evidoria

← Back to browse

Good practice Imported

ChatGPT-Assisted IEP Goal-Writing — Two Randomised Trials for Special-Education Teachers in Samsun, Turkey

Türkiye · Samsun · See the Türkiye profile

Evidence: Randomised controlled trial Top 24% 60/100 · Ask Evidence Copilot about this practice

Two randomised trials at Ondokuz Mayis University (Turkey) gave special-education teachers a ChatGPT tutorial before writing IEP goals for children with autism. Goal quality rose significantly vs. controls; novices worked faster. Neither trial addressed data privacy.

30 teachers
Teachers in Trial 1 (Çarşamba preschool autism study)
22 teachers
Novice teachers in Trial 2

Details

Maturity
Pilot
Promoter
Ondokuz Mayis University, Department of Special Education (with Çarşamba District Directorate of National Education)
Period
2023–2024 (two randomised controlled trials)
Region (NUTS)
TR83
Keywords
special education, teacher training, higher education

Context

Individualized Education Program (IEP) goal-writing is one of special education's heaviest paperwork burdens, and poorly written goals can leave children with disabilities without a legally sound, measurable plan. Two randomised trials led by Salih Rakap at Ondokuz Mayis University in Samsun, Turkey, tested whether a short ChatGPT tutorial could help.

Activities

In the first trial, 30 special-education teachers working with preschoolers with autism, in partnership with the Çarşamba District Directorate of National Education, were randomly split into a ChatGPT-assisted group and a control group; both received standard SMART-goal guidelines, but only the ChatGPT group got a 15-minute tutorial on using it. In a second, separate trial published in the Journal of Early Intervention, 22 novice special-education teachers were randomised the same way and asked to write IEP goals for five children with autism, scored against the Revised IEP/IFSP Goals and Objective Rating Instrument.

Results

In both trials, the ChatGPT-assisted group produced significantly higher-quality, more comprehensive goals than the control group, and novice teachers using ChatGPT took significantly less time to do so.

Conclusions

Neither published trial reports whether teachers were instructed to strip identifying student information before entering prompts, and neither addresses FERPA- or GDPR-style data-protection safeguards for what is highly sensitive disability data, a gap flagged more broadly by U.S. commentary on AI-assisted IEP writing. Both trials are small (n=30 and n=22) and described by their own authors as preliminary; neither reports adoption beyond the research setting.

Implementation
Indicative cost
Low (< €50k)
Time to results
Short (< 1 year)
Staffing & skills
Salih Rakap, Ondokuz Mayis University, Department of Special Education, Çarşamba District Directorate of National Education (Trial 1 partner)

Conditions for success

  • Short (15-minute) ChatGPT tutorial paired with existing SMART-goal guidelines
  • Random assignment to ChatGPT-assisted vs. control groups enables a causal comparison

Common failure modes

  • Neither trial reports whether teachers stripped identifying student information before prompting, or addresses FERPA/GDPR-style data-protection safeguards for sensitive disability data
  • Small samples (n=30, n=22), single-institution, and described by the authors as preliminary with no evidence of adoption beyond the research setting

Commonly funded by

Erasmus+ National / regional programmes

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Similar practices you may find useful