Recent advances in deep learning have significantly improved the accuracy of skin lesion classification models, supporting medical diagnoses and promoting equitable healthcare. However, concerns remain about potential biases related to skin color, which can impact diagnostic outcomes. Ensuring fairness is challenging due to difficulties in classifying skin tones, high computational demands, and the complexity of objectively verifying fairness, given the continuous and context-dependent nature of skin tone and the dependence of fairness conclusions on metric choice and subgroup representation. To address these challenges, we propose a fairness algorithm for skin lesion classification that overcomes the challenges associated with achieving diagnostic fairness across varying skin tones. By calculating the skewness of the feature map in the convolution layer of the Visual Geometry Group network (VGG) and the patches and the heads of the Vision Transformer (ViT), our method reduces unnecessary channels related to skin tone, focusing instead on the lesion area. Application on VGG11 and ViT-B16, showed improved fairness metrics by 15-20% on average while maintaining accuracy and F1-score within 0.01 of the baseline. Additionally, the method reduced model size by 16% for VGG11 and decreased memory footprint for ViT-B16, without requiring skin tone labels at inference. Thus, the approach lowers computational costs and mitigates bias without relying on conventional statistical methods. It potentially reduces model size while maintaining fairness, making it more practical for real-world applications.

Artificial intelligence (AI) is increasingly being explored in health and social care to reduce administrative workload and allow staff to spend more time on patient care. This paper evaluates a voice-enabled Care Home Smart Speaker designed to support everyday activities in residential care homes, including spoken access to resident records, reminders, and scheduling tasks. A safety-focused evaluation framework is presented that examines the system end-to-end, combining Whisper-based speech recognition with retrieval-augmented generation (RAG) approaches (hybrid, sparse, and dense). Using supervised care-home trials and controlled testing, we evaluated 330 spoken transcripts across 11 care categories, including 184 reminder-containing interactions. These evaluations focus on (i) correct identification of residents and care categories, (ii) reminder recognition and extraction, and (iii) end-to-end scheduling correctness under uncertainty (including safe deferral/clarification). Given the safety-critical nature of care homes, particular attention is also paid to reliability in noisy environments and across diverse accents, supported by confidence scoring, clarification prompts, and human-in-the-loop oversight. In the best-performing configuration (GPT-5.2), resident ID and care category matching reached 100

LLMs have demonstrated strong language-learning and human-like response-generation capabilities, and they are increasingly used to support decision-making in high-risk sectors. However, their internal decision processes remain difficult to interpret, and their responses may lack transparency. The literature has explored numerous approaches to address transparency challenges in LLMs, including Neurosymbolic AI (NeSy AI). NeSy AI approaches were primarily developed for conventional neural networks and may not transfer directly to the distinctive characteristics of LLMs. Consequently, there is a limited systematic understanding of how symbolic AI can be effectively integrated into LLMs. This paper aims to address this gap by first reviewing established NeSy AI methods and then proposing a novel taxonomy of symbolic integration in LLMs, along with a roadmap to merge symbolic techniques with LLMs. The taxonomy organises the literature across four dimensions: (1) the stage of LLM development at which symbolic information is integrated; (2) the coupling mechanism; (3) the architectural paradigm; and (4) the algorithm-level or application-level perspective. The review identifies commonly used benchmarks, recent advances and important research gaps, and uses these findings to outline directions for future research. By highlighting the latest developments and notable gaps in the literature, it offers practical insights for implementing frameworks for symbolic integration into LLMs to enhance transparency.

Discourse on artificial intelligence (AI) governance is dominated by the language of trustworthy, responsible, and ethical AI, while the role of assurance remains under-specified and often conflated with these neighbouring concepts. In this paper, assurance is treated as an evidence-based process to justify confidence in the behaviour of the system. Clarifying what must be assured therefore requires a consistent vocabulary of AI risk domains. We conducted a PRISMA-ScR scoping review of peer-reviewed literature published since 2021 on Web of Science, Scopus, and IEEE Xplore, examining how assurance-relevant language is expressed across AI Assurance and adjacent governance literatures, including Trustworthy AI, Responsible AI, and Ethical AI. The author and index keywords were assigned to assurance principles and risk domains, resulting in a consolidated vocabulary comprising four assurance principles and ten assurance categories. The results show that, despite the proliferation of governance frameworks and broad normative language, the underlying risk concerns across these discourses are comparatively stable. The proposed vocabulary therefore provides a portable conceptual layer that distinguishes generalisable AI risk domains from the governance structures and assurance processes used to manage them. Rather than introducing a new assurance framework, the study clarifies a shared language that existing audit, risk and compliance mechanisms can use to reason about AI systems. A qualitative comparison with major regulatory and standards frameworks, including the EU AI Act, NIST AI RMF and ISO/IEC AI standards, indicates broad alignment with current governance priorities while also highlighting differences in emphasis between academic discourse and practitioner concerns.

