Reconstructing 3D visuals from functional Magnetic Resonance Imaging (fMRI) data, introduced as Recon3DMind, is of significant interest to both cognitive neuroscience and computer vision. To advance this task, we present the fMRI-3D dataset, which includes data from 15 participants and showcases a total of 4,768 3D objects. The dataset consists of two components: fMRI-Shape, previously introduced and available at https://huggingface.co/datasets/Fudan-fMRI/fMRI-Shape, and fMRI-Objaverse, proposed in this paper and available at https://huggingface.co/datasets/Fudan-fMRI/fMRI-Objaverse. fMRI-Objaverse includes data from 5 subjects, 4 of whom are also part of the core set in fMRI-Shape. Each subject views 3,142 3D objects across 117 categories, all accompanied by text captions. This significantly enhances the diversity and potential applications of the dataset. Moreover, we propose MinD-3D++, a novel framework for decoding textured 3D visual information from fMRI signals. The framework evaluates the feasibility of not only reconstructing 3D objects from the human mind but also generating, for the first time, 3D textured meshes with detailed textures from fMRI data. We establish new benchmarks by designing metrics at the semantic, structural, and textured levels to evaluate model performance. Furthermore, we assess the model's effectiveness in out-of-distribution settings and analyze the attribution of the proposed 3D pari fMRI dataset in visual regions of interest (ROIs) in fMRI signals. Our experiments demonstrate that MinD-3D++ not only reconstructs 3D objects with high semantic and spatial accuracy but also provides deeper insights into how the human brain processes 3D visual information.

Brain leakage exposes covert cognitive computations in bodily movements.

The visual system may perceptually process conspecifics more efficiently when they are interacting, versus not, to support social cognitive functions such as group detection. In three experiments, young adult university students were briefly shown dyads (upright or inverted) and made speeded judgments of whether they attended the same location (joint attention) or different locations (non-joint attention). Participants performed worse with inverted stimuli, but this inversion effect was smaller in joint attention conditions. These findings indicate perceptual grouping of joint attention dyads into a single perceptual unit. This joint attention grouping effect was evident when dyads looked towards spatial locations (Experiment 1), towards objects (Experiment 2), and for asymmetrically composed stimuli (Experiment 3). The effect was weaker for non-social directional stimuli (Experiment 1). These data support the idea that two interacting individuals are coded as one socially bound perceptual unit, supporting efficient and rapid social cognitive computations.

Research claims that metacognitive experiences can be classified as types of metacognitive regulation. Formulated in terms of the theory of Attention as Internal Action, this view raises questions about the timing of metacognitive experiences that occur in response to internal experiences. To investigate these questions, this work presents a method for cognitive computation that simulates consecutive internal decisions occurring during the process of taking a digital exam. A new version of the General Internal Model of Attention is proposed and supported by research. It is applied as cognitive architecture in a simulation system to reproduce cognitive phenomena such as the cognitive cycle, internal decision-making, imagery, body actions, learning, and metacognition. Two corresponding groups of Markov Decision Processes were designed as information stores for goal influence and learning, and a Hebbian machine learning algorithm was applied as an operator on the learning models. The timing and consecutiveness of metacognitive experiences were analyzed based on the cognitive cycle results, and several hypotheses were derived. One of them suggests that the first engagement in a metacognitive experience for each question in the exam is delayed over the course of the exam-taking process.

