Human-centred artificial intelligence for adaptive and accessible user interfaces in augmented reality educational platforms with AI-Agent Interaction

The field of education is changing very fast, and the artificial intelligence and augmented reality are influencing it. Still, the development of augmented reality systems that are inclusive to everyone is a challenging task. The existing architectures are often used as black boxes, thus ignoring the learners with disabilities. This study therefore engaged in all-round research on human centred artificial intelligence and how the technology can transform the augmented reality. The overall idea is to make a hard tool flexible learning, responsive and reachable. The research employed structured literature review methodology, which involved conducting a survey of academic evidence from a variety of literature databases, such as IEEE Xplore, ACM Digital Library, ScienceDirect, SpringerLink, and Google Scholar, including 2005-2025. It was observed that there is a real need for an interactive link between the sense of interaction and the adjustment of the interface in real-time. In addition, the research study revealed that multi-agent roles may improve the engagement of a student, although the system has to regulate these agent interrelationships to avoid cognitive overload. The most effective way to minimize automation bias was shown to be approval-based monitoring, which requires the automation system to generate a change suggestion and for the human to confirm the change before it is implemented. Through the paper, the authors concluded that the disadvantages of immersive learning are caused by the gap that exists between the capabilities of technology and human needs. The proposed research is to bridge this gap by proposing an agent-driven inclusive augmented reality (ADI-AR), a four-layer model that integrates pedagogical scaffolding, interaction sensing and human governance.