Search Results (1,159)

Electroencephalography (EEG) has emerged as a promising non-invasive tool for the diagnosis of neurodegenerative disorders, and artificial intelligence (AI) has shown significant potential in this domain, as demonstrated by recent studies. However, strong inter-subject variability remains a major challenge, limiting the ability of AI-based models to learn disease-specific features that generalize across individuals, thereby hindering the development of clinically deployable subject-independent systems. In this work, we propose a cross-subject, AI-based EEG classification framework to distinguish between Alzheimer’s disease (AD), Creutzfeldt–Jakob disease (CJD), and healthy control subjects using clinical EEG data collected from a local hospital. A lightweight hybrid deep learning model is developed, combining a two-layer one-dimensional convolutional neural network with a two-layer Transformer encoder to capture both local temporal patterns and long-range dependencies in EEG signals. The proposed model achieves an average classification accuracy of $97\%$, representing a $3\%$ improvement over a baseline model evaluated on a cohort of 36 subjects. To assess deployment feasibility in real-time clinical settings, the trained model is implemented and evaluated on an edge-AI platform (NVIDIA Jetson AGX Orin), demonstrating energy efficiency for the inference with a compact model footprint. These results indicate that the proposed approach provides an accurate, efficient, and practically deployable solution for subject-independent EEG-based classification of neurological disorders. Full article

Recent advances in large language models (LLMs) have enabled new forms of software creation through natural-language interaction. However, many AI-assisted coding tools continue to assume familiarity with development environments, programming workflows, and technical conventions, which may limit accessibility for early-stage learners and communities historically underrepresented in digital participation. This challenge is particularly relevant in Aotearoa New Zealand, where Māori and Pacific peoples remain underrepresented across STEM and technology pathways. This paper introduces TechTahi, a browser-based, syntax-free AI-assisted platform designed to support low-barrier digital creation through natural-language prompts and immediate in-browser previews. The study had two aims: to describe the design rationale and workflow of TechTahi and to explore early learner perceptions following initial use of the platform. An exploratory pilot design was employed. Five participants completed a post-use survey after hands-on interaction with TechTahi. Responses were analysed descriptively, with open-ended feedback reviewed for recurring themes. Findings suggested generally positive perceptions of accessibility and ease of use, particularly the ability to create working applications without prior coding knowledge. Participants also identified opportunities for culturally relevant features, including language support and locally meaningful design elements, alongside areas for improvement such as clearer onboarding guidance and reduced information density. These preliminary findings suggest that syntax-free, culturally responsive AI creation tools may offer promising pathways for widening participation in digital learning. Further research with larger and more diverse samples is needed to evaluate longer-term educational impact. Full article

Large Language Models (LLMs) are increasingly explored for educational tasks such as grading, yet their alignment with human evaluation in real classrooms remains underexamined. In this study, we investigate the feasibility of using OpenAI GPT-4o to evaluate short-answer quizzes and project reports in an undergraduate Computational Linguistics course. We collect responses from approximately 50 students across five quizzes and receive project reports from 14 teams. LLM-generated scores are compared against human evaluations conducted independently by the course teaching assistants (TAs). Our results show that GPT-4o achieves strong correlation with human graders (up to 0.98) and exact score agreement in 55% of quiz cases. For project reports, it also shows strong overall alignment with human grading, while exhibiting some variability in scoring technical, open-ended responses. We release all code and sample data to support further research on LLMs in educational assessment. This work highlights both the potential and limitations of LLM-based grading systems and contributes to advancing automated grading in real-world academic settings. Full article