The Smart Study Buddy is an innovative AI-driven, web-based adaptive learning platform that redefines how students interact with and internalize academic content in the digital age. Built upon the convergence of machine learning (ML), natural language processing (NLP), and human-centered design, the system intelligently processes raw study materials—such as notes, lecture slides, research papers, and textbooks—to generate structured, context-aware learning outputs. These include automatically generated summaries, interactive quizzes, flashcards, crossword-style exercises, and keyword extractions, all aimed at enhancing comprehension, retention, and recall efficiency. At its core, Smart Study Buddy employs transformer-based deep learning models (GPT-2 and fine-tuned BERT architectures) to extract semantic meaning from unstructured text and produce concise, accurate summaries that maintain conceptual integrity. The quiz generation module leverages NLP-driven question-answering techniques to produce multiple-choice, true/false, and short-answer questions from the summarized content, encouraging active recall, which is a proven method for long-term memory consolidation. In parallel, the flashcard module and adaptive scheduling system promote spaced repetition, helping learners review difficult topics at optimal intervals based on past performance and engagement metrics. The platform architecture integrates a Python and Streamlit-based front-end interface for real-time interactivity, a SQL-backed database for persistent user data management, and OpenAI’s GPT API for semantic processing. The adaptive scheduler dynamically adjusts daily learning goals using predictive analytics to prevent cognitive overload and improve time management. The progress dashboard visualizes study trends, accuracy rates, and content mastery through analytics charts, while gamification elements such as points, streaks, and badges foster intrinsic motivation and consistent participation. Furthermore, the system features an AI-powered “Chat with Notes” assistant, allowing users to ask natural-language questions about their uploaded materials and receive contextually relevant explanations derived directly from their study corpus. This feature bridges the gap between passive content review and active conversational learning, simulating the experience of an intelligent personal tutor. Smart Study Buddy not only minimizes the effort required for manual summarization, note-making, and quiz creation but also enhances personalized learning through continuous feedback loops and intelligent progress evaluation. By incorporating AI personalization, behavioral analytics, and gamified engagement, it transforms traditional study methods into an adaptive, data-driven, and self-evolving learning ecosystem. Ultimately, this research demonstrates how AI-assisted educational platforms can foster autonomous learning, improve academic performance, and significantly reduce time spent on content preparation. The Smart Study Buddy exemplifies the future of intelligent education systems—an intersection of technology, pedagogy, and psychology—empowering learners to achieve higher efficiency, deeper understanding, and sustained motivation in an ever-expanding information landscape. .. Keywords: Artificial Intelligence · Adaptive Learning · Educational Technology · GPT-2 · Machine Learning · Natural Language Processing · Summarization · Quiz Generation · Flashcards · Gamification · Streamlit · Personalized Study Assistant · Cognitive Computing · Academic Automation

Emotion recognition is a crucial part of cognitive computing. Traditional emotion recognition systems include audio-visual modality. However, a recent trend in recognizing emotions is to use physiological signals such as the electroencephalogram (EEG). EEG signals, together with audio-visual and other physiological signals, improve the performance of emotion recognition systems. This paper presents a systematic literature review on EEG-based multimodal (multimedia) emotion recognition systems for the last five years. Three major research questions are addressed: (1) What kind of learning models are used in EEG-based multimedia emotion recognition? (2) What are the publicly available related datasets? (3) What are the challenges and future directions of this topic? The answers to the research questions are provided in different subsections.

The rapid emergence of intelligent decision support systems in businesses maximizes the transformative power of AI and cognitive computing to increase productivity, optimize decision-making processes and drive innovation. These technologies enable the integration of more complex data sets, providing real-time analytics, predictive modeling, and intelligent insights necessary for informed decision-making where AI-driven systems can learn from history in detail, identifying patterns and providing actionable recommendations, thereby reducing human error, improvements in strategy and even, cognitive computing Through various approaches do-it-yourself It enhances these systems, providing AI with more flexible interactions and adaptive learning.

Detecting deception in emotionally grounded natural language remains a significant challengedue to the subtlety and context dependence of deceptive intent. In this work, weuse a structured behavioral dataset in which participants produce truthful and deceptivestatements under emotional and social constraints. To maintain label accuracy and semanticconsistency, we propose a multilayer validation pipeline combining self-consistencyprompting with feedback-guided revision, implemented through the CoTAM (Chain-of-Thought Assisted Modification) method. Our results demonstrate that this frameworkenhances deception detection by leveraging a sentence decomposition strategy that highlightssubtle emotional and strategic cues, improving interpretability for both models andhuman annotators.

