Gains:
- Being able to distinguish where artificial intelligence saves real time in the special education workflow (draft writing, material production, summarization) and where decisions such as diagnosis, placement and goal setting are left to the expert and the IEP team, according to the task risk level.
- Ability to apply a discipline that verifies each artificial intelligence output by connecting it to the source, comparing it with the child's real profile, and passing it through an expert filter.
- Anonymizing student and family data within the scope of KVKK/privacy and gaining the habit of choosing a safe vehicle
A special education teacher's day goes by quietly under a pile of papers. One student's individualized education program (IEP) will be renewed, an adapted worksheet will be prepared for another, the behavior record of the third will be graphed, and a progress report will be written to the family of the fourth. All of these require knowing the child, creative thinking and expertise; But some of them are also repetitive, formal and time-consuming. Artificial intelligence (AI, or AI for short—computer systems that can generate text, recognize patterns, summarize and draft like humans) is a powerful helper in accelerating exactly this second group. It outlines an IEP goal, simplifies a text, and summarizes an observation note in minutes. But when used incorrectly, the same tool can carry a seemingly safe but inaccurate output into a child's education decision.
The first unit of this module is not a program introduction. Its purpose is to clarify where to put AI in your business and where not to put it at all. Because special education is a field that is both "workload-intensive" and directly "child-right-critical": a goal you write is what the child will learn during a year; The adaptation decision you make determines whether it will be understood in the classroom. Let's lay out the basic principle from the beginning: Artificial intelligence is an assistant, not an expert. Responsibility and final approval for diagnosis, placement, goal setting, and decisions affecting the child's future belong to the competent specialist, the IEP team, and the family.
Layers of special education business and the place of AI
It is useful to divide the special education specialist's job into three layers. The evaluation layer is the process of getting to know the child: observation, developmental history, reading the RAM report (educational evaluation report prepared by the Guidance Research Center). The planning layer is to produce IEP goals, materials and behavior plan based on this. The implementation and monitoring layer is daily teaching, data collection, and progress evaluation. AI can touch all three layers; but with a different authority in each. At the evaluation layer, AI only anonymously summarizes and edits, not diagnoses; is a powerful outline generator at the planning layer; In the monitoring layer, it summarizes the data and suggests interpretation, but does not make the decision.
Let's define a few basic terms from the beginning. IEP (Individualized Education Program) is a written plan prepared according to a student's needs and containing goals and measures. Outcome is the measurable learning outcome that the student is expected to achieve. Accommodation, facilitating the presentation without changing the content; modification is a simplification of the content itself. AAC (alternative and supplementary communication) is an approach that allows an individual with speech difficulties to communicate with symbols, cards or devices. We will explain these concepts one by one in the following units; For now, know this: AI gives you outlines and ideas for all of these concepts, but it does not make decisions.
The following table summarizes the role and risk level of AI by mission:
Quest
Role of AI
Risk level
Who approves
Material/worksheet draft
sketch generator
low
special education teacher
Text simplification, summary
Editor
low
teacher
Writing an IEP goal outline
Blueprint generator, customization required
medium
BEP team + family
behavior plan idea
recommendation generator
high
Expert + guidance
Progress review
data summarizer
medium
Teacher + expert
Diagnosis/placement decision
Not available
very high
RAM + authorized expert
Keep in mind the one line in this chart: as the stakes rise, the role of AI shrinks and human consent grows. AI has no place in the top row; Diagnosis and placement belong to legally and ethically authorized institutions.
Why "verification" is the heart of this business
Artificial intelligence language models seem confident in their answer, but they may not be sure. In technical language, this is called hallucination: it is the model's fabrication of non-existent information in a fluent sentence, just as if it were true. This is a serious trap in special education. The model may give you an imaginary “developmental norm” (“every 5-year-old child knows 200 words”); can invent a non-existent piece of legislation (a precise but false sentence such as "BEPs must be renewed every 6 months"); or he may suggest a "method" that has no scientific basis. Since he says them all with the same fluency, the only thing that separates right from wrong is your expertise and habit of verification.
The verification discipline consists of three steps:
- Link to the source: Not to the AI's memory for legislation, development norms, methods and criteria; Rely on official sources (MEB Special Education Services Regulation, RAM report, institutional procedures) and field literature. Use the AI to sketch out this information, not to remember it.
- Compare to the child's real profile: Compare every goal, material or suggestion the AI produces to the performance level of the real child you know. Is it heavy, light or suitable for him?
- Expert filter: Test with an expert eye whether the output is pedagogically, ethically and legally sound; Clean up wording that is stigmatizing, punishment-oriented, or unrealistic.
