Articles

Artificial Intelligence in Allied Health Practice: Current Evidence, Emerging Applications, And Implications for Patient Safety

/in /by

Artificial intelligence (AI) is moving from experimental research into routine healthcare workflows, including diagnostic support, clinical documentation, rehabilitation, laboratory medicine, medical imaging, patient education, and service management. Allied health professionals are increasingly involved in these AI-enabled pathways, yet evidence and implementation guidance are often concentrated in physician-led specialties. This narrative review synthesises current evidence on the use of AI across allied health practice and examines implications for patient safety, professional accountability, education, and health equity. Literature from peer-reviewed journals and authoritative guidance was considered, with emphasis on clinical applications, implementation barriers, bias, explainability, human-AI interaction, and generative AI. Current evidence supports potential benefits such as workflow support, image interpretation, pattern recognition, personalised rehabilitation, laboratory decision support, documentation assistance, and improved access to health information. However, technical performance alone does not establish clinical effectiveness or safety. Dataset shift, hidden bias, automation bias, inaccurate generated content, privacy risks, poor interoperability, insufficient local validation, and unclear responsibility can introduce new failure modes. Safe adoption should include a defined intended use, representative validation, transparent performance monitoring, user training, incident reporting, privacy safeguards, human oversight, and periodic re-evaluation after deployment. The review concludes that AI has meaningful potential across allied health practice, but sustainable adoption depends on evidence-based governance and continuous patient-safety monitoring rather than technology adoption alone.

From Gatekeepers to Gatekeeping Systems: Rethinking Journalism Theory in the Age of Algorithms and Generative AI

/in /by

Gatekeeping has remained one of the foundational theoretical approaches for explaining how information becomes news. From Kurt Lewin’s conceptualisation of gates and channels to David Manning White’s study of the newspaper wire editor, traditional gatekeeping theory located considerable power in human actors who selected, rejected and organised information before it reached audiences. The contemporary news environment, however, complicates this model. News selection is increasingly distributed across journalists, newsroom analytics, search engines, social media platforms, recommender systems and generative artificial intelligence. This paper re-examines gatekeeping theory through an interdisciplinary synthesis of journalism studies, platform studies, artificial intelligence research and communication theory. Drawing on a structured conceptual review of more than 29 major scholarly contributions, it argues that contemporary journalism is moving from identifiable gatekeepers towards interconnected gatekeeping systems. Within these systems, editorial judgement has not disappeared; rather, it interacts continuously with computational selection, platform infrastructures, audience behaviour and generative systems capable of producing, summarising and reorganising information. The paper proposes a five-layer model of contemporary gatekeeping comprising source, editorial, algorithmic, platform and generative gates. It further distinguishes between selection power, traditionally exercised by editors, and visibility power, increasingly exercised through technological infrastructures. The study concludes that gatekeeping theory remains valuable but requires conceptual expansion from the question of who controls the gate to the broader question of how interconnected human and computational systems determine what information becomes visible, credible and socially consequential.

Factors Influencing English Teachers’ Use of Generative Ai for Teaching Speaking Activities at a Language Center in Ca Mau

/in /by

Despite the potential of Generative AI (GenAI) to reduce Foreign Language Anxiety (FLA) in speaking instruction, its implementation in resource-constrained settings like a language center in Ca Mau province, Vietnam, remains underexplored. This study investigates the paradox of “digital native” instructors who possess high digital literacy but face infrastructural and pedagogical barriers in the Mekong Delta region of Vietnam. Employing an explanatory sequential mixed-methods design, quantitative data were collected from 58 EFL teachers via surveys, followed by semi-structured interviews. Utilizing the SAMR model as a diagnostic framework, the findings revealed the existence of a Substitution Plateau. Although teachers frequently used GenAI for administrative tasks (Substitution level, M = 3.91), its application for transformative, real-time voice interactions was notably limited (Redefinition level, M = 1.96). Stepwise multiple regression and thematic analyses demonstrated that infrastructural barriers (beta = -.52) and ethical concerns regarding “synthetic fluency” significantly inhibited advanced GenAI adoption, outweighing the positive influence of teachers’ TPACK competence. The study concludes that in peripheral educational settings, the primary limiting factor is not technological illiteracy, but rather a contextually driven “safety-first” pedagogical strategy. These findings challenge the universality of tech-integration models and provide localized implications for AI adoption in the Global South.

Generative AI in the Categorisation of Paediatric Pneumonia on Chest Radiographs

/in /by

Paediatric pneumonia is a leading cause of morbidity and mortality worldwide, necessitating accurate and timely diagnosis. This study explores the application of Generative AI for categorising paediatric pneumonia using chest radiographs. Leveraging deep learning techniques, including Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs), we enhance image quality, generate synthetic training data, and improve model generalizability. The proposed framework integrates AI-driven feature extraction, convolutional neural networks (CNNs), and attention mechanisms to improve diagnostic accuracy. The results demonstrate significant improvements in classification performance compared to traditional methods, with a focus on interpretability and clinical usability.