Aksesibilitas Peserta Didik Tunanetra Dalam Pembelajaran Berbasis Kecerdasan Artifisial: Kerangka Lima Lapis Hasil Analisis Dokumen

Abstract

Artificial intelligence creates opportunities for students with visual impairments to access learning materials in adaptable formats, obtain descriptions of visual information, and receive support that responds to their pace of interaction. The availability of an application, however, does not constitute accessibility. This study analyses the dimensions of accessibility, the barriers that emerge, and the conditions required for meaningful participation in such learning. A qualitative descriptive approach was employed through thematic document analysis of 23 sources, comprising 19 scholarly articles published between 2022 and 2026 and four policy or accessibility-standard documents. The analysis yields a five-layer framework: operability; informational equivalence and modality flexibility; output reliability and verifiability; user competence and pedagogical mediation; and governance and infrastructure. Artificial intelligence can strengthen learning independence, yet it also generates new barriers: unlabeled controls, loss of meaning in visual materials, inaccurate outputs, bias, privacy risks, and unequal access to devices. The framework is offered as an initial audit reference requiring field validation.