Mechanisms linking epistemic curiosity and learning performance: The multifaceted role of mind wandering from trait and state analysis

This systematic review synthesizes 64 empirical studies to examine how Generative AI (GenAI) shapes learning in Computer Science Education (CSE), particularly in programming, debugging, algorithmic reasoning, and computational problem-solving contexts. Grounded in Constructivist, Sociocultural, Cognitive Load, Adaptive Learning, and Metacognitive Learning theories, the review adopts an integrative perspective to analyze how GenAI-driven adaptivity, AI output qualities, hallucination dynamics, and cognitive–affective regulation influence learners’ interpretation, cognitive processing, and learning outcomes. Findings reveal a dual impact of GenAI in CSE. On the negative side, hallucinated or misleading outputs can increase extraneous cognitive load during programming and debugging and promote over-reliance on system-generated content. They may also perpetuate inequities due to limited access in low-resource settings or insufficient support for culturally and linguistically diverse learners. These effects can disrupt error detection, self-monitoring, and problem-solving, leading to impaired learning performance, and widened educational disparities. On the positive side, when embedded within structured, equitable, and pedagogically grounded environments, GenAI supports reflective programming practice by promoting self-monitoring, verification, and strategic adjustment, thereby enhancing problem-solving skills, engagement, and personalized learning outcomes. By framing learning performance, hallucination dynamics, and problem-solving as interconnected dimensions of GenAI-supported computing education, this review provides a theoretically coherent and pedagogically grounded lens for understanding how GenAI reshapes learning in CSE. The review’s novelty lies in its integrative conceptual framework, offering actionable insights for designing equitable, cognitively balanced, and instructionally effective GenAI-supported learning environments.

The monitoring of one's own learning progress is a key process in models of self-regulated learning and a key predictor of self-regulated learning and academic success. Judgments of learning (JOLs) are an established measure for assessing people's monitoring of learning and have been found to predict learners' subsequent performance as well as effort regulation. However, most studies have been conducted in laboratory settings, involving relatively artificial learning materials and low-stakes tests. We evaluate the predictive validity of JOLs for learning performance and effort regulation in an ecologically valid learning environment by requesting aggregate JOLs in an intelligent tutoring system. 90 German university students used an intelligent tutoring system that provided practice exercises for self-regulated preparation for a statistics exam over the course of a semester. Aggregate JOLs for each chapter of the statistics course were assessed once per week (279 assessments in total). Dependent variables were learning performance as well as absolute and relative learning effort for each chapter, derived from the intelligent tutoring system's log files. JOLs significantly predicted learning performance ( β = 0.20, p < .001) and effort regulation ( β absolute = −0.12, p < .001, β relative = −0.07, p = .002). The present research demonstrates that JOLs have predictive power in real-world learning. It thus bridges the gap between experimental cognitive research and applied educational research on metamemory and self-regulation. • Judgments of learning (JOLs) are predictions of one's own future performance. • Little is known about JOL accuracy in ecologically valid learning environments. • We examined JOL accuracy in an intelligent tutoring system used for exam preparation. • JOLs predicted effort regulation and performance during exam preparation.

Conventional grading in higher education assigns fixed weights to different assessment components, remaining unchanged across time and students, and offering limited flexibility to reflect individual learning trajectories. This study introduces a gamification-inspired assessment design termed Formative-Linked Exam Weighting (FLEX), integrated with Rhythmic Engagement Pedagogy (REP), where each student’s final exam initially carries 100% of the grade and is gradually reduced according to learning-progress points accumulated from formative components, such as quizzes, group tasks, flipped sessions, and inquiry-based milestones. Implemented in an 18-week undergraduate Circuit Theory course, the approach individualized exam weights; by semester’s end, the quartiles were 18.7%, 32.5%, and 44.7%. Compared with fixed-weight formats, FLEX increased voluntary attendance, reduced withdrawals, and improved exam outcomes. Correlation analyses showed that formative engagement aligned with professional-content mastery, REP sustained participation across scales, and FLEX supported learner autonomy. Thus, the proposed assessment provides a scalable framework for enhancing engagement and learning performance. • Formative-Linked Exam Weighting (FLEX) redefines the conventional fixed-weight grading. • FLEX uses internal gamification, converting formative progress into grade rewards. • Rhythmic Engagement Pedagogy (REP) structures alternating learning contrasts. • Combined DGS–REP model sustains engagement and reduces course withdrawals. • The approach enhances mastery, autonomy, and motivation in EMI STEM instruction.