ABSTRACTConstruction 4.0 (C4.0), a direct application of Industry 4.0 principles, is a substantial advancement in road construction projects (RCPs). It involves integrating digital technologies into various stages of RCPs. This study thus assessed the applicability of C4.0 in 112 ongoing RCPs in Tanzania. Data analyses were performed using SPSS 21.0 and AMOS 21.0 software. The exploratory and confirmatory factor analyses were utilized to determine the factors that affect C4.0 on RCPs and to illustrate the effect of C4.0 readiness on the applicability of C4.0 in RCPs. Some of the confirmed technologies include big data analytics, artificial intelligence (AI), building information modeling (BIM), computer simulation, cloud computing, 3D printing, Internet of Things (IoT), augmented reality (AR), automated survey, robotics, digital twin, drones, smart sensor, virtual reality (VR), machine learning, and cognitive computing. These technologies can enhance planning and design, optimize construction processes, and improve maintenance and management. BIM facilitates visualization, clash detection, and coordination between RCPs' stakeholders. Automation and robotics enhance speed, precision, and safety, while IoT and sensors deliver real‐time data for informed, proactive decision‐making. AI and machine learning optimize resource allocation and predict delays. AR and VR can visualize construction plans and guide workers. The RCPs are now ready to implement C4.0 technologies. The analysis revealed that communalities are above the minimum value of 0.5, accounting for 73.54% of the readiness level. The overall readiness level is 2.93 out of 5.0, indicating that companies' readiness to engage in RCPs is operating at Level 2 (intermediate level). Lastly, the study suggests that public road construction agencies adopt strategies to successfully implement C4.0 on road projects, including enhancing employee skills and knowledge, adapting regulatory frameworks, improving data management, securing external support, planning financial arrangements, implementing C4.0 technologies, strengthening existing technological capabilities within the company, and fostering collaboration.

Arabic Natural Language Processing (NLP) is still faced with the complexity of the language’s morphology and the limited availability of quality annotated resources. In this paper, we introduce an open-domain dataset of 5,009 Modern Standard Arabic (MSA) questions labeled according to AAFAQ framework that has11 linguistic and cognitive aspects, e.g., Question Particle, Question Particle Type, Intent, Answer Type, Cognitive Level, and Temporal Context. Based on the AAFAQ Framework (Arabic Analytical Framework for Advanced Questions), the dataset is designed to support semantic and cognitive understanding for Arabic Question Classification and related tasks. The dataset’s effectiveness was validated by fine-tuning state-of-the-art models. AraBERT achieved 100% accuracy on Question Particle Type classification and 94.95% on Intent classification. Integration within a generative question-answering system with Alpaca + Gemma-9B Unsloth improved evaluation metrics, including BLEU (+37.6%), ROUGE-1 (+132%), and BERTScore (+17.3%), validating the dataset’s value in both classification and generation tasks. Despite its broad coverage, the dataset includes underrepresented categories, e.g., Sociology and Volunteering, to be considered in future extensions. AAFAQ is a foundation benchmark for the advancement of Arabic question comprehension, with prospective applications in education, cognitive computing, and multilingual AI system creation.

Abstract The evolution from the Edge-Cloud Continuum to the Cognitive Computing Continuum (CCC) has introduced new challenges which necessitate advanced frameworks that integrate cognitive capabilities to enhance interoperability, adaptability, and resource efficiency. Considering insights from ongoing research and initiatives on the cognitive cloud, we identify the core concepts essential for transitioning to the CCC, including cognitive orchestration, distributed AI, and sovereign data management. We then introduce ENACT, a novel framework designed to embrace these concepts aiming to provide cognitive, highly adaptive orchestration to support modern hyper-distributed and data-intensive applications. Furthermore, ENACT employs bespoke mechanisms to enable dynamic continuum modelling and visibility as well as to facilitate application-level automation and adaptability. This paper presents the motivation behind the ENACT approach, reviews the state-of-the-art across its fundamental technological concepts and highlights its key innovations. Overall, it contributes to the formalization of the CCC architectural principles and, ultimately, to the realization of CCC.

