Conference Agenda

Technology has long elicited diverse attitudes and responses among different groups, often focusing on fundamentally different perceptions between those owning or controlling it and those being controlled by, or otherwise subjected, to it. This continues to be the case with artificial intelligence (AI) in its various forms and applications, and further applies to divergent views on the ethical use of generative AI (GenAI) for those employing it in different roles. This paper provides a comprehensive, and very current, background review of these asymmetries using academic, pseudo-academic and media sources.

The paper then explores whether perceptions of ethical GenAI use in higher education differ systematically based on the user’s or evaluator's role. It describes an international empirical ethics survey of attitudes towards the use of GenAI in universities and other higher education establishments. Two very similar questionnaires were produced to collect ethical perceptions on student and staff (faculty) use of GenAI. Each was completed by both staff and students, in several countries, giving four sets of data (‘student on student’, ‘student on staff’, ‘staff on student’ and ‘staff on staff’) to compare. Perceptions of ethical GenAI use in this environment do indeed prove to be asymmetric with regard to roles to a considerable extent.

The rapid expansion of artificial intelligence (AI) in education has created new opportunities for personalized learning, particularly in primary schools where foundational skills shape long-term academic outcomes. Despite this national commitment to digital transformation, primary teachers continue to face significant challenges in diagnosing learning gaps and producing timely, individualized remedial materials. This paper presents an AI-powered adaptive learning system that generates personalized remedial worksheets for Qatari primary students using student performance data. Adopting a Design Science Research methodology, the study develops a lightweight, Python-based, multi-agent system integrating rule-guided retrieval-augmented generation (RAG), deterministic performance classification, curriculum grounding, and automated critique-and-repair mechanisms. A mixed-method evaluation combining technical validation with teacher and student surveys was conducted across three use cases. Results indicate that the system reliably produces age-appropriate, curriculum-aligned materials, significantly reduces teacher workload, and improves timeliness of targeted instructional support. The findings demonstrate that a context-responsive, transparent AI system can meaningfully enhance personalized learning in primary education in Qatar, offering scalable insights for ethical and pedagogically grounded AI adoption in comparable educational contexts.

The integration of artificial intelligence (AI) in education has accelerated dramatically, with multi-agent systems (MAS) emerging as a promising paradigm for educational content generation. This systematic review synthesizes current research on MAS architectures and methodologies for generating educational content. From 285 initial records, 42 studies (2015– 2025) were analyzed. Our analysis reveals eight distinct MAS architecture types, with adaptive agent systems being most prevalent (52%) and large language model (LLM)-based systems (40%) showing significant growth post-2022. Seven primary content generation methodologies were identified, with personalized content generation dominating (76%), followed closely by adaptive generation (62%). Retrieval-augmented generation (RAG) emerged as an important technique in LLM-based systems (17%). Applications span four educational domains, with higher education representing 43% of studies. LLM-based approaches demonstrate superior natural language capabilities but face challenges in computational cost and potential hallucinations. RAG techniques effectively address accuracy concerns in LLM-based systems by grounding generation in verified knowledge sources. While MAS demonstrate significant potential for scalable, personalized content generation, challenges remain in evaluation standardization, ethical considerations, and system scalability. This review provides a comprehensive taxonomy of approaches, synthesizes effectiveness findings, and identifies critical gaps for future research in this rapidly evolving field.

Coming soon.