In prostate cancer diagnosis, a deep learning model shows increased susceptibility to MRI quality variations compared to radiologists

Beyond Human Variability: Deep Learning for Intravascular Ultrasound Segmentation With Noisy Labels.

A scoping literature review of prompt engineering for bridging students AI literacy in higher education

Behavioral and neurophysiological evaluation of pyrethroid effects on honey bee olfactory perception and learning.

Acori tatarinowii Rhizoma-Curcumae Radix herbal pair ameliorates cognitive impairment and suppresses neuro-inflammation via Ca2+/CaMKKβ/AMPK/mTOR pathway in Alzheimer's disease.

Deeper insights into the learning performance of Stochastic Configuration Networks

The influence of annotators' experience on radiomics-based machine learning performance in colorectal liver metastases characterization: Impact and mitigation strategy

Cross-validation strategies under data dependency: An example with anemia prediction in sheep using ocular conjunctiva images.

Effective agricultural pest management, crucial for global food security and ecosystem balance, demands robust identification systems capable of adapting to dynamic environments. While deep learning shows promise, current methods often fail in practical class incremental learning scenarios, suffering catastrophic forgetting when encountering learning new pest species. This limitation hinders the development of truly adaptive tools for incremental pest recognition. Addressing this gap, we aimed to create a framework integrating advanced artificial intelligence (AI) with ecological needs for continuous and reliable pest recognition. We propose PestCLIP, a framework for incremental pest recognition leveraging contrastive language-image pretraining (CLIP) model. To combat forgetting, PestCLIP employs dual-prompt tuning and a unique Concept Pool strategy that captures essential class features without extensive data replay. Crucially, it incorporates Prediction Distribution Calibration through incremental logit adjustment. Tested across diverse agricultural pest datasets (Li's, AgriInsect200, and Farm Insect) and general benchmark (mini-ImageNet), PestCLIP demonstrated superior class incremental learning performance, achieving 97.50% accuracy on Li's dataset with only a 5.55% drop when integrating new pest classes. Extensive testing on diverse agricultural pest datasets demonstrates the superiority of PestCLIP in incremental pest recognition tasks. The visualization results confirm that PestCLIP effectively preserves class-specific concepts and mitigates prediction bias through distribution calibration. The proposed PestCLIP marks a pivotal step in advancing incremental pest recognition, enhancing the adaptability and reliability of smart pest management systems in dynamic agricultural environments. © 2026 Society of Chemical Industry.

Evaluating adaptive and generative AI-based feedback and recommendations in a knowledge-graph-integrated programming learning system

Soft robotic actuators often require relatively high driving voltages, which limit their portability, safety, and compatibility with compact electronic systems in wearable haptic interfaces. Achieving strong electromechanical coupling at low voltage while maintaining mechanical compliance remains a key challenge for soft actuator design. Here, we present an origami-mediated low-voltage electret soft robotic actuator that integrates mechanical compliance and electrical functionality within a symmetric multilayer architecture. Two double-layer fluorinated ethylene propylene (FEP) electret films with enclosed micro air-cavity arrays are positioned on both sides of a folded copper origami structure. The origami layer acts as a compliant electrode with a tunable spring-like response, while the air-cavity arrays promote high surface potential and stable charge retention. By jointly optimizing electret charging and origami stiffness, the actuator produces perceptible vibrotactile feedback at driving voltages as low as 20V and supports reliable tactile digital recognition at 70V using a 7-segment actuator array. Stable output and durability are maintained over 10h of high-frequency operation. Application in a virtual reality piano training task further demonstrates statistically significant improvements in motor learning performance and perceived immersion. This approach offers a compelling pathway toward compact, low-voltage human-machine haptic interfaces with robust tactile performance.

