The rapid proliferation of wearable health technologies for older adults has not been matched by sustained adoption. Whilst existing literature predominantly focuses on theoretical acceptance models, it provides limited empirical guidance for practitioners seeking to improve usability. This systematic review analyses the usability and acceptance of wearable health devices amongst older adults aged 65 and older to identify usability challenges. Adhering to PRISMA protocols, we systematically searched Web of Science, CINAHL Complete, PubMed Central, Scopus, JMIR, IEEE Xplore, MDPI, SpringerLink, Taylor & Francis, and the ACM Digital Library for studies published between 2014 and 2025 evaluating the usability and/or acceptance of consumer wearable health technologies. Data extraction focused on usability challenges, acceptance frameworks, device ergonomics, and ecosystem integration. Of the 446 records identified, 11 studies involving 727 participants met the inclusion criteria. Key findings revealed predominant reliance on psychological models over engineering-focused usability research. Significant gaps emerged in physical design studies despite their influence on long-term use. Unstandardised training approaches limited scalability, and fragmented methodologies impeded evidence synthesis. Future research requires standardised evaluation frameworks, materials-focused usability studies, and longitudinal analyses. We propose an interdisciplinary research agenda bridging behavioural science and engineering design to enhance evidence-based development for geriatrics.
The emergence of Digital Human CEOs marks a significant development in the relationship between artificial intelligence, organizational authority, and corporate governance. This article develops a conceptual framework to explain how Digital Human CEOs may acquire, or fail to acquire, legitimacy as organizational leaders. Drawing on anthropomorphism theory, social presence theory, and transportation theory, the paper proposes an integrated mechanism through which human likeness, perceived presence, and narrative engagement shape stakeholder trust and acceptance. The analysis argues that legitimacy in AI-driven leadership does not depend on technical sophistication alone, but on the interaction between design realism, communicative effectiveness, transparency, and governance arrangements. Particular attention is given to the conditions under which trust may weaken, including excessive realism, limited explainability, cultural variation in authority expectations, and unresolved questions of accountability. The article also examines the governance implications of assigning executive authority to digitally embodied agents, emphasizing the need for clear oversight structures, auditable decision processes, and identifiable human responsibility. By clarifying the psychological and organizational pathways through which Digital Human CEOs may be evaluated, this paper contributes a structured basis for future empirical research while offering a governance-oriented perspective on the responsible development of AI-driven leadership.
Learning analytics dashboards (LADs) have been widely adopted in higher education to support academic engagement and provide pedagogical interventions. Despite persistent calls for theory-driven design to ensure that LADs provide meaningful support to students, a significant gap remains in the existing literature regarding how theories and frameworks are operationalised within LAD features to promote actionable insights and pedagogical interventions. While existing studies generally acknowledge the importance of theory, they nevertheless remain at the theory level, merely recommending theoretical perspectives without offering concrete guidelines for their translation into dashboard design elements. The study was conducted with the aim of systematically examining how theories and frameworks have been operationalised in the design of LADs to support tertiary students. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) approach was used to retrieve 35 peer-reviewed papers from six academic databases. These were then analysed. In contrast to earlier studies, the study revealed learning theories to be used to a large degree in LAD design, complemented by motivational, perception and user experience theories as well as LAD-specific design frameworks. A comprehensive analysis of the operationalisation of these theories and frameworks highlights key gaps and points to valuable implications for the design of future LADs.
Application icons are an essential part of almost all visual user interfaces. As such, the visual design of an application icon can have a major impact on the usability performance, and therefore, the consumer satisfaction and success of the corresponding application. However, there has been very little empirical research on the ways in which various visual properties of icons affect their performance during different phases of consumers’ interactions with them. In this paper, we present an online survey study conducted with design professionals to evaluate their understanding of the importance of different visual properties of mobile application icons—complexity, concreteness, and familiarity—during the customer journey, from the first exposure to assurance and continued the use of applications. Based on our findings, we evaluated the relevance of a set of existing guidelines for the visual design of application icons with the aim of improving their targeted customer journey. Designers generally agreed with most of the proposed guidelines. The highest-rated guidelines included the importance of strong contrast for rapid recognition during continued use, and the role of familiarity in supporting visual processing fluency during assurance and visual search during continued use.