Complex knowledge work depends on individual output and on public exchanges that document problems, evaluate contributions, route expertise, and preserve reusable knowledge. Software work makes this infrastructure unusually visible through GitHub issues, reviews, comments, mentions, and cross-project ties. As generative AI coding tools become private, on-demand sources of task support, it is unclear whether productive output remains tightly coupled with participation in this GitHub-visible public knowledge infrastructure. This study examines that question using a balanced panel of approximately 38,000 freelance developers on GitHub observed quarterly from 2019 to 2025 (approximately 1,080,000 person-quarter observations), estimating within-person changes in the association between a Productivity Index and a Social Connectivity Index. Two-way fixed effects models estimate a substantively large weakening after mid-2022 (−0.138 SD, about 44 percent of the pre-AI slope), and the pattern remains stable across alternative operationalizations, model specifications, and sample definitions. A survey-linked subsample (n = 237) provides individual-level triangulation: the weakening aligns with developers’ self-reported AI adoption dates, and heavier AI users exhibit larger decoupling. Decomposition by exchange function is selective: public exchanges with more direct private AI support pathways (information seeking, troubleshooting, preliminary evaluation) weaken more than exchanges anchored in contextual judgment and new-tie formation. This study documents a large-scale behavioral decoupling between productive output and visible GitHub-based public knowledge-infrastructure participation in a real-world problem-solving setting. The pattern is consistent with cognitive offloading as one micro-level pathway, while direct process evidence is left to future work. Full article

The emergence of Artificial Neural Networks (ANNs) and their deep learning form called Artificial Intelligence (AI) opened a new path to improve energy efficiency and the indoor environment. A small collaborating network team is now extending the passive house approach, in a book entitled Retrofitting, the Energy and Environment of Buildings (Gruyter Publishers), and presenting generalized AI modeling in the following paper. This concept uses a long-term neural network with a short-term memory (LSTM) and three stages (training, validation, and test) for optimalization to hourly data collected for one full year. The non-residential buildings are less affected by the space occupants. This paper examines the feasibility of a uniform, climate modified technology, as our objective is to create a universal and affordable approach to buildings assisting in slowing the rate of climate change. Hence, the idea of creating a generalized neural network for predicting electricity consumption linked with weather conditions was born. This network is to forecast the electricity consumption for buildings linked to the local weather conditions, but different categories of buildings are put together in one set. While this will lower the large set precision, still our question is if such a network would work. If so, in the future we will create multi-variant, local residential systems with the capability of predicting energy use. Full article

Geospatial Artificial Intelligence (GeoAI) enables the automated generation of built environment map features, such as building outlines/footprints, on a global scale. However, the integration of these AI-generated datasets into Volunteered Geographic Information (VGI) platforms like OpenStreetMap (OSM) risks incorporating `AI slop’, consisting of geometrically inconsistent/unreliable data, into the online map. While the OSM “Code of Conduct for Automated Edits” provides a policy framework for data ingestion, it lacks a machine-enforceable mechanism for real-time quality gating. This paper proposes a GeoAI-Gatekeeper to perform this task—an automated process that applies empirical Acceptable Quality Thresholds (AQT) to address the GeoAI data governance problem. Because the Gatekeeper utilizes an intrinsic, no-reference evaluation of geometric fidelity, it can assess incoming AI-generated data streams in real-time without requiring ground-truth benchmarks. Importantly, it focuses exclusively on the geometric validation of building footprints, acknowledging for now that semantic enrichment, such as tagging, remains a human-centric task. The presented GeoAI-Gatekeeper is a working prototype developed for a specific urban area, systematically triaging incoming AI-generated data into three tiers; Auto-Accept, Manual Review, and Reject. It provides a Web-GIS interface for Human-in-the-Loop (HITL) functionality to ensure the OSM community remains the final arbiter of acceptable data quality. Testing the Gatekeeper in Dublin (Ireland) demonstrates that our solution can auto-ingest 93.6% of features with a 14x reduction in human review effort while still adhering to OSM’s cartographic integrity standards. By implementing qualitative community guidelines into machine-enforceable thresholds, our approach introduces a viable methodology for next-generation hybrid VGI systems. Importantly, it ensures that the transition towards automated data ingestion reinforces, rather than undermines, the reliability of global crowd-source mapping datasets. Full article

