An Explainable AI Framework for Detecting Misinformation using Semantic NLP Features and Attention-based Models

Bird sound recognition poses challenges due to complex, overlapping spectral patterns. We propose a novel framework that combines multi-scale texture-aware modeling with interpretable deep learning. Central to our method is the Directional Laplacian of Gaussian Network (DLoGNet), a convolutional architecture with learnable orientation and scale parameters to capture directional acoustic textures. Additionally, we design the Frequency Band Recalibrated Spectrogram (FBRS), which adaptively selects energy-dense sub-bands via wavelet packet decomposition. Experiments on real-world datasets show that our method outperforms conventional CNNs, RNNs, and attention-based models in both accuracy and class separability. Visualizations of learned filters and t-SNE embeddings support its interpretability and effectiveness. This study highlights the importance of directional and multi-scale features in acoustic signal understanding and offers a robust solution grounded in the principles of explainable artificial intelligence (XAI), providing interpretable directional features and visual insights into model decisions for bird species identification.

Extracting fine-grained event ontology knowledge based on supply chain management (SCM) related corpus and constructing knowledge graph (KG) has important guiding significance and knowledge support for the efficient implementation and development of SCM in manufacturing enterprises. Recently, research on the KG of SCM has not gained sufficient attention. This paper aims to propose an event logical KG construction approach for SCM. Specifically, a stacked BiLSTM entity recognition model based on the binocular attention mechanism is proposed, called the SBBAN model. Firstly, the character feature attention mechanism is used to infer the key information that contributes greatly to entity recognition in the text sequence. Character weighted features and character features splicing are used as new character input features. Then the deep semantic abstract features of text sequence are obtained by stacked BiLSTM. In addition, a self-attention mechanism is added to obtain the deep context relevant features. Experimental results show that the model shows better performance in in comparison with the state-of-the-art algorithms to complete the matching of event argument entities and offer knowledge support for SCM.

This study develops a literature-driven conceptual framework for energy-burden-aware underwriting within the context of U.S. residential decarbonization finance. It addresses a critical and growing disconnect between the policy-driven urgency to decarbonize residential buildings and the financial constraints faced by households most in need of retrofit investments. Notably, those households that stand to benefit the most from energy-efficiency improvements are often underserved by traditional underwriting models. Conventional lending frameworks prioritize stable income streams, established credit histories, and property-based collateral signals, whereas energy-burdened households—characterized by high utility costs, inefficient housing conditions, and constrained liquidity—exhibit affordability dynamics that are poorly captured by these models but are highly relevant for socially beneficial investment. Drawing upon interdisciplinary literature spanning energy burden, building decarbonization, alternative data underwriting, explainable artificial intelligence, and algorithmic accountability, this paper proposes an AI-enabled underwriting framework that integrates utility-bill histories, housing attributes, retrofit cost-benefit dynamics, and consumer protection considerations. Importantly, the framework emphasizes transparency and interpretability, avoiding reliance on opaque risk-scoring mechanisms. Methodologically, the study employs a structured integrative review of academic and policy-oriented sources, synthesizing insights into a multimodal framework encompassing data governance, feature engineering, model development, explainability outputs, fairness diagnostics, and financing pathway design. The contributions of this research are threefold. First, it reconceptualizes utility payment data from a signal of financial distress into a meaningful input for transition finance. Second, it establishes a direct linkage between household energy burden and underwriting design principles. Third, it outlines governance and accountability conditions under which AI-assisted retrofit financing can advance decarbonization objectives without reinforcing systemic exclusion. Given the absence of original empirical data, the findings are presented as a conceptual architecture, evaluation framework, and forward-looking research agenda rather than as empirical validation. The proposed framework is intended to guide future implementation and testing by financial institutions, utilities, public green banks, and housing policy stakeholders. paper develops a literature-based conceptual framework for energy-burden-aware underwriting in U.S. residential decarbonization finance. The study addresses a widening disconnect between the climate-policy imperative to decarbonize homes and the financial reality that many households most likely to benefit from retrofit investment are also the least well served by conventional underwriting. Existing lending models privilege stable income, established credit files, and property value signals, whereas the households facing high utility costs, inefficient housing, and limited liquidity often present precisely the affordability patterns that make retrofit investment socially valuable. Drawing on scholarship on energy burden, building decarbonization, alternative-data underwriting, explainable artificial intelligence, and algorithmic accountability, the paper proposes an AI framework that uses utility-bill histories, housing characteristics, retrofit economics, and consumer-protection rules to assess payment feasibility without reducing underwriting to opaque risk scoring. Methodologically, the paper uses a structured integrative review academic and policy sources and synthesizes them into a multimodal framework comprising data governance, feature design, model development, explanation outputs, fairness diagnostics, and financing-channel selection. The paper’s contribution is threefold: it reframes utility bills from a distress signal into a transition-finance input; it links household energy burden to underwriting design; and it specifies governance conditions under which AI-assisted retrofit finance may advance decarbonization without reproducing exclusion. Because no original dataset is analyzed, the results are presented as a conceptual model, evaluation logic, and research agenda rather than empirical claims. The framework is intended to support future testing by lenders, utilities, public green banks, and housing-policy practitioners.

