As artificial intelligence systems scale in dimensionality, nonlinearity, and distributed representation, demands for transparency increasingly exceed what can be epistemically justified. Opacity in contemporary AI is not merely a problem of secrecy or insufficient disclosure, but a structural condition of complex, high-dimensional models. Even fully specified and deterministic systems may resist stable human comprehension.This paper proposes a layered governance architecture for structural opacity in Explainable AI (XAI). The architecture integrates four interdependent levels: (1) ontological analysis of complexity-driven opacity; (2) apophatic interpretability as disciplined epistemic boundary-setting; (3) accountable explainability as stakeholder-calibrated communication constrained by prior limits; and (4) institutional governance mechanisms that operationalize monitoring, auditability, and refusal conditions under acknowledged uncertainty.Rather than treating opacity as a transparency deficit that needs to be eliminated, we argue that it must be governed. By reframing interpretability as bounded epistemic warrant and explainability as accountable communication, the proposed architecture stabilizes trust not through exhaustive understanding but through enforceable safeguards under explicit limits. The result is a systematic framework for governing structural opacity in Explainable AI.

As artificial intelligence systems become integral across domains, the demand for explainability, called eXplainable artificial intelligence (XAI), grows. Existing efforts have focused primarily on generating and evaluating explanations for black-box models, while a critical gap in directly enhancing models remains through these evaluations. It is important to consider the potential of this explanation process to improve model quality with feedback on training as well. XAI may be used to improve model performance while increasing model explainability. Under this view, this paper introduces Transformation-Selective hidden input evaluation for learning dynamics (T-SHIELD), a regularization family designed to improve model quality by hiding features of input, forcing the model to generalize without those features. Within this family, we propose XAI-SHIELD (X-SHIELD), a regularization for explainable artificial intelligence that uses explanations to select specific features to hide. In contrast to conventional approaches, X-SHIELD regularization seamlessly integrates into the objective function, enhancing model explainability while also improving performance. Experimental validation on benchmark datasets underscores X-SHIELD’s effectiveness in improving performance and overall explainability. The improvement is validated through experiments comparing models with and without X-SHIELD regularization, with further analysis exploring the rationale behind its design choices. This establishes X-SHIELD regularization as a promising pathway for developing reliable artificial intelligence regularization.

This manuscript provides a systemic and data-centric view of what we term essential data science, as a natural ecosystem with challenges and missions stemming from the fusion of data universe with its multiple combinations of the 5D complexities (data structure, domain, cardinality, causality, and ethics) with the phases of the data life cycle. Data agents perform tasks driven by specific goals. The data scientist is an abstract entity that comes from the logical organization of data agents with their actions. Data scientists face challenges that are defined according to the missions. We define specific discipline-induced data science, which in turn allows for the definition of pan-data science, a natural ecosystem that integrates specific disciplines with the essential data science. We semantically split the essential data science into computational, and foundational. By formalizing this ecosystemic view, we contribute a general-purpose, fusion-oriented architecture for integrating heterogeneous knowledge, agents, and workflows-relevant to a wide range of disciplines and high-impact applications.

This paper introduces the LoBOX (Lack of Belief: Opacity & eXplainability) ethics governance framework, a governance-centric approach for managing artificial intelligence (AI) opacity when full transparency is infeasible. While transparency-centric approaches treat transparency as the social/ideal goal and therefore opacity as a design flaw, LoBOX suggests opacity is a condition which should be ethically governed through role-sensitive explanation and institutional accountability. The LoBOX framework comprises a three-stage pathway: reduce accidental opacity, bound irreducible opacity, and delegate trust through institutional oversight. Integrating the stakeholder-sensitive explanation described in the RED/BLUE XAI model, which is aligned with emerging legal instruments such as the EU AI Act, LoBOX offers a scalable and context-aware alternative to transparency-centric approaches. LoBOX reframes trust as an outcome of institutional credibility, structured justification, and stakeholder-sensitive accountability, and it is designed to remain aligned with evolving technological contexts and stakeholder expectations while ethically governing opacity. In the end, to ensure responsible AI systems, LoBOX moves from transparency ideals to ethical governance, emphasizing that trustworthiness in AI must be institutionally grounded and contextually justified.