Effective leadership for learning, innovation and sustainable organizational performance

Academic performance (AP) prediction is crucial for recognizing at-risk students and enhancing learning outcomes. Traditional statistical models often fail to capture temporal and behavioral patterns. Deep learning (DL) approaches offer improved accuracy and adaptability by leveraging multi-dimensional student data for predictive analysis. The objective is to advance a robust predictive model that predicts students' AP using multi-dimensional data, integrating temporal, behavioral, and demographic features. Students' learning performance data for n = 2000 is collected from multiple sources, including student grades, attendance, learning management system (LMS) interactions, psychometric surveys, and demographic records. Collected data undergoes preprocessing steps, including handling missing values using K Nearest Neighbor Imputation (KNNI), outlier removal, and normalization. Principal Component Analysis (PCA) is employed to decrease dimensionality and extract relevant characteristics from high-dimensional datasets. A novel Gated Long Short-Term Memory Unit is optimized with Dove (GateLSTMU-Dove) to capture temporal dependencies and student engagement patterns. GateLSTMU identifies time-dependent patterns in educational data to support accurate performance forecasting. Dove optimizes model parameters efficiently, enhancing convergence speed and predictive accuracy of the GatedLSTMU. Python 3.10-based experiments demonstrate the model's superior performance. GateLSTMU-Dove achieved lower error metrics and higher classification accuracy (98.85%) compared to baseline methods. Visualization of predictions confirmed accurate forecasting and interpretable temporal patterns in student performance. The GateLSTMU-Dove effectively predicted academic outcomes using multi-dimensional student data. It provides interpretable insights, supports early intervention strategies, and demonstrates a scalable, reproducible approach for data-driven AP management.

Accurate prediction of the temporal dynamics of biological systems is crucial for informing timely and effective interventions, e.g., in ecological or epidemiological contexts, or for treatment adjustments in therapy. While machine learning has proven its capabilities in generalizing the underlying non-linear dynamics of such systems, unlocking its predictive power is often restrained by the limited availability of large, curated datasets. To supplement real-world data, informing machine learning by transfer learning with synthetic data derived from simulations using ordinary differential equations has emerged as a promising solution. However, the success of this approach highly depends on the designed characteristics of the synthetic data. We suggest scrutinizing these characteristics, such as size, diversity, and noise, of ordinary differential equation-based synthetic time series datasets. Here, we demonstrate how to systematically evaluate the influence of such design choices on transfer learning performance. We conduct a proof-of-concept study on three simple, but widely used systems and four real-world datasets. We find a strong interdependency between synthetic dataset size and diversity effects. Good transfer learning settings heavily rely on real-world data characteristics as well as the data's coherence with the dynamics of the model underlying the synthetic data. We achieve a performance improvement of up to 95% in mean absolute error for simulation-based transfer learning compared to non-informed deep learning. Our work emphasizes the relevance of carefully selecting properties of synthetic data for leveraging the valuable domain knowledge contained in ordinary differential equation models for machine-learning based predictions. The code is available at https://github.com/DILiS-lab/opt-synthdata-4tl .

ABSTRACT Feeling included at school influences a child’s development positively. Inclusion encompasses the aspects of emotional inclusion, social inclusion and academic self-concept. While these three aspects of inclusion influence students’ engagement, self-esteem, learning and academic performance, research lacks clarity on inclusion concerning special education needs (SEN). Therefore, this study aimed to examine possible differences in the perception of inclusion due to low and high SEN from the students’ and teachers’ perspectives. Data from 74 inclusive classes, with 1374 students (3rd to 6th year, 279 low, 49 high SEN), was analysed, using the Perception of Inclusion Questionnaire (PIQ). Structural Equation Modelling revealed significant differences between students with high SEN and their peers without SEN across all aspects of inclusion (both perspectives). Teachers rated all aspects of inclusion of high SEN students significantly lower compared to low SEN students and students without SEN. High SEN students themselves rated their social inclusion significantly lower compared to low SEN students. Low SEN students rated their academic self-concept significantly lower compared to students without SEN. Students and teachers differed in their perceptions based on the SEN level. Effects ranged from r. = -.056 to r. = -.513. The study contributes to understanding disparities in inclusive education.

Gami-Yukmijihwang-Tang Ameliorates Scopolamine-Induced Cognitive Impairment by Restoring Cholinergic Function and Suppressing. Oxidative Stress and Neuroinflammation.