for all subsequent units.
Where does AI save time, where does the decision belong to the teacher?
It is useful to think of teaching duties in terms of risk level. Risk level is how much harm a task would cause to the student if done incorrectly.
Low-risk tasks that AI speeds up a lot: Drafting a lesson plan, producing three examples that explain a topic for different ages, writing the first version of a worksheet, drafting a parent bulletin, simplifying a text, brainstorming. Here AI solves the “blank page” problem; You correct and enrich it.
High-risk tasks where the decision is left to the teacher: Giving a grade to a student, making a judgment about a student such as "this child is weak/lazy/gifted", making a decision to violate academic integrity (e.g. "he/she cheated on this assignment"), guidance and psychological guidance, using expressions such as a definitive diagnosis to a parent. These decisions affect the child's future; The responsibility and final say always belongs to the teacher. AI offers at most a point of view here, it cannot make decisions.
Caution: "AI said" is not a justification. In areas affecting orientation, assessment, and student futures, AI output is not a substitute for approval from a competent expert (teacher). An unverified AI output is like an unverified rumor.
Why is verification necessary? Hallucination and bias
AI produces text not by "knowing" but by "predicting the possible word". That's why sometimes it gives extremely fluent but actually incorrect information. This is called a hallucination: a book that doesn't exist, an unspoken quote, a false date, a source that doesn't exist. For a teacher, this can be fatal because misinformation given to the student becomes a permanent misconception.
The second problem is bias: AI can repeat social patterns in the data it is trained on; may produce examples that are biased against certain gender, culture or level. For example, when offering differentiation, it may unfairly put a student in the "failure" mold.
Because of these two limits, we propose a simple validation discipline to apply to every output:
- Connect it to the source. Verify anything containing facts, dates, quotes, rules from a reliable source (syllabus, textbook, official document).
- Check again. Manually check a numerical or logical output (score distribution, time total).
- Pedagogical filter. “Is this appropriate for my class, this age, this culture?” Consult with your own professional knowledge.
- Take responsibility. The moment you approve it, that text is yours; Any mistakes are your responsibility.
Student data: privacy from the start
Student data is personal data; Information such as health, family situation, behavior and diagnosis are special personal data and require higher protection. In Türkiye, KVKK (Personal Data Protection Law) regulates this; Additionally, the teacher has an obligation of professional confidentiality. Writing a student's name, number, class, or sensitive status literally into a public AI tool means taking the data out of your control.
The rule is simple: Anonymize first. Remove identification information such as name-surname, school number, class-branch, parent name; Use general phrases such as "Student A", "12-year-old student". If possible, choose corporate tools provided by your school that have data processing assurance. We will deepen this principle in unit 9; Make it a reflex for now.
Tip: Before pasting a text into AI, search for name, number and organization name with "Ctrl+F". If you find it, replace it with general expressions before saving.
three mini cases
Case 1 — Made-up quote. A social studies teacher asked AI for "Ataturk's words about education" for a presentation. YZ gave 6 quotes; The teacher could not find the real source of 2 of them. If he had used it without checking, he would have taught 40 students a made-up word. Thanks to the verification step, he used only 4 quotes whose source was confirmed. The lesson was prepared 5 minutes longer, but incorrect information was prevented.
Case 2 — Time savings. An English teacher was preparing separate worksheets for 3 classes every week, spending approximately 3 hours on this. He started using AI as a draft generator: he had AI produce drafts that adapted the same outcome at three levels and corrected them himself. Time decreased from 3 hours per week to ~1 hour; He devoted the 2 hours he earned to individual feedback. The quality did not decrease because the final edit was the teacher's.
Case 3 — Return from confidentiality error. When a guidance counselor wanted to consult AI about a difficult student situation, he first wrote down the student's name and family information. He noticed the text before sending it, replacing them all with generic phrases like "a 13-year-old student" and "a domestic tension." The general pedagogical ideas he took worked again; but no personal data got out.
Copiable templates
1) Separating the task by risk level:
Your role: assistant to a teacher. Divide the following 8 teaching tasks into two lists: (A) low-risk jobs where the AI can confidently produce outlines, (B) high-risk jobs where the final decision must rest with the teacher. Write a one-sentence justification for each task. Tasks: [paste your own weekly tasks here]
2) Validation check prompt:
In the text below, mark each sentence that contains a fact, date, quote, source or numerical claim and make a note saying "must be verified". State clearly where you are unsure; fabricated information.Text: [paste draft]
3) Anonymization assistant:
In the following text, replace any personal information (name, number, grade, institution) that may identify the student, parent, or institution with general expressions (e.g., "Student A", "a student"). List what you changed.Text: [paste]
4) Pedagogical filter prompt:
Evaluate the following outline for a 6th grade, mixed-level Turkish public school classroom: name weaknesses in terms of age appropriateness, cultural appropriateness, and achievement alignment. Don't present it as the absolute truth, give it as a suggestion. Draft: [paste]
Weak prompt / Strong prompt
Weak prompt:
Write me a lesson plan.
There is no context: which course, grade, achievement, duration is not clear. AI is general, it will drop something that doesn't fit your class.
Powerful prompt:
Your role: an experienced science teacher.7. Grade "Force and Energy" unit, one lesson hour (40 minutes), outcome: "The student distinguishes between kinetic and potential energy with examples." Mixed-level class of 28 students, there is a projection. Give me a draft plan with a time distribution, with an introduction-body-conclusion flow. List separately the facts that I need to verify at the end.
Quest
Risk level
Role of AI
final decision
Lesson plan draft
low
Generates draft
teacher
First version of the working paper
low
Generates draft
teacher
Exam question + key
medium
Generates draft
teacher confirms
grading
high
Even the suggestion is limited
lonely teacher
Academic honesty decision
high
can't decide
lonely teacher
Common mistakes
- Giving the printout to the student without verifying it. Forgetting the risk of hallucinations is the most common and harmful mistake.
- Pasting personal data as is. Violating confidentiality for the sake of speed is a violation of KVKK and professional ethics.
- Making AI decide. Delegating decisions such as grades, diagnoses, and honesty judgments to AI does not eliminate responsibility, it only increases the risk.
- Contextless prompt. The desired output will not work without giving the class, level and achievement.
- Blindly trusting AI or rejecting it outright. Both are wrong; The correct thing is controlled use.
Tip: Think of each AI session like working with an intern: fast and diligent, but inexperienced. You never hand over a printout without signing or reading it.
In summary
AI is a powerful assistant that eases the teacher's repetitive writing burden; but due to hallucination and bias, every output requires verification. Separate tasks by risk level: use generously on low-stakes drafting work, keep high-stakes decisions like grading and pedagogical judgment to yourself. Always anonymize student data and choose a secure tool. Shortest rule: the blueprint is from the AI, the responsibility is from the teacher.
Application task
List 8 tasks you will do in your next work week. Mark each one as “low risk — AI can produce blueprints” or “high risk — up to me.” Then choose one of the low-stakes tasks and ask the AI for a draft with a contextual prompt, like the “Strong prompt” example above. Mark at least 3 facts that you need to verify in the output and write how you will verify it.
checklist
- [ ] Have I separated tasks by risk level?
- [ ] Do I have a plan to verify facts/quotes/numbers in the AI output?
- [ ] Have I anonymized personal data before giving the text to the AI?
- [ ] Am I sure that I keep the final decision (grade, judgment) with me?
- [ ] Have I included context in my prompt, such as role, grade level, outcome, and duration?