ACCESSIBILITY AND ARTIFICIAL INTELLIGENCE IN EDUCATION: A COORDINATED WORKFLOW OF PRE-MODELING AND TEACHER VALIDATION

Abstract

This study demonstrates the use of generative artificial intelligence tools in curricular flexibility tailored to inclusive education. The problem motivating this investigation is teacher overload in the manual production of accessible materials and the risk of uncritical use of generative AI, which can introduce biases and inconsistencies. It was investigated how the coordinated and sequential use of language models supports teaching routines through the pre-modeling of personalized pedagogical materials and adapted assessments. Methodologically, a tripartite automated workflow with functional task division across artificial intelligence platforms was developed, in which ChatGPT generated the initial version of the questions, Gemini refined communicational and linguistic accessibility, and DeepSeek performed restrictive quality control auditing, culminating mandatorily in critical pedagogical validation by the teacher. The results indicate that the sequential workflow of parameterized prompts has potential for the pre-modeling of activities for Special Education students based on data from individualized education plans. It is evident that assigning tasks by specialty acts as a safety filter and optimizes teacher planning time. It is concluded that the automated workflow operates strictly as an assistive resource to support pedagogical planning, with the teacher's technical domain and human analytical validation being mandatory, irreplaceable, and final requirements to mitigate biases and inconsistencies and to consolidate a fair, safe, and efficient teaching practice.