Artificial intelligence (AI) is increasingly used in scientific research, including psychiatry. Exercise addiction (EA) research relies on self-report instruments, most of which lack a functional impairment criterion and cannot account for obsessive passion, potentially contributing to inflated prevalence estimates and limited clinical specificity. This study examined whether Claude AI (Sonnet 4.6) could assist in synthesizing a brief, theoretically grounded screening tool for EA by integrating validated constructs from the Passion Scale (PS), the Exercise Dependence Scale (EDS), and the Exercise Addiction Inventory (EAI). Using structured prompting and instrument upload, Claude AI created a 10-item scale by systematically mapping source-scale content onto established EA frameworks, including the components model and DSM-5 criteria. The instrument included nine Likert-type items and one binary gate for functional impairment to enhance clinical specificity. Although no empirical validation data were collected, a Monte Carlo simulation (N = 500) was used solely to verify that the prespecified unidimensional simulation model produced internally coherent response patterns. Because the simulated data were generated from and analyzed under the same unidimensional model, these results constitute a computational plausibility check rather than empirical structural validation. AI-assisted theoretical synthesis may offer a novel methodological approach to instrument development, though this remains a working hypothesis requiring empirical corroboration. The proposed Exercise Addiction Screening Scale (EASS-10) is presented as a proof-of-concept tool that now requires empirical validation in clinical and exercise populations. Full article

Traditional talent development models were designed before the AI revolution and do not consider artificial agents as possible sources of development. artificial intelligence is quickly infiltrating education spaces—but our thinking about learning has not caught up with how we can productively pair learners with both human and artificial intelligence. Addressing this gap, we introduce Human–AI Symbiotic Theory (HAIST), a novel theoretical framework designed for AI-facilitated environments, which posits how learners can productively leverage both humans and AI as “development partners” across the entire talent development process. We begin with a comprehensive integration of ideas and theory from the literature on talent development, AI for learning, and human–AI collaboration and use these insights to build HAIST for the specific context of talent development. HAIST comprises three mechanisms—Complementary Intelligence Activation (CIA), Dynamic Adaptive Co-Regulation (DACR), and Agency-Preserving Scaffolding (APS)—that are grounded in prior theory and research on topics like sociocultural theory, self-regulated learning, and distributed cognition. We then demonstrate how HAIST can be applied throughout all phases of talent development while highlighting implications for traditionally underserved learners like adult learners, student veterans, multilingual learners, and first-generation learners. We provide an applied example of how the three mechanisms work in tandem to support talent development and discuss points of tension that must be navigated when applying HAIST (e.g., between adaptation and optimization vs. agency). Lastly, we highlight how considerations of ethics and learner rights (algorithmic bias, learner voice, etc.) should be considered when operationalizing HAIST. Overall, HAIST can serve as a foundational theory to not only understand how talent development should occur between learners and both humans and AI, but also to consider the process of instruction design in AI-mediated learning environments. Full article

Recent advances in forensic genetics are rapidly transforming the field from traditional DNA profiling toward integrative and predictive genomic approaches. While short tandem repeat (STR)-based typing remains the gold standard for human identification, emerging technologies such as massively parallel sequencing (MPS), forensic genetic genealogy (FGG), and artificial intelligence (AI)-driven bioinformatics are expanding the scope of forensic investigations, with MPS also widely established in clinical genomics, further supporting its application in complex and unresolved cases. This article presents a structured narrative and conceptual review of next-generation forensic genomics, based on selected peer-reviewed studies, technical guidelines, and recent review articles relevant to MPS-based marker analysis, FGG, DNA phenotyping, ancestry inference, AI-supported bioinformatics, validation, and ethical/legal issues. We discuss the transition from STRs to single nucleotide polymorphisms (SNPs) and microhaplotypes enabled by MPS, emphasizing their applications in mixture deconvolution, kinship analysis, and degraded DNA samples. The role of FGG in cold case resolution is examined, alongside methodological, legal, and ethical considerations related to the use of public genetic databases. Furthermore, we explore recent developments in DNA phenotyping and ancestry inference, focusing on predictive models of externally visible characteristics (EVCs) and their forensic utility. Particular attention is given to the growing impact of AI and machine learning in data interpretation, probabilistic genotyping, and pattern recognition across complex genomic datasets. Finally, we address current limitations, including technical standardization, population biases, data privacy concerns, and the need for robust validation frameworks. Rather than providing a systematic review, this work aims to synthesize current developments into an operational framework for integrated forensic genomics, distinguishing forensic intelligence, probabilistic interpretation, confirmatory testing, and evidentiary use. By integrating technological, analytical, and ethical perspectives, this review proposes a conceptual framework for integrated forensic genomics, in which genomic data are used not only for identification but also for forensic intelligence generation. Full article