Recently, the societal implementation of high-performance image classification models has expanded rapidly. While these models require vast amounts of training data to improve performance, securing sufficient real images is often impractical. As a means to compensate for this shortage, the use of synthetic data is becoming widespread. However, synthetic images are not necessarily equivalent to real images for training purposes. This study systematically analyzes the differences between two types of synthetic images created by different generation methods and real images from three perspectives: high-dimensional feature space, low-level statistics in color space, and the model training process. Furthermore, it experimentally verifies how synthetic data should be utilized by considering realistic data mixing scenarios. This enables the proposal of an evaluation and application strategy for performing preliminary assessments on synthetic images of unknown quality and safely incorporating them into training. This research aims to contribute to enhancing the reliability and safety of image classification models utilizing synthetic images.

Abstract Despite rapid progress in retrieval-augmented generation (RAG), existing systems still struggle to retrieve both fine-grained entity-level evidence and higher-level abstract knowledge from biomedical literature. Graph-based RAG offers a promising alternative to flat retrieval, and hierarchical Graph RAG extends this idea by enabling coarse-to-fine retrieval across multiple levels of abstraction. Its effectiveness, however, depends on hierarchy construction choices, uneven community connectivity, and query-dependent layer choice. These challenges are particularly important in literature-derived biomedical graphs, where the hierarchy is constructed automatically from documents rather than specified in advance. We introduce HAGRAG, a hierarchical attributed Graph RAG framework for PubMed-derived diabetes literature. HAGRAG constructs multi-level attributed communities, summarizes them across layers, indexes them using hierarchical navigable similarity search, and supports layer-aware filtering during retrieval. It further evaluates alternative community detection algorithms and layer-weighting strategies to determine how hierarchy design affects retrieval behaviour. Across comparisons with baseline systems, HAGRAG achieves the strongest overall performance, including accuracy of 0.80, relevancy of 1.00, recall of 0.54, and mean semantic similarity of 0.65. Answer quality is further assessed through NLI-based entailment checks and independent diabetes-clinician validation. The results show that Leiden produces the most coherent hierarchy, while retrieval quality remains sensitive to clustering and weighting choices. Overall, the findings show that effective hierarchical Graph RAG depends not only on graph augmentation, but on constructing, navigating, and weighting the hierarchy.

Edge Artificial Intelligence (Edge AI) enables the deployment of AI models directly on local edge devices, while such deployments are subject to strict resource constraints, particularly in clinical applications requiring local and timely inference. In such contexts, explainable artificial intelligence (XAI) can serve as a human-AI interface intended to support healthcare professionals' and patients' understanding of model predictions and informed decision-making. To fulfill this role, XAI method selection for TinyML deployments can be formulated as a human-centered multi-objective design problem that jointly considers qualitative stakeholder preferences, explanation quality, and proxy-based deployment cost. We propose a framework that integrates a large language model (LLM)-guided design interface that maps qualitative stakeholder preferences to candidate XAI methods, followed by deterministic feasibility filtering and Pareto-based optimization. The framework exposes trade-offs among explanation fidelity, stability, and proxy-based deployment cost while characterizing their implications for explanation quality and estimated deployment feasibility. A proof-of-concept evaluation on a skin lesion classification task illustrates how the framework systematically compares candidate XAI methods and identifies Pareto-efficient trade-offs. The present evaluation covers the computational selection stages, while physical MCU deployment and empirical human-expert validation remain outside the scope of this study.

