evidoria

← Back to browse

Good practice Imported

AccessiLearnAI — Transilvania University of Brașov's Accessibility-First AI E-Learning Platform

Romania · Brașov · See the Romania profile

Evidence: Descriptive / self-reported Top 73% 40/100 · Ask Evidence Copilot about this practice

Transilvania University of Brasov researchers built AccessiLearnAI, a proof-of-concept platform combining AI alt-text, summarisation and text-to-speech with WCAG-compliant design; early tests suggest better engagement, but authors call evaluation small-scale and unvalidated.

AccessiLearnAI — Transilvania University of Brașov's Accessibility-First AI E-Learning Platform

Details

Maturity
Pilot
Promoter
Transilvania University of Brașov (Dept. of Electronics and Computers)
Period
2024–2025
Keywords
higher education, web accessibility, generative AI, EdTech

Context

AccessiLearnAI, developed by researchers in the Department of Electronics and Computers at Transilvania University of Brașov, is a proof-of-concept e-learning platform built around 'accessibility-first' design rather than accessibility as an add-on. It combines WCAG 2.1/ARIA-compliant front-end markup with AI features — automatic alt-text generation for images via large language models, real-time content summarisation, translation, and Google Cloud text-to-speech — delivered through a Progressive Web App that also works offline.

Activities

The system routes AI-generated outputs (alt-text, summaries) through a human-in-the-loop review step before they reach learners, and the architecture is described as GDPR-aligned, with encryption and role-based access control. The team validated the platform with formative accessibility testing (the WAVE tool, screen-reader and keyboard-navigation checks) and comparative benchmarking against mainstream learning-management systems, reporting that 'early evaluations suggest' improved engagement, without publishing outcome or control-group data.

Results

The authors are candid that this is a small-scale, proof-of-concept evaluation: they flag reliance on third-party AI vendors (OpenAI, Google Cloud) as an operational risk, note only partial coverage of the full range of disability profiles, and call for large-scale validation before the platform could be considered classroom-ready.

Implementation
Indicative cost
Low (< €50k)
Time to results
Short (< 1 year)
Staffing & skills
Transilvania University of Brașov, Department of Electronics and Computers, research team

Conditions for success

  • Accessibility-first design from the outset (WCAG 2.1/ARIA compliance) rather than accessibility retrofitted later
  • Human-in-the-loop review of AI-generated outputs before learners see them
  • GDPR-aligned architecture with encryption and role-based access control

Common failure modes

  • Reliance on third-party AI vendors (OpenAI, Google Cloud) as an operational risk
  • Only partial coverage of the full range of disability profiles tested
  • No large-scale validation yet performed

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.

Attachments

Similar practices you may find useful