Human-AI collaboration (HAIC) increasingly mediates high-risk decisions in public and private sectors, yet many documented AI harms arise not only from model error but from breakdowns in joint human-AI work: miscalibrated reliance, impaired contestability, misallocated agency, and governance opacity. Conventional explainable AI (XAI) approaches, often delivered as static one-shot artifacts, are poorly matched to these sociotechnical dynamics. This paper is a position paper arguing that explainability should be reframed as a harm-mitigation infrastructure for HAIC: an interactive, iterative capability that supports ongoing sensemaking, safe handoffs of control, governance stakeholder roles and institutional accountability. We introduce co-explainers as a conceptual framework for interactive XAI, in which explanations are co-produced through structured dialogue, feedback, and governance-aware escalation (explain -> feedback -> update -> govern). To ground this position, we synthesize prior harm taxonomies into six HAIC-oriented harm clusters and use them as heuristic design lenses to derive cluster-specific explainability requirements, including uncertainty communication, provenance and logging, contrastive "why/why-not" and counterfactual querying, role-sensitive justification, and recourse-oriented interaction protocols. We emphasize that co-explainers do not "mitigate" sociotechnical harms in isolation; rather, they provide an interface layer that makes harms more detectable, decisions more contestable, and accountability handoffs more operational under realistic constraints such as sealed models, dynamic updates, and value pluralism. We conclude with an agenda for evaluating co-explainers and aligning interactive XAI with governance frameworks in real-world HAIC deployments.

Reproducing kernel Hilbert spaces provide a foundational framework for kernel-based learning, where regularization and interpolation problems admit finite-dimensional solutions through classical representer theorems. Many modern learning models, however – including fixed-architecture neural networks equipped with non-quadratic norms – naturally give rise to non-Hilbertian geometries that fall outside this setting. In Banach spaces, continuity of point-evaluation functionals alone is insufficient to guarantee feature representations or kernel-based learning formulations. In this work, we develop a functional-analytic framework for learning in Banach spaces based on the notion of featured reproducing kernel Banach spaces. We identify the precise structural conditions under which feature maps, kernel constructions, and representer-type results can be recovered beyond the Hilbertian regime. Within this framework, supervised learning is formulated as a minimal-norm interpolation or regularization problem, and existence results together with conditional representer theorems are established. We further extend the theory to vector-valued featured reproducing kernel Banach spaces and show that fixed-architecture neural networks naturally induce special instances of such spaces. This provides a unified function-space perspective on kernel methods and neural networks and clarifies when kernel-based learning principles extend beyond reproducing kernel Hilbert spaces.

Tools for automatic wood species identification are needed worldwide in order to support sustainable timber trade. This work explores the application of computer vision techniques to classify high-resolution macroscopic images of timber. The main challenge of this problem is that fine-grained patterns in timber are crucial in order to accurately identify wood species, and these patterns are not learned by convolutional neural networks (CNNs) trained on low resolution images. This work introduces the Timber Deep Learning Identification with Patch-based Inference Voting methodology, abbreviated TDLI-PIV methodology. This methodology exploits the concept of patching and the availability of high-resolution macroscopic images of timber in order to overcome the inherent challenges that CNNs face in timber identification. The TDLI-PIV methodology is able to capture fine-grained patterns in timber and, moreover, boosts robustness and prediction accuracy via a collaborative voting inference process. In this work we also introduce a new data set of marcroscopic images of timber, called GOIMAI-Phase-I, which has been obtained using optical magnification, and is openly published online in zenodo. Our experiments have assessed the performance of the TDLI-PIV methodology, including a comparison with other methodologies available in the literature, an exploration of data augmentation methods and the effect that the dataset size has on the accuracy of TDLI-PIV.

When artificial intelligence is becoming popular, the concern and the need for regulation are growing, besides other requirements of the data privacy. In this context, federated learning is proposed as a solution to data privacy concerns derived from different source data scenarios due to its distributed learning. The defense mechanisms proposed in the literature focus only on defending against adversarial attacks and maintaining performance, ignoring other important qualities such as explainability and fairness to poor quality clients, dynamism in terms of attack configuration and generality in terms of being resilient against different kinds of attacks. In this work, we propose RAB2-DEF, a resilient defense against byzantine and backdoor attacks which is dynamic, explainable and fair to poor clients via local linear explanations. We test the performance of RAB2-DEF on image datasets and defending against the byzantine and backdoor attacks considering the state-of-the-art defenses, and the result reveals that RAB2-DEF is a proper defense while also enhancing the other qualities toward trustworthy artificial intelligence.

