WS-STACK: A Weighted Stacking Ensemble with Multi-Criteria Feature Selection for Multi-Class Traffic Classification and Anomaly Detection in Heterogeneous Wireless Sensor Networks

Zainab Hussein; Nureize bt

doi:https://doi.org/10.54216/IJWAC.100206

Journal of Cognitive Human-Computer Interaction | 2026

Adaptive Interface Personalization through Real-Time Cognitive Load Detection

High-stakes computer work often requires users to interpret dense visual information while responding to timesensitive events. Static interfaces can become counterproductive in such conditions because the amount of information presented to the user does not change when mental demand rises. This paper presents an adaptive interface personalization approach that detects cognitive load from pupillometry, heart-rate variability, gaze behaviour, and interaction traces, then selects a transparent interface response. The proposed approach does not simply reduce screen content; it chooses between full, highlighted, simplified, and critical-only modes while preserving user control and explanation cues. A feature-level experimental analysis was conducted using a multimodal workload table structured around public cognitive-load datasets and high-stakes monitoring tasks. The results show that pupil expansion, lower HRV, response delay, gaze dispersion, and screen density jointly indicate rising cognitive load. The adaptation policy reduced predicted interaction errors and shortened response latency in high-load windows while maintaining explanation support for user trust. The findings suggest that cognitive-load detection should be treated as a personalization service rather than a hidden automation layer.

groups

Wadhah Abdullah mail -

Aygul Z. Ibatova mail

link https://doi.org/10.54216/JCHCI.110204

Volume & Issue

Vol. Volume 11 / Iss. Issue 2

Details open_in_new

Journal of Cognitive Human-Computer Interaction | 2026

Explaining AI Decisions to Mitigate Cognitive Biases in Human-AI Collaboration

Human-AI collaboration can improve decision quality only when users know when to rely on an AI recommendation and when to resist it. Explanations are often proposed as a remedy, but explanation content can also intensify automation bias or reinforce a user’s initial belief. This paper presents a cognitive explanation selection model for mitigating over-reliance and under-reliance in AI-assisted decision tasks. The study compares no explanation, feature-based, contrastive, example-driven, and hybrid explanations across simulated novice, intermediate, and expert decision makers using a public medical decision dataset as the task substrate. The analysis focuses on reliance behaviour rather than on model accuracy alone. The proposed model estimates when the user is likely to accept a wrong recommendation, reject a correct recommendation, or accept advice simply because it confirms an initial judgment. The results indicate that contrastive and hybrid explanations are more effective for reducing automation bias, while example-driven explanations preserve trust for lower-expertise users. The paper concludes with a transparent interface loop for high-stakes environments in which explanation style is selected according to user expertise, AI confidence, and human-AI agreement.

groups

Aiswan Aumanti mail -

Citra Dewi mail

link https://doi.org/10.54216/JCHCI.110205

Volume & Issue

Vol. Volume 11 / Iss. Issue 2

Details open_in_new

Journal of Cognitive Human-Computer Interaction | 2026

When the Story Knows You: Personalisation, Interactivity, and Emotional Transportation in Human-AI Collaborative Narrative Experiences

Stories have always been the primary medium through which human beings share emotions, build empathy, and make sense of experience. The emergence of large language models capable of generating coherent, contextually rich narratives raises a fundamental question for human-computer interaction: when a story is generated by a machine, does it still carry the emotional weight and imaginative pull of one written by a human, and can the design of the interaction itself amplify or diminish that pull? This paper reports a controlled within-subjects experiment in which thirty-six participants read or actively co-shaped stories produced by a large language model under four conditions that crossed two levels of interactivity—passive reading versus branching-choice interaction—with two levels of personalisation—generic narrative versus one adapted to the participant’s stated interests and preferences. Emotional engagement was measured through narrative transportation, positive and negative affect, sense of narrative agency, trust in the AI narrator, and perceived story quality. The study finds that both interactivity and personalization independently increase emotional transportation, and that their combined presence produces an amplified effect that is larger than either factor alone, while trust in the AI narrator emerges as a partial mediator of the personalization advantage. Individual differences in baseline narrative engagement propensity predict the magnitude of benefit from the most engaging condition, providing actionable guidance for adaptive storytelling interface design.

