福利在线免费

Mr Gabriel Borg
Department of Artificial Intelligence, Faculty of 福利在线免费 and Communication Technology

Immersive training environments, such as those used in medical simulation, aviation, and defence, require adaptive, realistic, and contextually appropriate content to support individualised learning. Generative AI (GenAI) technologies offer the potential to create such content dynamically, including realistic scenarios, dialogue, and feedback. However, their black-box nature and susceptibility to hallucinations, bias, and incoherence pose significant risks in high-stakes training. At the same time, Explainable AI (XAI) methods promise to enhance trust and transparency by revealing the reasoning behind AI-driven decisions. Yet, existing XAI techniques are often not optimised for generative models or real-time, multimodal feedback required in immersive simulations.

There is currently a lack of robust frameworks that tightly connect generative models with explainable mechanisms, especially in ways that ensure pedagogical soundness, user trust, and contextual fidelity in real-time training environments. This research addresses this gap by systematically exploring how GenAI and XAI models can be selected, adapted, and integrated to support trustworthy, adaptive, and explainable learning experiences.

This doctoral research is motivated by the technical challenges associated with integrating GenAI and XAI into a unified architecture. GenAI models – unlike current adaptive training systems, which use technologies such as large language models (LLMs), diffusion models, and generative video frameworks – offer the ability to dynamically create adaptive, domain-specific training content. However, their use in sensitive applications demands a high level of control, transparency, and coherence. These issues are particularly critical in immersive simulations where AI-driven decisions directly influence user actions, cognitive load, and learning outcomes.

Explainable AI techniques aim to mitigate these concerns by offering interpretable insights into AI behaviour. Nonetheless, most existing XAI tools are built for static classification tasks and are not inherently designed for the temporal, adaptive, and multimodal demands of generative systems. As such, this research will investigate new methods for aligning real-time content generation with layered, multimodal explanations that are understandable and actionable for users.

By shifting the focus from educational theory to system-level integration of AI components, this research contributes to the design of novel technical architectures that support interpretability and trust without sacrificing adaptivity or immersion. The goal is to engineer a modular, scalable framework in which GenAI and XAI are co-developed to meet the complex demands of high-stakes, personalised simulation environments.

Ms Charlene Ellul
Department of Artificial Intelligence, Faculty of 福利在线免费 and Communication Technology

Notarial registers, such as those preserved at the Notarial Registers Archive in Valletta, form one of the richest continuous sources for understanding social, economic, and legal life from the fifteenth century onwards. Yet the challenges they present, such as palaeographic difficulty, linguistic variation, and limited cataloguing, are shared by many local and international archives that hold comparable handwritten Latin and mixed鈥憀anguage collections. These materials remain largely inaccessible, under鈥慸escribed, and difficult to analyse at scale, despite their immense historical value.

This research investigates how artificial intelligence (AI) and semantic technologies can be integrated into archival workflows to enhance the discoverability and interpretability of historical notarial records. It addresses the research question of how knowledge extraction and ontology鈥慸riven representation can support both archival description and content鈥憀evel analysis of notarial deeds. A domain鈥憇pecific Notarial Ontology is being designed that extends the standard Records in Contexts Ontology (RiC鈥慜) to enable the integration of catalogue鈥憀evel metadata with deed鈥憀evel knowledge in a unified semantic framework. Furthermore, the research is investigating methods to extract structured information from Latin notarial deeds while also creating a curated, expert鈥慳nnotated dataset. Evaluation will consider data across centuries, linguistic shifts, and evolving legal terminology.

This research contributes to the NotaryPedia project, a collaboration between the Notarial Archives Foundation and UM, funded by the Ministry for the National Heritage, the Arts and Local Government. By bridging traditional archival standards with AI鈥慳ssisted content analysis, the study demonstrates how archives can be reimagined as dynamic, interconnected knowledge infrastructures.

Mr Keith Azzopardi
Department of Artificial Intelligence, Faculty of 福利在线免费 and Communication Technology

Adaptive learning systems offer significant potential to improve learner engagement and outcomes, tailoring the learning experience to students’ interests, backgrounds, and performance. However, their implementation is often restricted by the content authoring bottleneck; Developing the necessary components - designing the educational domain model, creating instructional and assessment content, and linking the content back to the domain model – is a labour-intensive and burdensome process for educators. This can hinder the initial uptake of adaptive learning systems and make it harder to update course syllabi and content.

This research proposes an Agentic AI platform that addresses these challenges using Generative Artificial Intelligence. By leveraging a multi-agent architecture, the proposed framework automates content generation and its subsequent mapping to domain models, while retaining the educator as the central orchestrator and verifier. It builds on ACCLAIMED, a UM project involving the design and implementation of a content creation platform for lecturers.

The research aims to demonstrate that a unified approach to content creation and modelling can significantly reduce authoring time while maintaining pedagogical validity. This presentation outlines the architectural vision for integrating these functionalities into a single platform and the planned research activities required to validate the platform’s effectiveness. It features a discussion on interoperability standards in education, which would enable integration of the tool with a wide spectrum of Learning Management Systems. Furthermore, the presentation explores how the study addresses the regulatory requirements of the EU AI Act by prioritising explainability and human oversight in the deployment of educational AI tools.

Mr Reno Yuri Camilleri
Department of Artificial Intelligence, Faculty of 福利在线免费 and Communication Technology

The integration of Artificial Intelligence (AI) within Virtual Reality (VR) is transforming immersive experiences across education, training, and entertainment. Yet, this intersection also introduces complex ethical challenges, including concerns over data privacy, user manipulation, and algorithmic bias. The European Union’s AI Act, recognised as the first comprehensive legislative framework for AI, aims to mitigate such risks by classifying AI systems according to their potential impact and establishing corresponding regulatory obligations.

This project investigates the ethical dimensions of AI-driven VR through the EU AI Act, analysing its implications for developers, designers, and users alike. To address these challenges, a modular framework is proposed to facilitate the ethical deployment of AI in VR environments. The model comprises four key components: the VR Environment, AI Agent, User Interface and Interaction, and Ethics Monitoring and Evaluation. Each module embeds mechanisms such as transparency protocols, data protection strategies, and real-time ethical oversight. The design aligns with the EU AI Act’s risk-based categorisation to ensure compliance while promoting innovation.

By integrating principles of responsible AI design, the approach prioritises inclusivity, fairness, and user autonomy. Through this structured framework, the project establishes a roadmap for ethical AI development in VR, offering practical guidance for stakeholders committed to creating immersive experiences that adhere to both legal and ethical standards. In balancing technological progress with moral accountability, this work contributes to the broader discourse on AI governance within virtual environments.