Explainable AI (XAI) frameworks are becoming essential in many areas, including the medical field, as they help us to understand AI decisions, increasing clinical trust and improving patient care. This research presents a robust and comprehensive Explainable AI framework. To classify images from the BloodMNIST and Raabin‐WBC datasets, various pre‐trained convolutional neural network (CNN) architectures: the VGG, the ResNet, the DenseNet, the EfficientNet, the MobileNet variants, the SqueezeNet, and the Xception are implemented both individually and in combination with SpinalNet. For parameter analysis, four models, VGG16, VGG19, ResNet50, and ResNet101, were combined with SpinalNet. Notably, these SpinalNet hybrid models significantly reduced the model parameters while maintaining or even improving the model accuracy. For example, the VGG 16 + SpinalNet shows a 40.74% parameter reduction and accuracy of 98.92% (BloodMnist) and 98.32% (Raabin‐WBC). Similarly, the combinations of VGG19, ResNet50, and ResNet101 with SpinalNet resulted in weight parameter reductions by 36.36%, 65.33%, and 52.13%, respectively, with improved accuracy for both datasets. These hybrid SpinalNet models are highly efficient and well‐suited for resource‐limited environments. The authors have developed a dynamic model selection framework. This framework optimally selects the best models based on prediction scores, prioritizing lightweight models in cases of ties. This method guarantees that for every input, the most effective model is used, which results in higher accuracy as well as better outcomes. Explainable AI (XAI) techniques: Local Interpretable Model‐agnostic Explanations (LIME), SHapley Additive ExPlanations (SHAP), and Gradient‐weighted Class Activation Mapping (Grad‐CAM) are implemented. These help us to understand the key features that influence the model predictions. By combining these XAI methods with dynamic model selection, this research not only achieves excellent accuracy but also provides useful insights into the elements that influence model predictions.

Machine learning (ML) has emerged as a critical enabling tool in the sciences and industry in recent years. Today’s machine learning algorithms can achieve outstanding performance on an expanding variety of complex tasks–thanks to advancements in technique, the availability of enormous databases, and improved computing power. Deep learning models are at the forefront of this advancement. However, because of their nested nonlinear structure, these strong models are termed as “black boxes,” as they provide no information about how they arrive at their conclusions. Such a lack of transparencies may be unacceptable in many applications, such as the medical domain. A lot of emphasis has recently been paid to the development of methods for visualizing, explaining, and interpreting deep learning models. The situation is substantially different in safety-critical applications. The lack of transparency of machine learning techniques may be limiting or even disqualifying issue in this case. Significantly, when single bad decisions can endanger human life and health (e.g., autonomous driving, medical domain) or result in significant monetary losses (e.g., algorithmic trading), depending on an unintelligible data-driven system may not be an option. This lack of transparency is one reason why machine learning in sectors like health is more cautious than in the consumer, e-commerce, or entertainment industries. Explainability is the term introduced in the preceding years. The AI model’s black box nature will become explainable with these frameworks. Especially in the medical domain, diagnosing a particular disease through AI techniques would be less adapted for commercial use. These models’ explainable natures will help them commercially in diagnosis decisions in the medical field. This paper explores the different frameworks for the explainability of AI models in the medical field. The available frameworks are compared with other parameters, and their suitability for medical fields is also discussed.

An Explainable AI framework for credit evaluation and analysis

This article examines the integration of artificial intelligence in financial evaluation and the vital role of explainability in building trustworthy decision support systems. As AI transforms traditional financial evaluation from forecasting to portfolio management, the inherent opacity of sophisticated algorithms creates tension with the financial sector's transparency requirements. The discussion explores how Explainable AI techniques—particularly SHAP values and LIME—enable financial professionals to understand AI-generated insights while maintaining regulatory compliance. Through examining real-world implementations, the article demonstrates quantifiable benefits of explainable models in reducing false positives, improving analyst confidence, and accelerating regulatory approval. The evaluation extends to comprehensive Responsible AI frameworks encompassing fairness and bias mitigation, privacy-preserving techniques, and adversarial resilience mechanisms. The discussion addresses how generative AI assistants revolutionize document evaluation by automating summarization and data extraction while confronting critical security challenges, including prompt injection attacks, data leakage, and regulatory compliance complexities. The article emphasizes human-in-the-loop paradigms and tiered governance frameworks that successfully balance innovation with appropriate oversight, while examining real-time explainability challenges and monitoring requirements. Forward-looking perspectives on regulatory harmonization and the convergence of explainable, privacy-preserving, and robust AI systems demonstrate the evolution toward trustworthy financial AI implementations.