groups

Nurdaulet Karabayev mail -

Sholpan Baumuratova mail

link https://doi.org/10.54216/JCHCI.110206

Volume & Issue

Vol. Volume 11 / Iss. Issue 2

Details open_in_new

Journal of Cognitive Human-Computer Interaction | 2026

Scene-Level Assessment of Comfort, Legibility, and Spatial Control in Virtual Reality Interfaces

Virtual reality interface quality is not determined by visual appeal alone. A scene may look convincing while still producing unstable gaze, uncomfortable depth switching, excessive head movement, or slow target selection. This paper presents a scene-level assessment framework for measuring comfort, legibility, and spatial control in VR interfaces. The work is deliberately organized as a design-science evaluation rather than as a conventional classifier study: it begins with interface failure mechanisms, defines observable headset and scene variables, computes a Virtual Reality Interface Comfort score, and then translates the results into review actions. The empirical analysis uses a processed feature-level extract aligned with public VR eye-tracking task structures and combines gaze stability, pupil variability, vergence error, head-turn demand, tracking loss, selection latency, contrast balance, target comfort, depth pressure, and spatial-memory support. The results indicate that comfortable VR scenes are characterized by stable fixation, consistent depth placement, strong spatial memory support, and modest interaction latency, while high-risk scenes are mainly associated with head-turn demand, tracking loss, pupil variability, and depth pressure. The paper contributes a transparent measurement model, a set of scene pattern diagnostics, and a practical governance workflow for deciding when a VR interface should be released, revised, or retested.

groups

Massila Kamalrudin mail -

Mustafa Musa mail

link https://doi.org/10.54216/JCHCI.110207

Volume & Issue

Vol. Volume 11 / Iss. Issue 2

Details open_in_new

Journal of Cognitive Human-Computer Interaction | 2026

Designing Algorithmic Accountability for Citizens: Developing and Validating a Three-Layer Transparency Framework for Public Sector Decision Systems Through Iterative Participatory Prototype Design

When governments use algorithmic systems to determine eligibility for housing support, welfare benefits, or social services, the citizens whose lives are most directly affected are often the least equipped to understand, scrutinise, or challenge the outcomes. Standard decision notices provide statutory reference numbers and outcome statements without any meaningful account of which data was used, why the algorithm produced the result it did, or what a citizen can realistically do next. This accountability gap is not merely a design inconvenience; it erodes the procedural fairness that democratic governance requires, and it disproportionately affects the most vulnerable service users. This paper reports a three-phase research programme in which a principled transparency framework for citizen-facing algorithmic decision interfaces was developed and validated through sustained engagement with end users. A needs assessment with 142 citizens and 18 civil servant interviews established what transparency citizens actually require. Three iterative co-design workshops with 24 citizens and 8 frontline officials produced progressively refined interface prototypes organised around three distinct transparency layers—process disclosure, rationale explanation, and contestation support. A subsequent think-aloud evaluation with 36 citizens compared four interface conditions ranging from the current opaque standard to the full three-layer framework. The fully layered interface substantially outperformed the existing standard and all partial implementations across trust, perceived actionability, comprehension, and transparency satisfaction. The paper contributes the framework itself as a theoretically grounded and empirically validated design resource, a set of evidence-based design guidelines derived from across all three study phases, and a replicable participatory methodology for involving affected citizens in the design of AI governance interfaces.