Caution: Putting an AI-generated IEP goal or behavior plan in the file without verifying it is like making an unsigned expert decision. Just because the output is fluent is not true; It is not suitable for children either.
Privacy: student data is private data
Student data (name, TR ID number, diagnosis, RAM report, family information, behavioral records) are protected under KVKK (Personal Data Protection Law) in Türkiye and GDPR in Europe; Health and disability information is special quality data. Pasting a student's name, identification number, diagnosis, or family history verbatim into a public AI tool is a serious violation. The rule is simple: anonymize data and don't share unnecessary. Instead of "Ahmet Yıldız, 9 years old, diagnosed with autism, protocol 2024-3312", write "9-year-old male student with limited verbal communication". If possible, choose corporate tools that have a data processing agreement and do not use your data in model training; Do not upload sensitive files from personal accounts.
Tip: Get into the habit of an “anonymization check”: before giving a text to the AI, replace names, ID/protocol numbers, address and school name with generic phrases like “student”, “family”, “school”. These 30 seconds protect you from a big risk.
three mini cases
Case 1 — Safe use. A special education teacher spent 4 hours per week creating a math worksheet adaptation for eight students. He gave the achievement and student level anonymously (no names) to the AI and requested drafts at three difficulty levels. AI produced drafts in 8 minutes; The teacher compared each one to the children's actual level, simplified two questions, and changed one visual. Duration: 50 minutes instead of 4 hours. AI gave the draft, the expertise remained with the teacher.
Case 2 — Unverified information trap. A teacher asked the AI, "How many digits can children with mild mental retardation add at most?" The AI gave a confident "Up to 3 digits" answer. The teacher wrote this as a boundary in the IEP. However, there is no such "rule"; Every child is individual and the real student was able to do the 4-step process. Mistake: Relying on a general "norm" made up by the AI rather than the child's actual performance.
Case 3 — Violation of privacy. One employee uploaded a list of the names, diagnoses and family phone numbers of all students in his class to a public tool and "generated a class report from those," he said. The data went to an external server and a KVKK complaint was raised. The correct way was to omit name, diagnosis, and contact information and work with only the anonymous level information (“three students at pre-reading level”).
Weak prompt / Strong prompt
Weak prompt:
Write an IEP goal for this student.
This claim is wrong: AI does not know the child, the performance level is not clear, there is no field and criterion. The result is a general sentence in the air that does not suit the child.
Powerful prompt:
Your role: an assistant helping the special education teacher. Student (anonymous): 8 years old, at the pre-reading level, partially establishing letter-sound relationships. Area: reading. Task: Write a DRAFT of 1 long-term and 3 short-term (measurable) goals. Include condition + observable behavior + criterion (percentage/trial) in each short goal. Do not make normative claims that you are not sure about; Assume this is a draft and will be adapted by the BEP team.
In this prompt, the role, anonymous profile, domain, output format, and "draft" expectation are clear. You still compare and adapt the output to the child's actual level.
Common mistakes
- Waiting for a diagnosis or placement decision from the AI. This authority belongs to RAM and authorized experts; AI cannot be used in this field.
- Sharing the child's identity. Name, ID, protocol and family information cannot be entered into any vehicle without anonymization.
- Relying on made-up “norms.” Substituting generalizations like "Every X child does Y" for the real profile.
- Using the draft as is. Carrying the unverified output verbatim into the IEP, family text, or behavior plan.
- Not paying attention to stigmatizing language. AI can sometimes produce negative labels such as “inadequate”, “failure”; These must pass through expert filtering.
In summary
Artificial intelligence is a powerful assistant in special education: it drafts, simplifies, summarizes, opens ideas and saves you time. But it does not make a diagnosis, it does not place it, it does not make the purpose certain. Each output is verified through three steps: link to source, compare with child's actual profile, expert filter. Student data is special quality data; It is not shared without being anonymized. As the risk increases, the role of AI becomes smaller and human responsibility grows. These principles are the backbone of the rest of the module.
Application task
Choose one of your own students and anonymize his profile (without name, ID, school name; just age, subject and performance level) and turn it into a text. Then ask the AI for an outline of the intent using the “Powerful prompt” template above. Examine the output with three verification steps (source, actual profile, expert filter) and correct at least two points. Write down in one sentence what you changed and why.
checklist
- [ ] I removed the student ID (name, ID, protocol, school) from the text.
- [ ] The role, anonymous profile and output format are clear in the prompt I gave to the AI.
- [ ] I compared the output to the child's actual performance level.
- [ ] I checked for any made-up "norm" or legislative claims.
- [ ] I did not leave the diagnosis/placement decision to the AI.
- [ ] I noted that final approval belongs to the specialist and the BEP team.