The Evidential K-Nearest Neighbor (EK-NN) classifier has demonstrated robustness in handling incomplete and uncertain data; however, its application in high-dimensional big data for feature selection, such as genomic datasets with tens of thousands of gene features, remains underexplored. Our proposed Granular–Elastic Evidential K-Nearest Neighbor (GEK-NN) approach addresses this gap. In the context of big data, GEK-NN integrates an Elastic Net within the Genetic Algorithm’s fitness function to efficiently sift through vast amounts of data, identifying relevant feature subsets. This process mimics human cognitive behavior of filtering and refining information, similar to concepts in cognitive computing. A granularity metric is further employed to optimize subset size, maximizing its impact. GEK-NN consists of two crucial phases. Initially, an Elastic Net-based feature evaluation is conducted to pinpoint relevant features from the high-dimensional data. Subsequently, granularity-based optimization refines the subset size, adapting to the complexity of big data. Before applying to genomic big data, experiments on UCI datasets demonstrated the feasibility and effectiveness of GEK-NN. By using an Evidence Theory framework, GEK-NN overcomes feature-selection challenges in both low-dimensional UCI datasets and high-dimensional genomic big data, significantly enhancing pattern recognition and classification accuracy. Comparative analyses with existing EK-NN feature-selection methods, using both UCI and high-dimensional gene datasets, underscore GEK-NN’s superiority in handling big data for feature selection and classification. These results indicate that GEK-NN not only enriches EK-NN applications but also offers a cognitive-inspired solution for complex gene data analysis, effectively tackling high-dimensional feature-selection challenges in the realm of big data.

The term Artificial intelligence (AI) was coined by Standford Professor John McCarthy at the Dartmouth conference in 1956. It refers to systems that display intelligent behaviour by analysing their environment and taking action – with some degree of autonomy – to achieve specific goals.[1] According to the Organisation for Economic Co-operation and Development, AI is a machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations, or decisions influencing real or virtual environments.[2] From these definitions, it can be construed that AI systems are designed to operate with varying levels of autonomy and it leverages computers and machines to simulate and mimic human intelligence. AI is typically implemented as a system comprised of both software and hardware.[3] AI utilizes computational power, deep learning algorithms, and graphics processing units (GPU) produced by Nvidia’ to process vast datasets and derive insights to emulate cognitive functions such as tackling language understanding, learning, logical reasoning, problem-solving, and/or decision-making[4]. Deep learning refers to a class of algorithms which are based on artificial neural networks optimized to work with unstructured data such as images, voice, videos and text.[5] While GPU is an electronic circuit which manipulates and modifies the memory for better image output.[6] Deep learning involves huge amounts of matrix multiplications and other operations which can be massively parallelized and thus done on GPUs.[7] AI uses such algorithms to learn how to complete tasks through many rounds of trial and error.[8] AI portfolio also involves Natural Language Processing (NLP), Robotics, Machine Learning, and Cognitive Computing.[9] Machine Learning (ML) is a particularly successful AI application.[10] It identifies patterns between variables in a large dataset. Most ML approaches can be classified as supervised, unsupervised or reinforcement learning.[11] Supervised ML uses labelled training data to develop models in which target results (such as a diagnosis) are known. In contrast, unsupervised ML recognizes patterns or aggregations that occur within data without requiring labelled data.[12] Reinforcement ML uses a system with reward and punishment methods to form a solution strategy to solve some problems.[13] Presently, the adoption and use of AI is rapidly increasing. AI is becoming more widely adopted and integrated into many aspects of daily life, including commerce, health, education, communication, and public service, holding significant impact in almost all areas where human intelligence is involved.[14] AI can be used by businesses and institutions to optimize operations, promote innovations, and empower and supplement staff.[15] For instance, AI models can run in cars to avoid accidents,[16] in smartphones to perform various tasks,[17] in banks to manage investments and loans,[18] and in law enforcement to help officials recover evidence and make law enforcement easier. In the same vein, AI is used in hospitals to aid doctors. AI can effectively analyze and provide valuable insights

The excessive consumption of information, amplified by hyperconnection and the abundance of digital stimuli, has become one of the main contemporary challenges for cognition and mental health. This article analyzes the relationship between information overload and the symptoms of stress and anxiety, increasingly prevalent phenomena in digital societies. The objective is to understand the effects of information excess on cognition and to discuss the possibilities of mitigation offered by generative artificial intelligence, especially large language models (LLMs). The methodology adopted was qualitative in nature, based on bibliographic and documentary review in fields such as cognitive neuroscience, psychology, sociology, and computer science, as well as reports from the World Health Organization. The results indicate that information overload compromises attention, working memory, and decision-making, raising stress and anxiety levels. On the other hand, LLMs have the potential to act as personalized filters, synthesizing and organizing information according to individual preferences, reducing exposure to dispersive digital flows. However, ethical risks, such as algorithmic biases and informational bubbles, must be considered. It is concluded that the conscious use of artificial intelligence can represent an effective strategy for preserving cognitive well-being, provided it is combined with information hygiene practices and policies that ensure diversity in access to knowledge.