groups

Arash Salehpour mail -

Laula Zhumabayeva mail

link https://doi.org/10.54216/JCHCI.110208

Volume & Issue

Vol. Volume 11 / Iss. Issue 2

Details open_in_new

International Journal of Wireless and Ad Hoc Communication | 2026

Fog-Assisted Trust and Anomaly-Aware Scheduling for Wireless Sensor IoT Devices

Wireless sensor Internet of Things (IoT) devices increasingly generate time-sensitive traffic that cannot be efficiently inspected only in a remote cloud. Fog computing reduces the distance between sensing devices and decision logic, but fog nodes must jointly manage latency, queue pressure, wireless channel variability, energy use and security risk. This paper presents FogSense-TSA, a trust-aware and anomaly-aware scheduling model for wireless sensor IoT traffic in fog computing environments. The model integrates traffic intensity, wireless link behaviour, fog-resource state and temporal trust into a local decision process that determines whether a device-window should be accepted, quarantined at the fog layer or escalated to cloud inspection. The empirical analysis is conducted using a reduced analysis-ready file aligned with a recent public IoT device-identification and anomaly-detection setting. The proposed formulation introduces three algorithmic components: online trust-risk scheduling, load-aware fog placement and adaptive threshold calibration. Mathematical analysis is provided for evidence aggregation, trust stability, latency decomposition, energy cost, constrained placement and computational complexity. The results show that fog placement substantially reduces service latency relative to cloud-only routing while preserving high anomaly-discrimination capability. The strongest predictors are trust score, flow intensity, jitter, fog CPU load, payload entropy and queue pressure, indicating that fog-layer security should be coupled with wireless access and resource conditions rather than treated as a separate classifier. The study provides a reproducible and interpretable basis for designing lightweight security and scheduling modules for wireless sensor IoT deployments.

groups

Arash Salehpour mail -

Tamara Zhukabayeva mail

link https://doi.org/10.54216/IJWAC.100202

Volume & Issue

Vol. Volume 10 / Iss. Issue 2

Details open_in_new

International Journal of Wireless and Ad Hoc Communication | 2026

Tiny Intelligence in Fog-Assisted Wireless Sensor IoT Networks: A Review of Deployment Patterns, Resource Trade-offs, and Open Challenges

Wireless sensor IoT networks are moving from simple measurement pipelines toward distributed systems where sensing, interpretation, filtering, and coordination are divided across devices, fog nodes, and cloud services. This review examines that transition through the lens of tiny intelligence, with special attention to how small models, local event filters, federated updates, service placement, and privacy controls reshape fog-assisted wireless sensor deployments. The paper does not treat fog computing as a generic latency layer. Instead, it studies fog as a governance and orchestration layer that decides which data should stay at the device, which events should be aggregated nearby, and which models require cloud-level supervision. A structured comparison of prior studies is provided across architecture, TinyML, federated learning, placement, security, benchmarking, and lifecycle coverage. The synthesis shows that the literature has matured in modelling fog resources and building lightweight inference functions, but remains fragmented in lifecycle management, cross-layer wireless awareness, privacy accounting, and reproducible evaluation. The review concludes with a research agenda for sensor-to-fog intelligence pipelines that are adaptive, auditable, energy-aware, and suitable for long-lived cyber-physical deployments.

groups

Aygul Z. Ibatova mail -

Baumuratova Dilaram mail

link https://doi.org/10.54216/IJWAC.100203

Volume & Issue

Vol. Volume 10 / Iss. Issue 2

Details open_in_new

International Journal of Wireless and Ad Hoc Communication | 2026

A Deep Reinforcement Learning Framework with Solar Energy Forecasting for Adaptive Routing and Lifetime Extension in Energy-Harvesting Wireless Sensor Networks

Battery-powered sensor nodes expire when their energy reserves are depleted, terminating data collection regardless of the physical integrity of the hardware. Solar harvesting offers a viable path to perpetual operation, but only when the routing layer can continuously track the time-varying energy state of every node and steer traffic away from nodes likely to be power-starved in the near future. Classical clustering and chain-based protocols select forwarding paths without regard to harvested energy, leading to premature node death even when sufficient solar income would have been available to sustain operation. This paper presents a deep reinforcement learning framework in which each sensor node operates an independent Deep Q-Network agent that adapts its next-hop forwarding decision based on local battery state, short-horizon solar energy forecasts, link quality estimates, and the residual energy levels of candidate neighbours. A lightweight LSTM sub-model provides the solar prediction horizon that the agent uses as part of its state representation, enabling it to distinguish nodes that are temporarily depleted but will recover from those whose batteries are trending toward permanent failure. Extensive simulation across a 100-node deployment over 3,000 operational rounds confirms that the proposed approach substantially extends network lifetime, improves packet delivery, and reduces wasted harvested energy compared with five competitive baselines. Reward function ablation, scalability experiments, and an energy neutrality verification further validate the design choices and confirm stability across a wide range of deployment conditions.