Student teachers can seek help from large language models when writing their academic texts during their initial teacher training, and teacher educators can use large language models to evaluate the submitted academic texts. However, existing research presents inconsistent evidence regarding whether large language models assess academic work in ways comparable to human evaluators. To our knowledge, few studies have examined evaluations made by both student teachers and teacher educators alongside those generated by large language models. This study addresses two research questions concerning how academic texts are evaluated by student teachers, teacher educators, and large language models. First, we found that the two large language models showed agreement with each other but did not consistently align with the evaluations provided by either the student teachers or the teacher educator. Second, the large language models produced substantially longer evaluation texts that closely followed the structure of the assessment criteria but struggled with evaluating the discussion sections. Although the large language models offered practical suggestions for improving academic texts, their feedback did not emphasize the same aspects highlighted by the teacher educator. Implications for practical use of generative AI-tools and needs for further research are discussed. Full article

Peripheral nerve blocks can effectively reduce the use of general anesthesia and opioids in situations where robust pain management is critical, such as severe extremity trauma and hip, femur, and knee surgeries. Despite these benefits, nerve blocks are underutilized due to the high skill required to accurately insert a needle and safely deliver local anesthetic. To overcome this challenge, ultrasound image guidance enabled by artificial intelligence (AI) offers a semi-automated solution for regional anesthesia delivery by non-specialists. As a first step towards realizing an integrated platform for AI-guided nerve blocks, the main objective of this study is to develop and characterize deep learning algorithms to interpret anatomical landmarks on ultrasound images in real time and identify aimpoints for needle placement. Our AI system was trained on over 55,000 images from 20 porcine models and demonstrated an average area under the precision–recall curve of 0.92 (SD = 0.03) for in vivo landmark detection in the femoral nerve region. In prospective live animal testing, aimpoint identification had a 98.3% success rate with an average time of 40.5 s (SD = 33.5). Future work will focus on integrated testing with handheld robotics towards a more accessible method for delivering regional anesthesia in settings from point of injury to medical transport to hospitals. Full article

Despite its many limitations, peer review is the most preferred research assessment scheme for research proposal assessment at the individual level. Although scientometric assessment offers effective assessment frameworks, certain limitations, including the proven and potential misuse of scientometric indicators, hinder its wide adoption. Informed peer review is viewed as an effective way of harnessing the advantages of peer review and scientometric/quantitative assessment wherein one may complement the limitations of the other. Informed peer review frameworks are still prone to many inherent challenges in scientometric assessment and peer review. The importance of intelligent review frameworks that can be more advanced and effective than informed review frameworks lies there. With the advent of AI and generative AI (GenAI), a plethora of opportunities are available to convert informed peer review frameworks to intelligent review frameworks but not without challenges and concerns. In this work, we discuss the possible opportunities for effective AI intervention in an existing informed peer review framework to transform it into an intelligent review framework. Although the selected existing informed peer review framework emphasized the ‘novelty first’ policy, it did not provide any means or guidelines to execute it. The proposed conceptual ‘intelligent review framework’ addresses this very well by exploring the effective use of AI/ML techniques for the process and is envisioned to have the flexibility to adapt to future technological developments in AI, GenAI, etc. Possible challenges and a roadmap for possible evolution with anticipated technological changes, etc., are also discussed. Full article