This article introduces an enhanced Internal State Dynamics Model (MDEI), a mathematical-computational framework for modeling cognitive-affective states in Artificial Intelligence systems. Cognitive computing represents one of the most challenging frontiers in modern AI, evolving from simplified models to sophisticated dynamic approaches inspired by brain functioning. In MDEI, each internal state is represented by an adaptive three-dimensional vector, surpassing traditional discrete symbolic approaches. The formalism is built on solid foundations of vector algebra, differential calculus, and dynamical systems theory, with an emphasis on didactic clarity and conceptual depth. A strategic literature review, including studies from MIT, Stanford, and recent high-impact journals, situates MDEI within the perspective of AI as a cognitive extension, highlighting its role in facilitating more natural human-machine interactions. Recent research demonstrates that advanced generative systems exhibit behaviors aligned with human cognitive functions, indicating potential for human-machine synergy. MDEI provides a robust framework for developing adaptive, resilient AI in the face of emotional complexity, pointing to applications in cognitive assistants, mental health, and education. Finally, the empirical validation of the model's parameters is critically discussed, underscoring the need for future experimental work based on rigorous emotional assessment methods.

Abstract: The emergence of Organoid Intelligence (OI) marks a transformative shift in artificial intelligence by integrating living neuronal networks with silicon-based systems. This study explores a bio-digital hybrid framework that combines cerebral organoids—three-dimensional neural tissues derived from human stem cells—with neuromorphic computing architectures to emulate advanced cognitive processes such as learning, memory, and adaptive decision-making. A robust methodological pipeline was implemented involving multi-electrode array (MEA) interfaces, signal transduction layers, and predictive modeling algorithms. The electrophysiological outputs of organoids were translated into digital signals using machine learning models including regression, AUC-ROC analysis, and error metrics such as RMSE and nDCG.Comparative experiments between the OI hybrid system and baseline models—Deep Neural Networks (DNNs) and Spiking Neural Networks (SNNs)—demonstrated superior performance of the OI system, with higher accuracy (94.7%), lower error rates, and greater memory retention capacity. Regression analysis confirmed a strong adaptability trend (R² = 0.89) over repeated trials, underscoring the capacity for biological learning within the integrated system.The implications of this research are profound, offering new directions for intelligent computing architectures that transcend traditional silicon limitations. Potential applications include adaptive robotics, brain-machine interfaces, and energy-efficient cognitive processing systems. This work establishes a foundational step toward ethically scalable, biologically integrated AI platforms. Keywords Organoid Intelligence, Hybrid AI, Cerebral Organoids, Neuromorphic Computing, Living Neuronal Networks, Machine Learning, Bio-silicon Interface, Predictive Modeling, Cognitive Computing, Adaptive Systems

How humans resolve the explore-exploit dilemma in decision making is central to how we flexibly interact with both social and non-social aspects of dynamic environments. However, how individual differences in the cognitive computations underlying exploration relate to social and non-social psychological flexibility traits remains unclear. To test this, we probed decision-making strategies in a cognitive flexibility task, a restless three-armed bandit task, and examined how individual differences in cognitive strategy related to social and non-social traits measured by the Broad Autism Phenotype Questionnaire (BAPQ), a well-validated, clinically-relevant, community instrument, in a large (N = 1001) online sample. In contrast to prior links found between exploratory behavior and cognitive rigidity, we found that differences in choice behavior and exploration were primarily associated with social phenotypes as captured by the BAPQ aloof subscale. Higher scores on the BAPQ aloof subscale, indicative of reduced social interest and engagement, were associated with decreased shift rates, increased win-stay/lose-shift behavior, heightened sensitivity to negative outcomes, and reduced exploration. Reinforcement learning (RL) modeling further revealed that reduced exploration in high aloof individuals was driven by lower decision noise rather than increased cognitive rigidity, suggesting that decreased exploratory behavior may reflect a reduced tendency for stochastic exploration rather than an inflexible learning process. Sparse canonical correlation analysis reveals that the strongest loading for these non-social reward-related measures are in fact socially coded items. These results suggest that differences in motivation to seek information, especially in social contexts, may manifest as decreased exploratory behavior in a non-social decision-making task. Our findings additionally highlight the potential for using computational approaches to reveal general cognitive mechanisms underlying social functioning.