groups

Suhasini Monga mail -

Damandeep Kaur mail

link https://doi.org/10.54216/IJWAC.100204

Volume & Issue

Vol. Volume 10 / Iss. Issue 2

Details open_in_new

International Journal of Wireless and Ad Hoc Communication | 2026

Frequency-Aware Antenna Configuration for Reliable Wi-Fi Communication Networks

Dense Wi-Fi deployments are often tuned by changing channel width or adding access points, while the joint effect of antenna gain, operating frequency, wall loss, and network interference receives less systematic attention. This paper presents a frequency-aware antenna configuration model for Wi-Fi communication networks operating in the 2.4, 5, and 6 GHz bands. The model combines an indoor link budget, antenna-pattern classes, bandwidth dependent noise, an airtime-overlap penalty, and a coverage-assurance score that balances signal quality, throughput, latency, and packet error. A reproducible design-space table is generated from a validated Wi-Fi engineering model and analyzed across five deployment scenarios, three antenna families, four channel widths, and multiple client distances. The results show that higher frequency bands improve short-range capacity but deteriorate faster under distance and wall loss, while directional antenna gain can recover a substantial part of the lost link margin. The paper provides planning rules for selecting antenna type, frequency band, and channel width according to coverage, capacity, and interference risk. The work is intended for Wi-Fi network designers who need interpretable engineering evidence rather than a black-box prediction model.

groups

Mohammed. I. Alghamdi mail -

Abdul Rahaman Wahab Sait mail

link https://doi.org/10.54216/IJWAC.100205

Volume & Issue

Vol. Volume 10 / Iss. Issue 2

Details open_in_new

International Journal of Wireless and Ad Hoc Communication | 2026

WS-STACK: A Weighted Stacking Ensemble with Multi-Criteria Feature Selection for Multi-Class Traffic Classification and Anomaly Detection in Heterogeneous Wireless Sensor Networks

Heterogeneous Internet-of-Things deployments expose wireless sensor networks to a diverse and continuously evolving threat landscape encompassing distributed denial-of-service flooding, network reconnaissance scanning, and brute-force credential attacks. Existing intrusion detection approaches predominantly adopt single-classifier architectures and binary labelling, which are ill-suited to the multi-class, class-imbalanced traffic characteristic of real-world IoT sensor deployments. This paper proposes WS-STACK, a Weighted Stacking ensemble that combines five heterogeneous base learners—Random Forest, XGBoost, Support Vector Machine, K-Nearest Neighbours, and Gradient Boosting—under an ℓ2-regularised Logistic Regression meta-learner trained on cross validationgenerated probability features. A three-stage feature engineering pipeline comprising mutual information filtering, variance inflation factor pruning, and correlation-based elimination reduces the 83 dimensional RT-IoT2022 feature space to 20 informative features, and the Synthetic Minority Over-Sampling Technique corrects the six-fold class imbalance prior to training. Evaluated on 83,000 labelled network flow records from the publicly available RTIoT2022 benchmark spanning four benign traffic patterns and seven attack categories, WS-STACK achieves 99.61% classification accuracy, a weighted F1-score of 0.9960, and an AUC-ROC of 0.9978, outperforming every individual base classifier and five recently published state-of-the-art baselines. The false positive rate is reduced to 0.0006, and ten-fold cross-validation confirms μacc = 0.9959 (σ = 0.0004). Ablation experiments identify SMOTE as the single most critical preprocessing component, and noise robustness tests confirm 98.81% accuracy under 20% Gaussian feature perturbation. The framework is grounded through a formal variance-reduction proof and a channel-energy anomaly model that establishes the physical motivation for packet-rate features as the dominant intrusion detection signal in constrained wireless sensor networks.

groups

Zainab Hussein Arif mail -

Nureize bt Arbaiy mail

link https://doi.org/10.54216/IJWAC.100206

Volume & Issue

Vol. Volume 10 / Iss. Issue 2

Details open_in_new