This article presents a reflective analysis concerning the evolution of the “Living Classroom” concept, traditionally associated with rural settings, and its re-conceptualization within urban socio-ecological systems. By synthesizing global scientific trends and contemporary research, the study explores how metropolitan green infrastructures—such as urban forests, biocultural canals, and school gardens—function as high-complexity laboratories for Inquiry-Based Learning (IBL). The methodological approach combines a PRISMA-inspired systematic qualitative synthesis with a bibliometric analysis of 2,732 open-access articles indexed in the Scopus database (2015–2026), applying the Authors–Journals–Contributions (ARA) methodology and VOSviewer for scientific mapping. The findings reveal the scientific leadership of China and several European countries, a predominance of publications in Q1 journals, and four major thematic clusters related to sustainable development, human health and education, interdisciplinary knowledge integration, and socio-ecological learning environments. The reflection argues that the urban landscape is not a substitute for nature but a primary pedagogical mediator that enhances cognitive development, health, and environmental citizenship. Furthermore, the discussion integrates findings from international research to propose a necessary shift in natural science education: the city can be understood as a living laboratory where scientific inquiry bridges the gap between urban development and ecological literacy. Building on this synthesis, the study develops a conceptual framework that addresses the phenomenon of environmental blindness by reconceptualizing urban living classrooms as pedagogical mediators that connect scientific inquiry with everyday socio-ecological experiences. The proposed framework supports the pedagogical reinterpretation of riverside territories and urban ecosystems as authentic contexts for scientific inquiry. It also provides a theoretical reference for future educational research on inquiry-based science education in urban contexts.

Obstructive Sleep Apnea (OSA) is a well-known sleep disorder that can lead to major health consequences if
left untreated. The traditional diagnostic method, polysomnography, is precise but costly and labor-intensive. This research presents an Evolutionary Feedforward Neural Network (FFNN) optimized using a Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and the Crow Search Algorithm (CSA) to diagnose OSA. We utilized a dataset comprising demographic and physical measurements. The adopted features include age, sex, weight (lb), height (in), neck circumference, body mass index (BMI > 30), (Neck > 17), airway evaluation via the modified Friedman grade (MF > 2), the Berlin apnea-hypopnea questionnaire (BAN), and the identification of apnea as indicated by the apnea hypopnea index (AHI ≥ 15). We used this methodology to optimize the FFNN’s weights and biases, improving generalization and avoiding local minima that traditional backpropagation can trap the model in. The results indicate that GA, PSO, and CSA achieved competitive, comparable performance, with PSO showing a marginally better performance across most evaluation metrics, rather than a significant superiority of any single optimization method.

Field service management (FSM) has been faced with the need to efficiently assign tasks to technicians in field service operations (FSOs). The problem of inefficient task allocation has led to either an over utilization or an under utilization of the organization’s workforce and resources across cities and states. This study focuses on developing a resource-efficient multi-criteria algorithm (Proximity-Skill-Priority, PSP algorithm) for optimized field service task management. A mathematical modelling approach was used to design a unified score that takes into consideration the proximity of the technician to the task location, the skillset of each technician and the priority level of the task at hand while ensuring effective workload balance to ensure unbiased task assignment. The study used adaptive weighting coefficients to ensure that real-time adjustments are made when there are varying time conditions. The model was evaluated through a simulation experiment and benchmarked against three single-rule baselines: Proximity-First, Skill-First, and Priority-First assignment. The performance of the algorithms was measured in a 100-run simulation conducted on a synthetic dataset that is parameterized with the attributes of technicians, task priorities, and geospatial information.  The lowest average response time and travel distance were achieved by Proximity-First and Priority-First rules, whereas Skill-First produced the highest first-time-fix-rate surrogate at the cost of substantially higher travel and response values. PSP produced a more balanced trade-off: it improved first-time-fix performance when compared to the proximity-first and priority-first heuristics. At the same time, it balances the severe travel penalty observed under Skill-First assignment. Statistical analysis using Friedman and Wilcoxon signed-rank tests shows that there are significant differences in operation across the different algorithms for the evaluation metrics. These findings show that PSP is best interpreted as a compromise strategy that balances service quality and operational efficiency rather than maximizing any single metric in isolation.

The face recognition systems (FRS) can have latency bias, with some of their inputs or subsets having disproportionate inference delays, which result in unfair quality-of-service despite apparently similar predictive performance. Although much research on fairness in face recognition research has been done on accuracy-based disparities, there has been little work on runtime behavior. This study seeks to understand whether there are differences in performance in real-time face recognition systems based on visually complex sub-groups, and whether these differences can be considered as operational fairness concerns. In this work, we investigate the latency problem of real-time FRS, with especially great attention to tail inference as the first-class fairness measure. We utilize a label-free auditing framework for conducting runtime fairness assessment, and propose the tail inference latency metric for evaluating the fairness level of the deployment. Instead of focusing on prediction scores as in prior work, our work focuses on the fairness level of our deployment. The MeGlass dataset (noted as a publicly available Kaggle release) is used as the base of experiments, which is an image-only face dataset to test the strength against eyeglasses, and the MeGlass dataset is a base to assess the reality of deployment through the Kaggle publicly available release. Using eyeglass wearers as a sample visually complex group, we conduct a label-free Responsible AI audit in the conditions when we do not have labels of the sensitive attributes. We observed that average performance difference in mean is not statistically significant (p = 0.36) by the permutation test, while mean difference in p99 is 16.85 ms between cohorts, bringing up fairness differences not captured by averages. Doing systems-level regression, we attribute such variances to pipeline-level attributes (e.g. detection confidence, input complexity) as opposed to the use of these attributes directly as causal factors. We also evaluate mitigation policies and determine trade-offs among the latency, fairness and recognition fidelity. The findings here show that fairness inequities can arise at worst-case runtime behavior, not in average latency, suggesting that the use of metrics focused on the tail of a run-time fairness distribution should be considered in fairness assessments of run-time deployed face recognition systems.