Medical edge-AI systems must operate under a difficult tension: delivering reliable diagnostic inference while running on devices with limited battery capacity, memory, and compute. In dermatology, this problem is amplified by real-world image degradation caused by smartphone capture, poor lighting, blur, compression, and heterogeneous edge sensors. To handle these degraded inputs, deploying a heavyweight model can improve reliability, but it rapidly increases the energy burden on resource-constrained devices. Conversely, always using a lightweight model saves energy but may be less reliable on ambiguous or degraded inputs. This paper introduces EcoFair, a vertically partitioned inference framework for dermatology classification in which image and tabular inputs remain local to edge clients while only learned modality-specific representations are transmitted for server-side fusion. EcoFair first processes each sample using a lightweight image encoder and then decides whether additional heavyweight computation is necessary. Escalation is triggered when the lightweight prediction exhibits high uncertainty, a narrow separation between safe and high-risk classes, or elevated metadata-derived risk from patient age and lesion location. Across HAM10000, BCN20000, and PAD-UFES-20, EcoFair is evaluated using multiple lightweight–heavy backbone pairings to quantify the trade-off between energy consumption, diagnostic performance, and worst-group malignant-case recall. Results show that EcoFair can reduce per-sample image-inference energy by up to 68% relative to always using the heavyweight encoder, while selectively allocating additional computation under difficult data regimes to support inference reliability. Group-level analysis further shows configuration-dependent effects, with improvements in selected model–dataset settings and mixed behaviour in others.

Skin color has historically been a focal point of discrimination, yet fairness research in machine learning for medical imaging often relies on coarse subgroup categories, overlooking individual-level variations. Such group-based approaches risk obscuring biases faced by outliers within subgroups. This study introduces a distribution-based framework for evaluating and mitigating individual fairness in skin lesion classification. We treat skin tone as a continuous attribute rather than a categorical label, and employ kernel density estimation (KDE) to model its distribution. We further compare twelve statistical distance metrics to quantify disparities between skin tone distributions and propose a distance-based reweighting (DRW) loss function to correct underrepresentation in minority tones. Experiments across CNN and Transformer models demonstrate: (i) the limitations of categorical reweighting in capturing individual-level disparities, and (ii) the superior performance of distribution-based reweighting, particularly with Fidelity Similarity (FS), Wasserstein Distance (WD), Hellinger Metric (HM), and Harmonic Mean Similarity (HS). These findings establish a robust methodology for advancing fairness at individual level in dermatological AI systems, and highlight broader implications for sensitive continuous attributes in medical image analysis.

The proliferation of IoV technologies has revolutionized the use of transport systems to a great level of improvement in safety and efficiency, and convenience to users. On the other hand, increased connectivity has also brought new vulnerabilities, making IoV networks susceptible to a wide range of cyber-attacks. The contribution of this paper is the in-depth study of the development and evaluation of advanced machine learning (ML) models that detect and classify network anomalies in IoV ecosystems. Several classification models have been studied in our research to achieve high accuracy for discriminating between benign and malicious traffic. This work further harnesses Explainable AI (XAI) methodologies through the LIME framework for enhanced interpretability of models' decision-making processes. Experimental results strongly advocate the strength of Random Forest and XGBoost, proving to be better on the binary and multi-class classification tasks, respectively. Due to resilience, preciseness, and scalability these models are a practical choice in real-world IoV security frameworks. Explainability integrated not only strengthens model reliability but also closes the gap between performance and interoperability in vehicular networks.

LLMs have demonstrated highly effective learning, human-like response generation,and decision-making capabilities in high-risk sectors. However, these models remain black boxes because they struggle to ensure transparency in responses. The literature has explored numerous approaches to address transparency challenges in LLMs, including Neurosymbolic AI (NeSy AI). NeSy AI approaches were primarily developed for conventional neural networks and are not well-suited to the unique features of LLMs. Consequently, there is a limited systematic understanding of how symbolic AI can be effectively integrated into LLMs. This paper aims to address this gap by first reviewing established NeSy AI methods and then proposing a novel taxonomy of symbolic integration in LLMs, along with a roadmap to merge symbolic techniques with LLMs. The roadmap introduces a new categorisation framework across four dimensions by organising existing literature within these categories. These include symbolic integration across various stages of LLM, coupling mechanisms, architectural paradigms, as well as algorithmic and application-level perspectives. The paper thoroughly identifies current benchmarks, cutting-edge advancements, and critical gaps within the field to propose a roadmap for future research. By highlighting the latest developments and notable gaps in the literature, it offers practical insights for implementing frameworks for symbolic integration into LLMs to enhance transparency.