In Artificial Intelligence, there is an increasing demand for adaptive models capable of dealing with a diverse spectrum of learning tasks, surpassing the limitations of systems devised to cope with a single task. The recent emergence of General-Purpose Artificial Intelligence Systems (GPAIS) poses model configuration and adaptability challenges at far greater complexity scales than the optimal design of traditional Machine Learning models. Evolutionary Computation (EC) has been a useful tool for both the design and optimization of Machine Learning models, endowing them with the capability to configure and/or adapt themselves to the task under consideration. Therefore, their application to GPAIS is a natural choice. This paper aims to analyze the role of EC in the field of GPAIS, exploring the use of EC for their design or enrichment. We also match GPAIS properties to Machine Learning areas in which EC has had a notable contribution, highlighting recent milestones of EC for GPAIS. Furthermore, we discuss the challenges of harnessing the benefits of EC for GPAIS, presenting different strategies to both design and improve GPAIS with EC, covering tangential areas, identifying research niches, and outlining potential research directions for EC and GPAIS.

This paper examines the ethical significance of habituation in human–AI interaction. It argues that repeated reliance on artificial intelligence can progressively reshape the capacities through which ethical agency is exercised. Drawing on Aristotelian virtue ethics, autonomy-supportive design, and accounts of techno-moral formation, the paper shifts attention from isolated outputs and individual decisions to the diachronic effects of recurrent interaction. It develops a normative framework organised around six analytically distinct but causally interconnected dimensions of ethical agency: human agency, moral judgment, ethical awareness, autonomy, human intention, and second-order agency. The analysis identifies interactional and institutional mechanisms through which these capacities may be weakened, including habitual default acceptance, sycophantic reinforcement, deceptive design, goal drift, opaque delegation, and reverse-centaur workflows. On this basis, the paper proposes ten actionable recommendations for AI design, education, and governance. These recommendations specify conditions under which AI can support reflective judgment, meaningful control, revisable delegation, and co-intelligent collaboration rather than habituated dependence. The central claim is that AI systems should not be evaluated solely by their technical performance or immediate decision quality. They should also be assessed by the human dispositions and capacities that their repeated use helps to cultivate.

The deployment of artificial intelligence (AI) in criminal justice is no longer experimental but routine. In Catalonia, this shift is embodied by RisCanvi, a risk assessment system used to guide decisions on imprisonment, conditional release, and rehabilitation. Initially built on expert-defined rules to standardize classifications and reduce bias, RisCanvi has gradually evolved into an opaque, data-driven infrastructure at the core of penal governance. A key turning point came in 2019, when the system moved from a transparent weighted-sum model to a logistic regression embedded in the prison administration’s digital architecture. While this change sought better predictive performance, it also concealed variables, weights, and thresholds from most professionals and all affected individuals. As a result, those whose liberty is at stake cannot meaningfully understand, question, or contest the algorithmic logic that shapes their trajectories. This paper proposes a governance and explainability framework for high-risk AI systems like RisCanvi. Building on the Trustworthy AI paradigm, and aligning with the EU AI Act and ISO/IEC 42,001, we introduce a three-layer model of intelligibility: algorithmic explainability, human-centred interface design, and stakeholder-informed narrative communication. The framework aims to enhance auditability, support procedural fairness, and restore agency and accountability in AI-assisted criminal justice decisions.

The rapid proliferation of Artificial Intelligence (AI) and the subsequent demand for large-scale data processing have intensified concerns regarding data privacy and security. Regulatory frameworks, such as the General Data Protection Regulation and the European AI Act, have further necessitated the development of privacy-preserving machine learning paradigms. Federated Learning (FL) has emerged as a robust solution, enabling collaborative model training without the centralized collection of raw data. However, many existing FL frameworks lack seamless support for edge devices, particularly the Android ecosystem, which constitutes a significant portion of the global mobile market. This paper introduces FLEXible, a modular and extensible FL platform designed to bridge the gap between simulation and real-world deployment. We detail the integration of an idiomatic Android library into the FLEXible ecosystem, leveraging modern communication protocols and software design patterns to facilitate efficient mobile-based federated training.

Deep learning models have an intrinsic privacy issue as they memorize parts of their training data, creating a privacy leakage. Membership inference attacks (MIAs) exploit this to obtain confidential information about the data used for training, aiming to steal information. They can be repurposed as a measurement of data integrity by inferring whether the data were used to train a machine learning model. While state-of-the-art attacks achieve significant privacy leakage, their requirements render them infeasible, hindering their use as practical tools to assess the magnitude of the privacy risk. Moreover, the most appropriate evaluation metric of MIA, the true positive rate at a low false positive rate, lacks interpretability. We claim that the incorporation of few-shot learning techniques into the MIA field and a suitable qualitative and quantitative privacy evaluation measure should resolve these issues. In this context, our proposal is twofold. We propose a few-shot learning-based MIA, termed the FeS-MIA model, which eases the evaluation of the privacy breach of a deep learning model by significantly reducing the number of resources required for this purpose. Furthermore, we propose an interpretable quantitative and qualitative measure of privacy, referred to as the Log-MIA measure. Jointly, these proposals provide new tools to assess privacy leakages and to ease the evaluation of the training data integrity of deep learning models, i.e., to analyze the privacy breach of a deep learning model. Experiments carried out with MIA over image classification and language modeling tasks, and a comparison to the state of the art, show that our proposals excel in identifying privacy leakages in a deep learning model with little extra information.