Gifted learners require instructional environments that transcend conventional pedagogical boundaries, offering adaptive challenge, metacognitive scaffolding, and high degrees of learner autonomy. This study presents the design, implementation, and empirical evaluation of the AI-Enhanced Self-Directed Learning (AESDL) framework an integrated adaptive system grounded in Treffinger's Independent Learner Model (ILM) and operationalized through an ensemble of five artificial intelligence (AI) technologies: the Google Gemini API (as central orchestration controller), cognitive computing, multi-agent systems (MAS), expert systems, and automatic speech recognition. The model is built upon four core ILM components Guidance, Self-Development, Enrichment, and Seminars/In-Depth Study each computationally instantiated through dedicated AI subsystems. A quasi-experimental pre-test/post-test control group design was employed with 50 gifted secondary school learners (25 experimental, 25 control) drawn from model gifted-education classrooms in Khartoum State, Sudan. The experimental group received eight weeks of instruction via the AESDL system; the control group received equivalent instruction through conventional electronic resources. Outcome measures self-directed learning skills, problem-solving capacity, and academic enrichment achievement (each scored on a 50-point scale) were analyzed using independent-samples t-tests with Cohen's d effect sizes. Results indicated statistically significant and practically large superiority of the AESDL condition: self-directed learning (t(48) = 7.91, p < .001, d = 2.24), problem-solving (t(48) = 5.52, p < .001, d = 1.56), and academic achievement (t(48) = 6.50, p < .001, d = 1.84). These findings advance the empirical evidence base for AI-mediated gifted education and provide actionable design principles for intelligent adaptive learning systems. The control condition comprised conventional electronic learning resources (digital textbooks, instructional videos, and static online exercises). Ninety-five-percent confidence intervals for the between-group mean differences were [9.13, 15.35], [5.62, 12.06], and [7.02, 13.30] for self-directed learning, problem-solving, and academic achievement, respectively. Given the modest sample (n = 50) and the near-ceiling experimental-group scores, these findings should be interpreted with caution regarding potential ceiling effects and limited external validity, and warrant independent replication with larger, more diverse samples.

Quantum Computing is gradually gaining importance as a transformative technology with the potential to revolutionize business applications by solving complicated problems beyond the capabilities of classical computers. Using basic principles such as superposition and entanglement, quantum systems can concurrently handle a wide range of combinations of possibilities. It offers several benefits in optimization, financial modeling, supply chain management, risk analysis, and machine learning. Despite the limitations of existing hardware in the NISQ era, rapid developments are enhancing scalability and reliability. This study analyses the practical implications, opportunities, and challenges of integrating quantum computing into business environments, spotlighting its capacity to drive innovation, competitive advantage, and long-term strategic growth throughout industries. By situating quantum computing within a transdisciplinary engineering and business ecosystem, this review demonstrates that quantum technologies are not merely enablers of computational speed but catalysts for systemic innovation.

There exists an ongoing problem among technical and vocational education and training institutions in Kenya regarding validation of demonstrated practical competencies, since the verification method depends on periodic observation which is paper-based and hence prone to subjectivity and potential corruption. In this article, we propose a quantum-optimized decentralized ledger system in which trainees wear wearable Internet of Things sensors and a consensus algorithm referred to as Quantum-Optimized Decentralized Proof-of-Skill is used to record and validate skill demonstration. The sensors detect motion, physiological, and environmental signals that are converted into skill proficiency ratings using an edge-based feature-processing pipeline; validator nodes are then chosen by solving a quadratic unconstrained binary optimization problem with a quantum-inspired algorithm. Assessment records are written to a permissioned ledger accessible only to training institutions, national qualification authorities, and prospective employers. According to a discrete-event simulation of the proposed consensus algorithm relative to proof-of-work, proof-of-stake, and practical Byzantine fault tolerance, the proposed algorithm has a reduced mean confirmation time of 148 ms and an increased sustained throughput of 386 tx/s under the modeled conditions of the Kenyan network environment, alongside skill-scoring consensus (F1 overall = 0.88) with the assessor ground truth obtained through simulations based on the accuracy of the wearable-assessments in published reports.

Medical image analysis plays an important role in early breast cancer detection through mammographic image analysis. But, some information in the GLCM based Haralick texture features is redundant and irrelevant, which can impact performance in classification. In this study, a feature selection framework based on Cuckoo Search Optimization (CSO) is presented to select the optimum Haralick features for the breast cancer classification. The proposed approach consists of the preprocessing of mammograms, extraction of features from the GLCM, optimization of the GLCM features using the CSO technique, and classification using machine learning algorithms. Fourteen extracted GLCM attributes were used in the experiments performed on MIAS. The CSO-KNN model obtained the best classification rate of 90.00% with 6 selected features, which means that 8 attributes have been removed. The proposed framework significantly reduces feature dimensionality and improves the classification efficiency in computer-aided breast cancer diagnosis.