Humans and animals can learn new things, improve their skills, and share what they know with others for the rest of their lives. Neurocognitive systems that help with skill growth, memory consolidation, and adaptation are what make this lifelong learning process possible. For models that need to process and adapt to information that is always changing, lifelong learning is very important in artificial intelligence. But AI systems have trouble with “catastrophic forgetting,” which is when new information replaces old information. This makes adaptive learning very hard. Continuous learning is especially helpful for personal AI assistants since it lets them improve their understanding of user preferences, learn new tasks, and remember past encounters. This makes it easier for them to give personalized, context-aware answers, which makes the user experience smooth and easy to understand. This paper suggests a new model of continual learning called the elastic prototype-aligned contrastive (E-PAC). The network’s settings were changed while it was learning so that it could better remember how to forecast existing categories while simultaneously learning new ones. The experiment has been set up, and the recommended method is tested against the existing continual learning technique, elastic weight consolidation (EWC) on the “16 personality type” dataset that is available on Kaggle. The experimental results demonstrate that the proposed E-PAC model achieves a test accuracy of 98.9%, outperforming the baseline EWC method, which achieves 97.2% accuracy. This makes the classification and prediction model much more scalable and intelligent.
E-assessment of STEM subjects is limited given the poor user experience when producing diagrams and equations on the computer with traditional mouse and keyboard interfaces. However, stylus input could help overcome the limitations of STEM e-assessment. The aim of this work is to compare stylus input to paper for STEM e-assessments, focussing particularly on usability and cognitive load. Our prediction is that styli can provide a similar user experience to that of paper. A longitudinal study with students assessed the usability and cognitive load of direct and indirect stylus input devices compared to paper in an e-assessment context. The results show that overall stylus input was not significantly different to paper in terms of cognitive load and usability. For self-reported cognitive load, the direct and indirect styli were equivalent to paper. The direct stylus was also equivalent to paper in terms of ease of use, intuitiveness and confidence. Most students also preferred the devices over paper for mathematics assessments. Over time, cognitive load decreased, and usability and preferences increased. These results indicate a strong case for the use of stylus input for STEM e-assessment. This is an important finding as assessment is digitised, particularly in terms of providing adequate support for students completing STEM e-assessments.
Web accessibility ensures that all individuals, including people with disabilities, can interact effectively with digital portals. This study evaluates and classifies the levels of web accessibility of the official portal of the Decentralized Autonomous Government (GAD) of the canton of Portoviejo, Ecuador, using an unsupervised fuzzy clustering approach. Automated metrics aligned with the Web Content Accessibility Guidelines (WCAG 2.1, Level AA) were collected using the web accessibility test (TAW) tool. Subsequently, exploratory factor analysis (EFA) was applied to reduce the dimensionality of 21 initial criteria, retaining seven latent factors representative of web accessibility. Based on these factors, the fuzzy C-means (FCM) algorithm was implemented to group the evaluated pages into accessibility levels. The results reveal the existence of clusters with overlapping boundaries, confirming that web accessibility constitutes a continuum rather than a binary condition. Finally, the study provides accessibility profiles and practical recommendations by the cluster, contributing to the progressive improvement of digital inclusion in the context of local public administration.
The rapid proliferation of generative artificial intelligence and automated systems has radically transformed human interaction, introducing emerging forms of functional affective expression that profoundly challenge traditional biological definitions. These advancements rely on what the existing literature terms 'artificial empathy', a computational capacity for affect detection and proxy generation known for nearly 2 decades. However, the social sciences urgently require a broader structural categorization to evaluate its systemic implementation. Therefore, this study conceptually extends and systematizes this existing literature by proposing the advanced sociological construct of 'Teleological Empathy'. The research is framed within a qualitative, theoretical-constructive methodology, based on a critical documentary review of interdisciplinary sources from psychology, sociology, artificial intelligence, alignment ethics and the philosophy of language. Through abductive reasoning and robust theoretical triangulation, the study identifies categorical gaps in the current academic treatment of nongenuine, functional empathy. The main result is the formal definition of Teleological Empathy as a form of goal-oriented and purposefully designed empathic expression that successfully fulfils specific, legitimate communicative functions in human or technological contexts, completely without requiring a subjective, biological affective experience on the part of the agent. To operationalize this construct, five key defining characteristics are rigorously distinguished: strategic intentionality, discursive or algorithmic design, absence of internal emotional experience, contextual coherence and functional legitimacy. Moving beyond the outdated ontological debate of 'real versus fake'& raquo; empathy, this proposal provides specific operationalization guidelines for future empirical validation, ultimately offering a robust analytical framework specifically designed to evaluate interactional ethics, value alignment and organizational standardization in an increasingly automated world.