The increasing prevalence of cyber threats across Internet of Medical Things (IoMT) ecosystems poses critical challenges for safeguarding patient safety and data integrity, necessitating a dynamic, resilient intrusion detection system (IDS). In this work, we present a comprehensive machine learning framework for classifying cyberattacks in IoMT settings using biometric and network traffic data from the publicly available WUSTL-EHMS-2020 dataset. We conduct a unique comparative analysis using three paradigms: a Graph Neural Network (GNN) model to capture structural dependencies a Transformer deep learning model to capture contextual relationships and a lightweight baseline classifier, Logistic Regression. We undertook extensive data preparation, including label encoding, normalisation, and stratified sampling to maintain class balance. The Transformer achieved the highest overall classification accuracy in the IoMT ecosystem (93.5 parcent), outperforming both GNN (88.7 parcent) and Logistic Regression (92.8 parcent) across all evaluation metrics. Our research demonstrates the superior ability of attention-based models to identify complex threat patterns in heterogeneousIoMT data. Our study provides a reproducible benchmarking framework and lays the groundwork for future efforts related to hybrid modelling, explainable AI, and federated learning to improve the cybersecurity of Smart Healthcare Systems.

The quest to explain the output of artificial intelligence systems has clearly moved from a mere technical to a highly legally and politically relevant endeavor. In this paper, we provide an overview of legal obligations to explain AI and evaluate current policy proposals. In this, we distinguish between different functional varieties of AI explanations - such as multiple forms of enabling, technical and protective transparency - and show how different legal areas engage with and mandate such different types of explanations to varying degrees. Starting with the rights-enabling framework of the GDPR, we proceed to uncover technical and protective forms of explanations owed under contract, tort and banking law. Moreover, we discuss what the recent EU proposal for an Artificial Intelligence Act means for explainable AI, and review the proposal’s strengths and limitations in this respect. Finally, from a policy perspective, we advocate for moving beyond mere explainability towards a more encompassing framework for trustworthy and responsible AI that includes actionable explanations, values-in-design and co-design methodologies, interactions with algorithmic fairness, and quality benchmarking.

The explosion of Internet of things (IoT) devices has led to pretty much saturated network infrastructures, thus congestion often becomes severe, especially in applications where users require low latency, high throughput and real-time responses. Hence, traditional congestion control mechanisms like static routing protocols, Active Queue Management (AQM) and TCP variants are not suitable for this dynamic and heterogeneous IoT environment as they are reactive, rigid and cannot adapt to dynamic changes. The contribution is to investigate the potential of AI paradigms—such as ML, DL, RL and hybrid models- to provide Intelligent and proactive congestion Control in IoT systems. The paper compares the strengths and weaknesses of each method, such as RL’s generalizability and DL’s ability to capture patterns, as well as limitations in terms of scalability, interpretability, and computational resource requirements. Moreover, it emphasizes important research gaps in model generalization process, evaluation criteria and cross-platform fusion. Next steps in research discussions on the future of research take into account lightweight AI architectures, Explainable AI (XAI) frameworks, scalability of FL, and standard benchmarking data sets. The hybrid-AI congestion prediction model built in this work is validated across simulation tools and real-data sets and is observed to result in a significant decrease in latency, packet loss and energy consumption. This paper paves the way for scalable, intelligent and secure AI-based congestion management solutions for various IoT networks.

Web accessibility stands as a fundamental requirement for inclusive financial services because people with disabilities need to access online banking and trading platforms and financial services. The W3C (2025) maintains WCAG 2.1 as a global standard which provides web designers with accessible design guidelines while U.S. and EU regulators enforce compliance through ADA guidance and the 2024 EU Accessibility Act . The majority of websites contain accessibility barriers because research shows that 95% of 1,000,000 websites fail to meet accessibility standards and financial websites become targets for legal actions and regulatory investigations because of their inaccessible design . The process of testing web applications for accessibility through manual methods and rule-based systems (WAVE and axe) proves to be time-consuming and restricted in its capabilities. The research introduces FinAccAI as an AI system which performs automated accessibility testing for financial web applications. The system design implements rule-based verification for WCAG requirements including alt-text and label presence alongside machine learning and deep learning content analysis. The system uses a conceptual algorithm to extract HTML data and perform CNN/NLP-based visual analysis of charts before identifying accessibility problems. The system includes an Explainable AI (XAI) framework which generates human-friendly explanations for each detected violation through LIME/SHAP feature attribution methods . The research evaluates three testing methods (rule-based and ML and DL) through their ability to detect problems and their scope of detection while previous research demonstrates AI systems can decrease human work requirements . The AI-enhanced testing method demonstrates potential to achieve expert-level results according to our conceptual assessment which draws from research on finance site auditing .The system requires additional research to address model bias and generative errors which affect its performance while maintaining open-source model availability for reproducibility. The proposed system creates essential components for automated accessibility testing in financial services which helps in both regulatory compliance as well as business development. The research demonstrates that AI-based accessibility testing identifies more accessibility problems than traditional tools while delivering fast and dependable compliance verification results. The FinAccAI system enables large-scale intelligent accessibility audits which support national digital inclusion initiatives while helping institutions meet regulations and minimizing their exposure to risks and expanding financial services to 10 million American users. The research creates essential foundations for upcoming AI-based accessibility testing research while establishing open accessible solutions which will benefit the U.S. financial and technological sectors.

