DeVry University

This paper presents a dynamic gamification architecture for an Extended Reality–Artificial Intelligence (XR-AI) virtual training environment designed to enhance STEM education through immersive, adaptive, and kinesthetic learning. As a supplement to online coursework, advanced realistic scenarios can be replicated in the virtual world through XR-AI based remote labs and collaborative activities for engineering training with options to manipulate variables in real-time and analyze the corresponding outcomes. Different XR platforms, such as Unity, can integrate AI to support adaptive learning environments. These environments enable students to manipulate real-time variables in simulated scenarios and receive immediate feedback, thus fostering deeper conceptual understanding. Preliminary design concepts and future implementation plans are presented to demonstrate the viability of this approach. Security and privacy are discussed via a defense-in-depth approach spanning client, middleware, and backend layers, incorporating AES-256 encryption, multi-factor authentication, role-based access control (RBAC) and GDPR/FERPA compliance. Factors such as sensor exploitation, perceptual manipulation, and virtual physical harm are identified, with mitigation strategies embedded at the design stage. Finally, the importance of maintaining learning integrity and ensuring educator oversight is emphasized.

Administrative access to managed Internet of Things (IoT) and communication devices is often evaluated through identity, session, and authorization controls without directly incorporating the current security posture of the managed device. This paper presents a modest risk-adaptive identity-assurance method, RAIA, that combines four signals: biometric continuity, QR/token session integrity, role-based authorization risk, and supply-chain/device posture. The method does not replace any of these mechanisms; it aggregates them into a bounded access-risk score and maps the score to allow, step-up/restrict, or deny decisions. Validation uses a fully reproducible synthetic experiment with 30,000 development-regime sessions and a separate 13,200-session stress regime. Thresholds are calibrated on held-out development data under an 8% benign-friction constraint. On the development test split, RAIA achieves AUROC 0.976, attack containment 0.911, benign friction 0.075, and balanced accuracy 0.918. Under distribution shift, the original threshold preserves high attack containment (0.932) but benign friction rises to 0.379, exposing a calibration limitation. Recalibrating only on benign stress-regime sessions restores benign friction to 0.080 while retaining 0.706 attack containment, slightly above a static four-factor baseline (0.695) and clearly above QR+RBAC (0.566) and biometric+QR (0.549). The results support a limited claim: cross-context access decisions can benefit from combining user identity, session integrity, authorization, and device posture, but threshold calibration remains deployment dependent. The study is simulation validated and does not claim production or hardware validation.

Este trabalho de conclusão de curso explora o uso de ferramentas digitais para avaliação formativa, com base em artigos dos Periódicos da CAPES. A pesquisa analisa como essas ferramentas podem ser integradas à avaliação formativa para otimizar o aprendizado e atender às necessidades da era digital. A metodologia é qualitativa, por meio de revisão bibliográfica, buscando identificar as principais ferramentas, descrever as práticas pedagógicas que as utilizam e examinar os desafios e oportunidades de sua implementação. A avaliação formativa é apresentada como um processo contínuo que visa aprimorar a aprendizagem e adaptar o ensino às necessidades dos alunos. A evolução tecnológica exige a adaptação das práticas pedagógicas e dos métodos de avaliação. Ferramentas como Moodle, Kahoot!, Mentimeter, Google Forms, Google Meet e Google Classroom são destacadas por promoverem engajamento, interatividade e feedback imediato, facilitando a autoavaliação e o acompanhamento do progresso do estudante. Apesar do potencial, desafios como infraestrutura tecnológica, conectividade e formação docente são mencionados. Contudo, as oportunidades superam os obstáculos, pois a tecnologia permite mudar o paradigma da aprendizagem, promover o pensamento crítico, a autonomia e o trabalho em equipe. Conclui-se que a integração estratégica da avaliação formativa e das ferramentas digitais é crucial para uma educação de qualidade e alinhada às demandas atuais.

The growing demand for immersive and adaptive educational tools has accelerated the use of extended reality (XR) technologies in engineering education. This work presents the design and development of an XR based framework for interactive electrical circuit simulation, aimed at enhancing student engagement and supporting diverse learning styles. The proposed architecture comprises of four layers: Presentation and Interaction layer, Simulation and Physics layer, Adaptation and Assessment layer, and Integration and Deployment layer. Unity XR provides the user interface for manipulating virtual components, while real-time circuit simulation is supported with haptic feedback and visual cues to aid kinesthetic learning. Learner actions are logged for performance analysis, and an adaptive assessment mechanism provides tailored feedback based on pre-defined thresholds. A cloud backend manages analytics, user profiles, and content synchronization, with support for multi-user interaction through middleware APIs. Initial implementation demonstrates the feasibility of accurate component modeling using C# scripting within Unity XR, establishing a scalable foundation for more complex educational scenarios.

This model proposes a Modified Item Response Theory (M-IRT) framework, extending the traditional Two-Parameter Logistic (2 PL) model by redefining the discrimination parameter as a function of multiple contextual variables. While conventional Item Response Theory (IRT) models primarily account for item difficulty and learner ability, M-IRT incorporates additional passive factors that influence response behavior and learning outcomes. Specifically, the model integrates: (i) response time, (ii) encoded learning style, (iii) lesson duration, (iv) instructor reputation, and (v) participation score- each contributing to a dynamic discrimination slope. The paper presents the mathematical formulation of the model, supported by a computational algorithm designed to estimate these parameters effectively. The proposed framework offers a more nuanced understanding of learner performance and holds promises to enhance adaptive assessment and personalized instruction in the educational environment.