B-Tree is a well-known indexing data structure for managing large persistent data. In its native form, it is used by database management systems and not humans. It is seen as a tree whose nodes are disk blocks. In this paper, an adjustment on B-Tree is proposed to become easier to use for humans. This is done by relaxing the restriction that its node has to be a disk block. Instead, tree nodes are implemented as files. All B-Tree algorithms were adjusted and implemented to adapt to this change. The main gain of this work is better usability of a B-Tree for humans. Another gain is the complete freedom in setting the degree of the tree and the size of the data record. An experimental study was conducted to study how the performance of insertion and deletion changes with the degree of the tree t. It was found that the performance of the algorithms is very much affected by the number of levels of the tree. Very large values of t caused a degradation in performance. However, this degradation was found to be low compared with the change in t. This work opens the door for more research to enhance the usability of the B-Tree: like supporting the adjusted B-Tree with a GUI that makes it even easier to build and manipulate B-Trees by a human.
This empirical study analyzes 26,399 education-focused user reviews of ChatGPT from the Google Play Store to investigate factors shaping perceptions of educational usability. Using quantitative methods including feature engineering, sentiment analysis, and ordinary least-squares regression, we examine how emotional sentiment, content quality, and review scores influence usability perceptions. Results reveal a counterintuitive emotional pattern: while surprise and anticipation enhance perceived usability, joy and disgust are associated with lower usability perceptions. Overall sentiment emerges as the strongest predictor of usability (coefficient = 0.8565, p <= 0.001), with content quality also exerting a significant positive effect (coefficient = 0.0334, p <= 0.001). In contrast, review scores show a small but significant negative relationship with usability (coefficient = -0.0210, p <= 0.001). Trust exhibits a slight negative effect, whereas fear, anger, and sadness show no significant associations. These findings highlight the critical role of emotional factors in educational AI usability, suggesting that developers should prioritize both high-quality content and positive emotional engagement. This research contributes to HCI and educational technology by providing a replicable framework for analyzing user perceptions of AI tools while emphasizing the importance of affective dimensions in technology adoption.
This study aimed to optimize emotion classification using electroencephalography (EEG) signals by evaluating the impact of window size, overlap, classification models, and wavelet transform selection. Understanding these factors is crucial for enhancing EEG-based emotion recognition systems, which play a key role in improving human-computer interaction, adaptive user interfaces, and neurophysiological therapy. A comparative analysis was performed using supervised classification models with varying levels of window overlap and wavelet transforms to extract characteristics. The study followed four stages: (1) data preprocessing and wavelet selection for signal decomposition, (2) determination of optimal overlap and feature extraction, (3) application of supervised classification models, and (4) comparative evaluation using metrics such as the Area Under the Curve (AUC), accuracy (ACC), kappa coefficient, and F1-score. The findings indicate that Random Forest (RF) and Support Vector Machines (SVM) achieve superior performance with overlap levels between 10% and 30%, regardless of the wavelet transform applied. In contrast, Logistic Regression (LR) and Decision Tree (DT) models exhibited lower accuracy and did not show significant improvements with varying overlap levels. Excessive overlap, however, degrades performance, emphasizing the importance of selecting an appropriate overlap level. Additionally, while wavelet transforms were used for feature extraction, the specific wavelet type (Coiflets, Daubechies, Symlet, Haar, or Meyer) did not significantly influence classification accuracy. These results highlight the importance of selecting appropriate overlap levels to optimize classification accuracy. This study provides valuable information for improving EEG-based emotion recognition systems, which can be applied in fields such as brain-computer interfaces, mental health monitoring, and human-computer interaction. Future research should explore advanced feature extraction techniques and deep learning approaches to further improve classification performance.
In the last few years, recent advances in alternative methods for neurorehabilitation have witnessed outstanding progress including electrical stimulation (ES) and its derivative functional electrical stimulation (FES), a treatment that applies a transcutaneous electrical current to induce muscle contractions and is commonly used in individuals with motor disabilities, such as poststroke, amyotrophic lateral sclerosis (ALS), multiple sclerosis (MS), and spinal cord injury (SCI). Similarly, brain–computer interfaces (BCIs) have been shown to be a powerful tool in rehabilitation processes specifically for people in motor disabilities situations or injuries associated with the brain. By doing a bibliometric analysis, this work presents some of the most important advances in the integration of these two approaches, BCI and FES, for motor recovery showing at the same time their main aspects, the most used methods and future challenges. The bibliometric analysis allows researchers to identify possible ways to explore current developments, challenges and future perspectives of a specific field of study. Scopus tool and open software VOSviewer were used. As a result, it is concluded that FES is a technique applied in different scenarios from sports and fitness, robotics, and physical therapy, among others, and that the integration of this technology with a BCI-based control could improve neurorehabilitation processes.