Abstract The growing necessity for power‐efficient and cognitive computation mechanisms has driven progress in neuromorphic computing which seeks to imitate the synaptic mechanisms underlying human brain functionality. The drawbacks of traditional computation paradigms which involve a large amount of power utilization and restricted data communication drive the quest for alternative materials and technologies. In this context, 2D materials have proven themselves an especially valuable class of materials for new‐generation neuromorphic devices because of their atomic thickness and distinct electronic attributes. The present review gives a detailed account of advanced techniques enabling the fabrication of neuromorphic devices using 2D materials with a focus on deposition methods and device engineering strategies to enhance synaptic functionalities for energy‐efficient signal processing. This article also explores the role of 2D materials in establishing effective synthetic synapses which are predominant in supporting key functions such as short‐term plasticity (STP) and long‐term plasticity (LTP). This article further addresses key fabrication challenges such as scalability, contact/interface issues, and variability, along with emerging solutions like atomic‐thickness control and heterostructure integration. Through a carefully designed roadmap, this article attempts to blend fabrication processes and 2D material's neuromorphic device physics hence presenting valuable insights in constructing brain inspired computational devices.

Intelligence is increasingly recognized as a critical factor in successful behavioral and emotional regulation. Neuroimaging techniques coupled with machine learning algorithms have proven to be valuable tools for uncovering the neural foundations of individual cognitive abilities. Nevertheless, current EEG studies primarily focus on classification tasks to predict the intelligence category of subjects (e.g., high, medium, or low intelligence), rather than providing quantitative intelligence-level forecasts. Furthermore, the outcomes obtained are significantly impacted by the specific data processing pipeline chosen, which could potentially compromise result generalizability. In this study, we implemented a connectome-based predictive modeling approach on high-density resting-state EEG data from healthy participants to predict their nonverbal intelligence level. This method was applied to three independently collected data sets (n = 255) with different functional connectivity methods, parcellation atlases, threshold p values, and curve fitting orders used to ensure the reliability of the findings. Prediction accuracy, measured as correlation between predicted and observed values, varied significantly across pipeline configurations. The most consistent results across data sets were found in the alpha frequency band. Furthermore, we employed a computational lesioning approach to identify the valuable edges that made the most significant contribution to predicting intelligence. This analysis highlighted the crucial role of frontal and parietal regions in complex cognitive computations. Overall, these findings support and expand upon previous research, underscoring the close relationship between alpha rhythm characteristics and cognitive functions and emphasizing the critical consideration of method selection in result evaluation.

The burgeoning capacity of artificial intelligence (AI) to generate artworks has ignited substantial interdisciplinary interest. However, the absence of a shared conceptual framework has hitherto impeded effective communication and collaboration among cognitive science, computer science, and art theory. This study addresses this lacuna through a comprehensive literature review by developing a transdisciplinary definition of an AI-generated artwork. It is proposed that an AI-generated artwork constitutes the confluence of three essential elements: (1) an autonomous AI-production of a new and surprising idea or artifact, (2) which passes an internal evaluation mechanism embedded in the very same AI, and (3) is considered a candidate of appreciation by a human audience. This definition provides a unified conceptual foundation to facilitate interdisciplinary research and deepen understanding of the nature of AI-generated art. Subsequent research should explore the applicability of this definition to diverse forms of AI-generated artworks and evaluate its implications for artistic practices.