Fairness is a critical component of Trustworthy AI. In this paper, we focus on Machine Learning (ML) and the performance of model predictions when dealing with skin color. Unlike other sensitive attributes, the nature of skin color differs significantly. In computer vision, skin color is represented as tensor data rather than categorical values or single numerical points. However, much of the research on fairness across sensitive groups has focused on categorical features such as gender and race. This paper introduces a new technique for evaluating fairness in ML for image classification tasks, specifically without the use of annotation. To address the limitations of prior work, we handle tensor data, like skin color, without classifying it rigidly. Instead, we convert it into probability distributions and apply statistical distance measures. This novel approach allows us to capture fine-grained nuances in fairness both within and across what would traditionally be considered distinct groups. Additionally, we propose an innovative training method to mitigate the latent biases present in conventional skin tone categorization. This method leverages color distance estimates calculated through Bayesian regression with polynomial functions, ensuring a more nuanced and equitable treatment of skin color in ML models.

Harmful chemicals and metals from e-waste affects our environment and human health. The UK produced 24.9kg e-waste per person in 2016, 10kg more than the EU average. A typical laptop manufacturing produces 331kg of CO2 emissions. Early routine maintenance typically extends the use cycle of 3 to 5-year-old computers to at least 9 years. While circular economy (CE) frameworks present a promising approach to reducing this environmental burden, putting circularity into practice remains a significant challenge. A barrier to repairing and reusing electronic products is the lack of information and efficient decision systems to allow people to make informed choices. This paper introduces Circul8, a real-time monitoring and analytics tool designed to enhance data-driven circular asset management. Developed through a Knowledge Transfer Partnership (KTP) between the University of Hull and Techbuyer UK Ltd, Circul8 continuously collects non-intrusive telemetry data from core hardware components, including CPU, memory, storage, and battery usage. The platform features a lightweight agent for data collection, secure MQTT-based data transmission, a time-series analytics backend, and an interactive dashboard for visualising trends and detecting anomalies. Deployed across 70 laptop devices, Circul8 demonstrated its ability to identify performance degradation, thermal anomalies, and component stress-empowering timely and cost-effective maintenance decisions. The tool's open-source release further promotes replicability and cross-sector adoption, embodying co-creation principles and open innovation in sustainable IT.

Vertical Federated Learning (VFL) offers a privacy-preserving paradigm for Edge AI scenarios like mobile health diagnostics, where sensitive multimodal data reside on distributed, resource-constrained devices. Yet, standard VFL systems often suffer performance limitations due to simplistic feature fusion. This paper introduces HybridVFL, a novel framework designed to overcome this bottleneck by employing client-side feature disentanglement paired with a server-side cross-modal transformer for context-aware fusion. Through systematic evaluation on the multimodal HAM10000 skin lesion dataset, we demonstrate that HybridVFL significantly outperforms standard federated baselines, validating the criticality of advanced fusion mechanisms in robust, privacy-preserving systems.

Artificial intelligence (AI) is transforming the healthcare landscape globally. However, the unregulated use of AI tools and models raises concerns about patient safety, particularly in developing countries. This study aims to explore the necessity of a Responsible AI (RAI) framework in developing nations, focusing on Nepal and Ghana. By reviewing established AI guidelines, engaging with stakeholders from Nepal and Ghana, observing their existing digital healthcare systems for AI readiness, and conducting a survey, we identify key components around opportunity and ethical concerns for AI adoption in their healthcare system. Our analysis reveals a lack of advanced medical technology infrastructure, and the risks associated with unregulated AI use, which compromise patient safety. This advocates the urgent need for an RAI framework to ensure ethical AI integration tailored to local contexts, addressing risks related to AI model transparency, data bias, privacy, and safety. The findings provide practical insights for regulators in developing countries for designing, implementing, and governing AI-driven healthcare solutions using the localised RAI Framework.