Autism spectrum disorder (ASD) encompasses a range of neurodevelopmental conditions characterized by difficulties in social interaction, repetitive behaviors, speech, and nonverbal communication. Affecting approximately 1 in 36 children globally, ASD poses significant diagnostic challenges, particularly in early childhood, due to the absence of definitive medical tests. Current diagnostic practices rely heavily on behavioral observations and developmental history, often leading to delays in intervention. Early and accurate detection is essential for initiating timely support and improving developmental outcomes. In this study, we present a novel deep learning framework that integrates a one-dimensional convolutional neural network (1D-CNN) with the synthetic minority oversampling technique (SMOTE) to detect ASD using only 10 behavioral screening features (A1-A10). The dataset, derived from publicly available screening questionnaires, exhibited significant class imbalance, which was mitigated using SMOTE to improve model sensitivity and fairness. The proposed CNN + SMOTE model demonstrated superior performance, achieving an accuracy of 99.31%, surpassing traditional classifiers such as random forest (98.0%) and XGBoost (98.5%). Model evaluation was conducted using standard metrics including accuracy, precision, recall, F1-score, and AUC, with validation via k-fold cross-validation. The results underscore the effectiveness of lightweight deep learning models augmented by oversampling techniques in developing accurate, interpretable, and scalable tools for ASD screening particularly in resource-limited settings where access to advanced diagnostic technologies may be constrained.
This study explores the integration of augmented reality (AR) in energy trading platforms, focussing on user experience within electromobility charging networks. It begins by introducing AR concepts and their applications across industries. Integrating AR into energy trading platforms for electromobility charging networks presents innovative potential; however, user experiences and adoption rates remain underexplored. This review seeks to address this gap by analysing the existing AR-aided platforms, focussing on user engagement, satisfaction and the overall effectiveness of these systems in enhancing the charging experience. The current state of electromobility charging networks is analysed, highlighting the existing energy trading paradigms and user experience challenges. Innovations in peer-to-peer energy sharing and trading, particularly blockchain technology, are examined along with the potential advantages of integrating AR. This review also discusses AR technologies and tools applicable to electromobility, user-centric design principles and the impact of AR visualisation on energy trading decision-making. This article reviews and evaluates security and privacy implications, case studies, challenges, future directions, regulatory and ethical considerations and user adoption factors, culminating in actionable recommendations for industry stakeholders.
Effective electronic record management systems (ERMSs) are crucial for modern organizations, offering benefits such as streamlined document management, enhanced security, and improved institutional memory. However, poor usability often hinders ERMS adoption and effectiveness. While various usability evaluation methods exist, a comprehensive approach integrating multiple techniques is often lacking, particularly in the context of ERMS, where factors like data security and regulatory compliance are paramount. This paper presents a novel hybrid usability assessment model that combines heuristic walkthrough, statistical log analysis, server log path analysis, and user testing to provide a holistic evaluation of ERMS usability. This integrated approach, suitable for continuous evaluation throughout the software lifecycle, addresses limitations inherent in individual methods, capturing both expert insights and real-world user behavior at scale, and generating complementary insights for diverse stakeholders (executives, developers, procurement, UX researchers). Furthermore, this study offers specific heuristics for evaluating ERMS, such as "standardized terminology" and "automatic suggestions for the standard file plan." Applied to Hacettepe University's ERMS, the model revealed usability challenges such as poor search functionality, inefficient workflows, and nonintuitive design. The study demonstrates how the combined insights from these methods provide a more nuanced understanding of ERMS usability than single-method approaches, generating actionable and cost-effective recommendations for system improvement. This research contributes a practical framework for enhancing ERMS usability and highlights the importance of multimethod evaluations for complex web applications.
In this research, we hypothesize that attitudes toward artificial intelligence (AI) are shaped by individuals’ perceived competence in using and managing it, as well as their assessment of the importance of AI’s understandability and transparency, often facilitated by explainable artificial intelligence (XAI) interfaces. Similarly, competence in AI implies the ability to identify and interpret XAI within AI systems, and a high valuation of XAI importance suggests an interest in seeking out XAI. We explore the relationships between individuals’ self-perceived AI competency, their appreciation of XAI importance, belief in XAI availability, and attitudes toward AI (measured through the fear and acceptance of AI). We investigate the mediating role of belief in XAI availability between AI competency and belief in XAI importance (predictors) and its impact on both the acceptance and fear of AI (outcomes). Our study, conducted through an online survey across two distinct cultural frameworks, the Arab and UK contexts, reveals consistent results, confirming the associations and mediating relationship. The findings underscore the significance of making XAI interfaces visible and understandable, even for less interested and technically competent AI users, as a means to enhance acceptance and alleviate fear.