Evolutionary and bioinspired computation are crucial for efficiently addressing complex optimization problems across diverse application domains. By mimicking processes observed in nature, like evolution itself, these algorithms offer innovative solutions beyond the reach of traditional optimization methods. They excel at finding near-optimal solutions in large, complex search spaces, making them invaluable in numerous fields. However, both areas are plagued by challenges at their core, including inadequate benchmarking, problem-specific overfitting, insufficient theoretical grounding, and superfluous proposals justified only by their biological metaphor. This overview recapitulates and analyzes in depth the criticisms concerning the lack of innovation and rigor in experimental studies within the field. To this end, we examine the judgmental positions of the existing literature in an informed attempt to guide the research community toward directions of solid contribution and advancement in these areas. We summarize guidelines for the design of evolutionary and bioinspired optimizers, the development of experimental comparisons, and the derivation of novel proposals that take a step further in the field. We provide a brief note on automating the process of creating these algorithms, which may help align metaheuristic optimization research with its primary objective (solving real-world problems), provided that our identified pathways are followed. Our conclusions underscore the need for a sustained push towards innovation and the enforcement of methodological rigor in prospective studies to fully realize the potential of these advanced computational techniques.

Large Language Models (LLMs) are increasingly being deployed in high-risk domains where opacity, bias, and instability undermine trust and accountability. Traditional explainability methods, focused on surface outputs, do not capture the reasoning pathways, planning logic, and systemic impacts of agentic LLMs. We introduce TAXAL (Triadic Alignment for eXplainability in Agentic LLMs), a triadic fusion framework that unites three complementary dimensions: cognitive (user understanding), functional (practical utility), and causal (faithful reasoning). TAXAL provides a unified, role-sensitive foundation for designing, evaluating, and deploying explanations in diverse sociotechnical settings. Our analysis synthesizes existing methods, ranging from post-hoc attribution and dialogic interfaces to explanation-aware prompting, and situates them within the TAXAL triadic fusion model. We further demonstrate its applicability through case studies in law, education, healthcare, and public services, showing how explanation strategies adapt to institutional constraints and stakeholder roles. By combining conceptual clarity with design patterns and deployment pathways, TAXAL advances explainability as a technical and sociotechnical practice, supporting trustworthy and context-sensitive LLM applications in the era of agentic AI.

Despite the transformative potential of artificial intelligence (AI), small and medium-sized enterprises (SMEs) continue to face significant challenges in its effective adoption. While prior studies have emphasized strategic benefits and readiness models, there remains a lack of operational guidance tailored to SME realities—particularly regarding implementation barriers, resource constraints, and emerging demands for responsible AI use. This study presents an analysis of AI adoption in SMEs by integrating the technology–organization–environment (TOE) framework with selected attributes from the diffusion of innovations (DOI) theory to examine adoption dynamics through a dual structural and perceptual lens. Empirical insights from sectoral and regional contexts are also incorporated. Ten critical challenges are identified and analyzed across the TOE dimensions, ranging from data access and skill shortages to cultural resistance, infrastructure limitations, and weak governance practices. Notably, the framework is expanded to incorporate responsible AI governance and democratized access to generative AI—particularly open-weight large language models (LLMs) such as LLaMA, DeepSeek-R1, Mistral, and FALCON—as emerging technological and ethical imperatives. Each challenge is paired with actionable, context-sensitive solutions. The paper is a structured, literature-based conceptual analysis enriched by empirical case study insights. As a key contribution, it introduces a structured, six-phase roadmap methodology to guide SMEs through AI adoption—offering step-by-step recommendations aligned with technological, organizational, and strategic readiness. While this roadmap is conceptual and has yet to be validated through field data, it sets a foundation for future diagnostic tools and practical assessments. The resulting study bridges theoretical insight and implementation strategy—empowering inclusive, responsible, and scalable AI transformation in SMEs. By offering both analytical clarity and practical relevance, this study contributes to a more grounded understanding of AI integration and calls for policies, ecosystems, and leadership models that support SMEs in adopting AI not merely as a tool, but as a strategic enabler of sustainable and inclusive innovation.