Recent advancements in skin lesion classification models have significantly improved accuracy, with some models even surpassing dermatologists' diagnostic performance. However, in medical practice, distrust in AI models remains a challenge. Beyond high accuracy, trustworthy, explainable diagnoses are essential. Existing explainability methods have reliability issues, with LIME-based methods suffering from inconsistency, while CAM-based methods failing to consider all classes. To address these limitations, we propose Global Class Activation Probabilistic Map Evaluation, a method that analyses all classes' activation probability maps probabilistically and at a pixel level. By visualizing the diagnostic process in a unified manner, it helps reduce the risk of misdiagnosis. Furthermore, the application of SafeML enhances the detection of false diagnoses and issues warnings to doctors and patients as needed, improving diagnostic reliability and ultimately patient safety. We evaluated our method using the ISIC datasets with MobileNetV2 and Vision Transformers. Our code for the experiment is available on GitHub (https://github.com/Kuniko925/ExplainForSafe).

Generative AI, such as Large Language Models (LLMs), has achieved impressive progress but still produces hallucinations and unverifiable claims, limiting reliability in sensitive domains. Retrieval-Augmented Generation (RAG) improves accuracy by grounding outputs in external knowledge, especially in domains like healthcare, where precision is vital. However, RAG remains opaque and essentially a black box, heavily dependent on data quality. We developed a method-agnostic, perturbation-based framework that provides token and component-level interoperability for Graph RAG using SMILE and named it as Knowledge-Graph (KG)-SMILE. By applying controlled perturbations, computing similarities, and training weighted linear surrogates, KG-SMILE identifies the graph entities and relations most influential to generated outputs, thereby making RAG more transparent. We evaluate KG-SMILE using comprehensive attribution metrics, including fidelity, faithfulness, consistency, stability, and accuracy. Our findings show that KG-SMILE produces stable, human-aligned explanations, demonstrating its capacity to balance model effectiveness with interpretability and thereby fostering greater transparency and trust in machine learning technologies.

The detrimental effects of air pollutants on human health have prompted increasing concerns regarding indoor air quality (IAQ). The emergence of digital health interventions and citizen science initiatives has provided new avenues for raising awareness, improving IAQ, and promoting behavioural changes. The Technology Acceptance Model (TAM) offers a theoretical framework to understand user acceptance and adoption of IAQ technology. This paper presents a case study using the COM-B model and Internet of Things (IoT) technology to design a human-centred digital visualisation platform, leading to behavioural changes and improved IAQ. The study also investigates users' acceptance and adoption of the technology, focusing on their experiences, expectations, and the impact on IAQ. Integrating IAQ sensing, digital health-related interventions, citizen science, and the TAM model offers opportunities to address IAQ challenges, enhance public health, and foster sustainable indoor environments. The analytical results show that factors such as human behaviour, indoor activities, and awareness play crucial roles in shaping IAQ.

Large language models have demonstrated exceptional performance in multiple domains. However, practical deployment in the healthcare sector has distinctive challenges. These challenges include hallucination, inconsistency, explainability, reasoning, authenticity, and validity of information sources. Hallucinations in LLM often emerge due to unstructured and obsolete training data and the incompetence to upgrade the model data post-training. Retrieval-augmented generation (RAG) integration with LLM decision-making helps access real-time information from external resources. However, further improvements are needed to improve accurate response generation. A knowledge Graph is a structured data comprising nodes as entities and edges as relationships. When integrated with RAG, Knowledge Graph-based retrieval offers better contextu-ally relevant responses, traceability, and explainability of generated responses than RAG alone. This study proposes a novel knowledge graph-based RAG framework with a refined retrieval pipeline, robust chunking mechanism, and source traceability for enhanced diabetes-focused LLM. The retrieval pipeline integrates three robust retrieval strategies: keyword, graph, and vector. To ensure the authenticity of responses, a knowledge base focusing on diabetes is designed from validated sources. This verified knowledge base is preprocessed and converted to a knowledge graph to design A graph-based RAG pipeline. The empirical results demonstrate effective performance in diabetes-focused LLM, achieving a Rouge 1 score of 82.19%.

As machine learning is increasingly making decisions about hiring or healthcare, we want AI to treat ethnic and socioeconomic groups fairly. Fairness is currently measured by comparing the average accuracy of reasoning across groups. We argue that improved measurement is possible on a continuum and without averaging, with the advantage that nuances could be observed within groups. Through the example of skin cancer diagnosis, we illustrate a new statistical method that works on multidimensional data and treats fairness in a continuum. We outline this new approach and focus on its robustness against three types of adversarial attacks. Indeed, such attacks can influence data in ways that may cause different levels of misdiagnosis for different skin tones, thereby distorting fairness. Our results reveal nuances that would not be evident in a strictly categorical approach.