To examine if artificial intelligence (AI) has the potential to complement human health providers in telemedicine, this experiment tests different factors that may affect patients' intention to adhere to prescriptions. Participants (N = 261) were randomly assigned to one of 12 conditions: 2 (doctor type: human vs. AI) x 2 (illness severity: high vs. low) x 3 (message richness: text-only prescription vs. with audio vs. with audiovisual). They were then asked to indicate their perceived credibility of the doctor and intention to adhere to the prescription. Results showed that people reported greater perceived credibility toward a human doctor than an AI doctor which, in turn, was positively associated with adherence intention. Moreover, there was a significant interaction between doctor type and message richness, such that the difference in adherence intention for human versus AI doctor was significantly smaller in the text-with-audiovisual condition than that in the text-only condition, indicating that people may start paying less attention to the source and more to the message cues as the message gets richer. These results reflect user responses to a specific type of AI system within a controlled interaction context, and future research should explore how alternative system designs might influence credibility and adherence intention differently.
This systematic literature review examines the integration of artificial intelligence (AI) into automated and remote usability and user experience (UX) evaluation methods. Synthesizing insights from 55 peer-reviewed articles published between 2014 and 2024, the review identifies key AI technologies, such as machine learning, large language models (LLMs), generative AI (GenAI), and ChatGPT, and their roles in enhancing UX evaluation practices. While these technologies contribute to behavior modeling, sentiment analysis, feedback generation, and user simulation, their increasing use also introduces critical challenges. AI models often function as opaque “black boxes,” raising concerns about transparency, hallucinated outputs, contextual misinterpretation, and data bias. The review underscores the need for explainable, human-in-the-loop AI systems, standardized evaluation frameworks, and responsible deployment practices. Specific implications for practice include integrating explainable AI (XAI) methods such as SHAP and counterfactual explanations to improve the transparency of UX insights; adopting domain adaptation and transfer learning to improve generalizability across platforms, demographics, and task contexts; and ensuring human oversight through human-in-the-loop mechanisms to mitigate risks such as hallucinations and contextual misinterpretation. This study offers a comprehensive foundation for both future research and informed adoption of AI technologies in UX workflows, supporting more efficient, data-driven, and human-centered evaluation strategies.
This paper reports on the investigation of low learning management system (LMS) adoption by Ghanaian lecturers and proposes a statistical model to understand and predict the influence of perceived usability (PUsab) on lecturers' LMS usage. A total of 255 lecturers from four higher education institutions (HEIs) in Ghana responded to the survey during the 2023/2024 academic year. Partial least squares structural equation modeling (PLS-SEM) was applied to test the proposed research model. The PLS-SEM results confirmed that human factors, organizational support, and social influence have a statistically significant impact on the PUsab of LMSs, and specifically, PUsab was found to have significantly influenced LMS use among lecturers. In addition to proposing the inclusion of the usability construct in modeling LMS adoption, the paper makes a knowledge contribution by mapping factors from the technology acceptance literature (attitude acceptance) to those from the usability literature (behavior acceptance) toward understanding LMS adoption. The practical contribution of this study is twofold: first, to inform designers and guide HEI managers in considering the contextual factors during the prepurchase evaluation of LMSs; and second, to provide a validated questionnaire which can be employed in similar studies as a reliable tool for assessing LMS adoption.
Alzheimer’s disease (AD), a progressive neurodegenerative disorder whose symptoms become apparent late in the disease process but are only in the early stages of development, creates challenges that demand that this disorder be diagnosed early to reduce its progression. This research work has also suggested a lifelong learning system that insists on the combination of MRI and spike neural signals (EEG data) for the early detection of AD using automated deep learning. By recurrent catastrophic forgetting through elastic weight consolidation (EWC) and memory replay, the model learns from new data while retaining past knowledge that is vital in health-care-related environments where patient information is ever-expanding. Three levels of integration between MRI and spike neural data have been used in this work: early, mid-, and late fusion. Experimental results show that mid-fusion gives better performance than other approaches of 86% accuracy, 84% sensitivity, and 88% specificity for AD identification, which can provide the structure of MRI and temporal EEG signals to identify the AD patient. Early fusion proved a capability to integrate general MRI-EEG correspondences effectively, as integrated late fusion showed the capacity for handling diverse qualities of inputs by analyzing both models independently. The results of the study affirm the effectiveness of the continuous learning multimodal fusion strategy in improving both sensitivity to early AD biomarkers and data heterogeneity. The proposed approach seems to be promising for scalable and real-time diagnostic solutions suitable to bring heterogeneous clinical data and the incremental nature of medical data into the diagnosis of early-stage AD and patient management.