Recent advances in medical image segmentation have introduced a wide spectrum of deep learning paradigms, including 2D/3D convolutional neural networks (CNNs), transformer-based architectures, vision-language models (VLMs), prompt-driven foundation models such as Segment Anything Model (SAM), and diffusion-based approaches. Although these methods have demonstrated remarkable performance across MRI, CT, PET, ultrasound, and endoscopic imaging, the rapid proliferation of architectures has created methodological uncertainty regarding optimal model selection under varying clinical and data constraints. Existing surveys primarily focus on architectural categorization, yet provide limited guidance for decision-oriented model selection. This study presents a comprehensive and decision-guided survey that systematically analyzes segmentation paradigms across imaging modalities, task types, dataset characteristics, and evaluation protocols. Beyond taxonomy, we propose a practical model selection framework that links clinical scenarios, such as small lesion detection, multi-organ 3D segmentation, limited-data regimes, and domain shift, to appropriate segmentation strategies. Furthermore, robustness, generalization, annotation variability, and benchmarking reproducibility are critically examined. By integrating architectural taxonomy, cross-modal comparative analysis, and a structured decision framework, this work provides a clinically oriented roadmap for selecting segmentation methods and highlights future research directions toward reliable and reproducible medical AI systems. Full article

Generative artificial intelligence (AI), especially large language models (LLMs) that can write and debug code, is changing how students approach programming work in engineering education. Unlike more open-ended conceptual or modeling tasks, programming fits closely with what these systems do well: generating syntax, fixing errors, building procedural logic, and completing code structures. Hence, programming coursework may be one of the areas in which AI changes performance patterns in a measurable way. This study examines whether that shift appears in actual student outcomes. Using a retrospective pre/post design, it compares results from a pre-AI period (2021–2022) with results from a post-AI period (2023–2025), when generative AI tools became widely available to students. The focal assessment is a comprehensive programming project graded with the same rubric across multiple sections and terms. Performance is evaluated through descriptive statistics, distributional comparisons, and mastery thresholds (≥80%). The post-AI period shows a rise in overall scores, along with strong clustering near the top of the scale. Lower- and middle-range scores become much less common, most students fall in the highest score band, and overall variability declines. These results suggest that generative AI acts as a procedural equalizer in programming contexts, referring to the role of generative AI in reducing performance differences by assisting with rule-based, syntax-driven, and execution-oriented aspects of tasks, thereby raising baseline outcomes while compressing variation among students. It appears to raise lower-end performance and make outcomes more consistent, but it also narrows the spread among stronger students and creates a ceiling effect. That pattern raises questions about assessment validity, skill differentiation, and what “mastery” means when AI can handle much of the procedural work. Using multi-term data from authentic online courses, this study adds empirical evidence to the growing literature on AI in engineering education and identifies programming coursework as a setting where generative AI may have already changed performance dynamics in a structural way. Full article

The importance of system safety elevates with the introduction of novel technologies in the aviation industry. With the rise of system complexity, regular safety practices include iterative workflows and heavy reliance on expert knowledge. For the development of modern, efficient aircraft systems, there is a need for innovative approaches. This paper presents a tool, OSIRIS (operational safety and integrated risk analysis), that supports safety and risk analyses utilizing artificial intelligence (AI) concepts. Developed as a key safety feature within the HADES modeling framework, OSIRIS aligns with an architecture-based design approach for abstract system modeling, adhering to model-based systems engineering (MBSE) principles and standards. It currently aids safety engineers in formulating system failure cases consistent with functional hazard assessments (FHA), representing model-based safety assessment (MBSA) in compliance with SAE ARP4761A. The methodological concepts and their implementation in OSIRIS are demonstrated considering an abstract system model from aeronautical applications. The generated results were evaluated against the system context to confirm compliance with the FHA process required for certification. Further, the future work will explore refining OSIRIS’s capabilities and its application cases. Full article