Out-of-Distribution (OOD) detection is a critical task in machine learning, particularly in safety-sensitive applications where model failures can have serious consequences. However, current OOD detection methods often suffer from restrictive distributional assumptions, limited scalability, and a lack of interpretability. To address these challenges, we propose STOOD-X, a two-stage methodology that combines a Statistical nonparametric Test for OOD Detection with eXplainability enhancements. In the first stage, STOOD-X uses feature-space distances and a Wilcoxon-Mann-Whitney test to identify OOD samples without assuming a specific feature distribution. In the second stage, it generates user-friendly, concept-based visual explanations that reveal the features driving each decision, aligning with the BLUE XAI paradigm. Through extensive experiments on benchmark datasets and multiple architectures, STOOD-X achieves competitive performance against state-of-the-art post hoc OOD detectors, particularly in high-dimensional and complex settings. In addition, its explainability framework enables human oversight, bias detection, and model debugging, fostering trust and collaboration between humans and AI systems. The STOOD-X methodology therefore offers a robust, explainable, and scalable solution for real-world OOD detection tasks.

The rapid development of artificial intelligence systems has amplified societal concerns regarding their usage, necessitating regulatory frameworks that encompass data privacy. Federated learning (FL) is proposed as a potential solution to the challenges of data privacy in distributed machine learning by enabling collaborative model training without data sharing. However, FL systems remain vulnerable to Byzantine attacks, where malicious nodes contribute corrupted model updates. Although Byzantine resilient rules have emerged as a widely adopted robust aggregation algorithm to mitigate these attacks, their effectiveness drops significantly in high-dimensional parameter spaces, sometimes leading to poor-performing models. This paper introduces Layerwise Cosine Aggregation, a novel aggregation scheme designed to enhance the robustness of these rules in such high-dimensional settings while preserving computational efficiency. A theoretical analysis is presented, demonstrating the superior robustness of the proposed Layerwise Cosine Aggregation compared to the original robust aggregation rules. Empirical evaluation in diverse image classification datasets, under varying data distributions and Byzantine attack scenarios, consistently demonstrates the improved performance of Layerwise Cosine Aggregation, achieving up to a 16% increase in model accuracy.

This paper introduces LoBOX (Lack of Belief: Opacity & eXplainability) governance ethic structured framework for managing artificial intelligence (AI) opacity when full transparency is infeasible. Rather than treating opacity as a design flaw, LoBOX defines it as a condition that can be ethically governed through role-calibrated explanation and institutional accountability. The framework comprises a three-stage pathway: reduce accidental opacity, bound irreducible opacity, and delegate trust through structured oversight. Integrating the RED/BLUE XAI model for stakeholder-sensitive explanation and aligned with emerging legal instruments such as the EU AI Act, LoBOX offers a scalable and context-aware alternative to transparency-centric approaches. Reframe trust not as a function of complete system explainability, but as an outcome of institutional credibility, structured justification, and stakeholder-responsive accountability. A governance loop cycles back to ensure that LoBOX remains responsive to evolving technological contexts and stakeholder expectations, to ensure the complete opacity governance. We move from transparency ideals to ethical governance, emphasizing that trustworthiness in AI must be institutionally grounded and contextually justified. We also discuss how cultural or institutional trust varies in different contexts. This theoretical framework positions opacity not as a flaw but as a feature that must be actively governed to ensure responsible AI systems.

Deep learning models have an intrinsic privacy issue as they memorize parts of their training data, creating a privacy leakage. Membership Inference Attacks (MIA) exploit it to obtain confidential information about the data used for training, aiming to steal information. They can be repurposed as a measurement of data integrity by inferring whether it was used to train a machine learning model. While state-of-the-art attacks achieve a significant privacy leakage, their requirements are not feasible enough, hindering their role as practical tools to assess the magnitude of the privacy risk. Moreover, the most appropriate evaluation metric of MIA, the True Positive Rate at low False Positive Rate lacks interpretability. We claim that the incorporation of Few-Shot Learning techniques to the MIA field and a proper qualitative and quantitative privacy evaluation measure should deal with these issues. In this context, our proposal is twofold. We propose a Few-Shot learning based MIA, coined as the FeS-MIA model, which eases the evaluation of the privacy breach of a deep learning model by significantly reducing the number of resources required for the purpose. Furthermore, we propose an interpretable quantitative and qualitative measure of privacy, referred to as Log-MIA measure. Jointly, these proposals provide new tools to assess the privacy leakage and to ease the evaluation of the training data integrity of deep learning models, that is, to analyze the privacy breach of a deep learning model. Experiments carried out with MIA over image classification and language modeling tasks and its comparison to the state-of-the-art show that our proposals excel at reporting the privacy leakage of a deep learning model with little extra information.