Android is a popular target for malware attacks due to its rapid proliferation, which emphasizes the need for efficient and understandable detection methods. A novel ensemble‐based framework that combines semantic feature engineering (SFE) and explainable artificial intelligence (XAI) is presented in this paper. It addresses the generalizability and transparency issues with current models. The purpose of combining low‐level Android permissions into high‐level semantic categories is to improve interpretability and reduce feature dimensionality, making the model easier to understand. Grid search optimization is used to assess a stacked ensemble classifier that consists of k‐nearest neighbors (KNN), random forest (RF), gradient boosting (GB), and support vector machine (SVM) models with a logistic regression (LR) as a meta‐learner on the TUANDROMD dataset. The model demonstrates strong class separation and generalization with a high test accuracy of 97.54%, with an F1‐score of 0.9847 and a ROC‐AUC of 0.9929. For explainability, permutation feature importance (PFI), partial dependence plots (PDPs), and a global surrogate decision tree (GSDT) are used. Results show that network and location permissions are the most important features. This study offers a robust and interpretable solution for Android malware detection. Its combination of SFE, ensemble learning (EL), and XAI makes it suitable for real‐world cybersecurity applications, especially on mobile devices.

Abstract: The problem of customer churn is a continuous challenge for profitability in the telecommunications sector, particularly given significant competition within the industry. Therefore, the ability to use advanced analytics to identify compressed subscribers in advance, and potentially retain them, is critical. This study describes an end-to-end explainable AI framework for predicting and intervening with telco churn. We used the IBM Telco Customer Churn dataset and simulated relevant behavioural features, so we could train our system via informative preprocessing (handling missing values and inappropriate encodings), feature engineering (where we devised aspects such as charge ratio, tenure, and service stability), and XGBoost modelling (where our modelling is optimized via stratified cross-validation and business-based target metrics) to monitor our predictions. Importantly, we provide SHAP-based explainability for every prediction, where our predictions contain both global and local explainability features to provide an actionable bridge between the power of algorithmic manipulation and business consideration for the customer. Our prediction offers the functionality of a business logic layer allowing churn probabilities to be mapped to risk formulations, loyalty tier definitions, and ways to intervene with the customer prior to their leaving as well as knowing later via a secure Flask-based web app with live analysis and audit of actions performed. Our experiment produced strong prediction values (AUC-ROC value of 0.95; retention of churn class recall up to 0.93), consistent with feature ranking providing interpretably relevant information, and identifiable space for practical business consideration. In conclusion, we have provided what we believe to be a scalable and transparent model for interpretable churn management with an aim of providing timely, informed decisions and accountability to operate the retention strategy in the telecommunications environment. Index terms: Customer churn, Churn prediction, Telecom retention strategy, Explainable AI, XGBoost, SHAP (SHapley Additive exPlanations).

Discovering injury severity risk factors in automobile crashes: A hybrid explainable AI framework for decision support

This article explores the emerging paradigm at the intersection of behavioral science and data operations in artificial intelligence systems. As organizations progress beyond technical implementations toward human-aligned AI, a sophisticated synthesis is developing that integrates psychological insights with advanced data frameworks. This integration transforms how AI systems are designed, deployed, and optimized across industries by incorporating the understanding of cognitive biases, decision heuristics, and contextual influences. The article examines how real-time behavioral feedback loops, human-in-the-loop frameworks, and behavioral metrics are revolutionizing traditional data operations. Strategic applications across real estate, financial services, and technology sectors demonstrate how this convergence creates adaptive ecosystems that enhance user experience while driving operational excellence. Implementation challenges including ethical considerations, privacy implications, organizational readiness, and measurement complexity, are analyzed alongside emerging trends such as explainable behavioral AI, cultural adaptation, longitudinal optimization, and collective intelligence frameworks. This human-centered approach represents a critical evolution in AI development—one that balances technological capability with psychological resonance to create systems that not only perform efficiently but meaningfully align with human decision-making processes.