This technical report presents the REA System (Real-time Educational Analysis), a proprietary software platform developed at Centro Educativo Pugliautismo (Bari, Italy) for the quantitative monitoring of educational activities involving children and young adults on the autism spectrum. The system integrates computer vision (Face-API.js, MediaPipe), local artificial intelligence (Ollama/Gemma 3 12B), and structured data collection (SQLite) to generate automated periodic observation reports. It operates in a purely educational, non-clinical context under Italian Law 107/2015, with a fully local, GDPR-compliant, privacy-first architecture where no participant data leaves the center's premises. Empirical results are reported from a cohort of five participants (ages 6–19, autism support levels 1–3) totaling 309 educational sessions and 1,795 individual trials over monitoring periods of 1 to 18 weeks. Key findings include: Overall positive trend in 4 out of 5 participants Mean session scores ranging from 33.6% to 75.2% Global success rate of 62.8% across all trials 10 skills fully acquired, 17 in progress, 5 in regression Measurable increases in learner autonomy (decreasing prompt-dependency) Emotional profiling via validated computer-vision pipelines (66 analyzed sessions) Postural self-regulation tracking with neurodiversity-affirming interpretation The system's analytical engine produces 9-section automated reports covering: quantitative summaries, per-area trend analysis, per-activity learning curves (linear regression), autonomy evolution, variability profiling, and 4-week projections. Core principle: Results = (Observation × Data) / Time Both English and Italian versions of the full report are included as uploaded files.
Claudio Trevisi (Sat,) studied this question.