Cognitive Systems Research Articles

Cognitive Emotion Aware Systems Using Multimodal Signals and Reinforcement Learning

Predicting human behaviour is a complex task. Traditional methods often rely on explicit user input or external observation, which can be restrictive and impractical in real-world scenarios. As an alternative, Brain-Computer Interfaces (BCIs) offer a more direct and specific means of accessing cognitive and emotional states, providing valuable insights into human intentions and decision-making processes. This paper proposes a novel method that predicts and suggests personalised emotion-based activities for individual users based on multi-modal sensory data collected from the brain, body, and environment. Our method overcomes the limitations of conventional systems by incorporating a multi-modal data collection set throughout the day to understand user context and intent better. By analysing this data, we predict the emotions-based practice of the user's day. We train our method using state-of-the-art, nature-inspired reinforcement learning algorithms and agent technology to optimise its optimisations and personalised continuously. The performance evaluation shows that the accuracy and F1 score for the proposed method achieved 95.6% and 84%, respectively, achieving 2 to 3% more accuracy than AI-based emotion state-of-the-art detection methods.

Egocentric mental rotation in individuals with multiple sclerosis: relationship with disability and cognitive parameters.

This study aims to examine whether egocentric-based mental rotation is affected in individuals with Multiple Sclerosis (MS) and to analyze its relationship with disability, functional system scores (FSS), and cognitive parameters. The study involved 57 right-dominant individuals with MS and 40 right dominant healthy subjects. Disability was assessed using the Neurostatus-Expanded Disability Status Scale (Neurostatus-EDSS). Participants' performance on mental rotation tasks was evaluated with the Recognise Hand App™, Recognise Foot App™, measuring reaction times (seconds) and accuracy (%) for hand and foot stimuli. Cognitive functioning was assessed using the Brief International Cognitive Assessment for MS (BICAMS). MS patients exhibited lower performance in right hand mental rotation reaction time (p = 0.042), right hand accuracy percentage (p < 0.001), right foot accuracy percentage (p = 0.035). Positive correlations were found between disease duration, EDSS total score, cerebellar FSS, bladder-bowel FSS, and ambulation scores with reaction times, while negative correlations were observed with accuracy percentages (p < 0.05). Additionally, a positive correlation was identified between accuracy percentages and BICAMS (p < 0.05). The study revealed substantial impairments in right-hand and right-foot performance in right dominant individuals with MS. Mental rotation abilities were found to be related to disease duration, higher EDSS and FSS scores, as well as cognitive functioning. Identifying the cognitive domains and functional systems associated with egocentric mental rotation will contribute to better understanding this underexplored area and developing potential treatment strategies to enhance functionality.

Transferable Modulation of Cognitive Control: The Cross-Task Role of Conflict Adaptation in Thematic Roles Assignment in Chinese

Conflict adaptation reflects the dynamic modulation of information processing by the cognitive control system following conflict detection. A central question in language processing research concerns whether control elicited by non-linguistic tasks generalizes across tasks to influence higher-order processes such as sentence comprehension. The present study employed color-word Stroop tasks of varying complexity and, in conjunction with eye-tracking technology, examined their cross-task regulatory effects of conflict adaptation on thematic role assignment in Chinese. Across two experiments, participants read sentences containing either congruent or conflicting thematic roles following Stroop trials with congruent or incongruent stimuli. The temporal dynamics of syntactic processing were captured via eye movement measures. Results indicated that both conflict tasks triggered cross-task conflict adaptation, as evidenced by accelerated syntactic processing and reduced regression behaviors when thematically incongruent sentences followed conflict trials. Notably, the more complex color-word Stroop task imposed greater demands on cognitive control resources and elicited earlier cognitive adaptation effects during comprehension. Theoretically, these findings extend conflict monitoring theory to the domain of language processing, demonstrating that cognitive control mechanisms contribute to real-time syntactic parsing. Methodologically, the use of eye-tracking to examine thematic role assignment provides fine-grained empirical evidence for the interaction between domain-general control and language-specific processing.

TUNeS: A Temporal U-Net With Self-Attention for Video-Based Surgical Phase Recognition.

To enable context-aware computer assistance in the operating room of the future, cognitive systems need to understand automatically which surgical phase is being performed by the medical team. The primary source of information for surgical phase recognition is typically video, which presents two challenges: extracting meaningful features from the video stream and effectively modeling temporal information in the sequence of visual features. For temporal modeling, attention mechanisms have gained popularity due to their ability to capture long-range dependencies. In this paper, we explore design choices for attention in existing temporal models for surgical phase recognition and propose a novel approach that uses attention more effectively and does not require hand-crafted constraints: TUNeS, an efficient and simple temporal model that incorporates self-attention at the core of a convolutional U-Net structure. In addition, we propose to train the feature extractor, a standard CNN, together with an LSTM on preferably long video segments, i.e., with long temporal context. In our experiments, almost all temporal models performed better on top of feature extractors that were trained with longer temporal context. On these contextualized features, TUNeS achieves state-of-the-art results on the Cholec80 dataset. This study offers new insights on how to use attention mechanisms to build accurate and efficient temporal models for surgical phase recognition. Implementing automatic surgical phase recognition is essential to automate the analysis and optimization of surgical workflows and to enable context-aware computer assistance during surgery, thus ultimately improving patient care.

Cognitive styles and behavioral systems: Linking looming cognitive style and reinforcement sensitivity

Cognitive styles and behavioral systems: Linking looming cognitive style and reinforcement sensitivity

Pumping up your predictive power for cognitive state detection with the proper GAINS.

Pumping up your predictive power for cognitive state detection with the proper GAINS.

Childhood P300 predicts development of depressive disorders into adolescence.

The prevalence of depressive and anxiety disorders dramatically increases as children enter adolescence. It is critical to identify etiological factors that can assist in the identification of children most at risk. Event-related potentials are one measure that has demonstrated promise and feasible application in children and adolescents. In particular, the P300 has been extensively employed to examine cognitive system deviations associated with depressive and anxiety disorders. However, this work has primarily focused on adults, and there have been limited prospective investigations, making it unclear whether the P300 can prospectively predict the development of later depressive and anxiety disorders during critical developmental periods, such as adolescence. The present sample included 272 9-year-old children with no history of psychopathology who completed the doors task while continuous electroencephalography was recorded to measure the choice- and feedback-locked P300s. Participants completed follow-up diagnostic interviews through age 15 to determine onset of later depressive and anxiety disorders. A smaller choice-locked, but not feedback-locked, P300 in childhood predicted an increased likelihood of developing first-onset depression by mid-adolescence. Neither P300 predicted the development of anxiety disorders. The present study indicates a blunted choice-locked P300 indexes risk for depressive disorders in adolescence. The choice-locked P300 might be a valuable neural measure for further understanding pathways unique to increasing depression in adolescence.

On the ethno-cultural specificity of color names: Kazakh ⱪonyr ‘brown’ vs. Polish brązowy ‘brown’

The article is based on an assumption of anthropological linguistics, in particular the biocultural theory of meaning (J. Zlatev), that the semantic system of language is determined by a combination of external factors, including environmental determinism. The need to adapt the cognitive system to the environment and geophysical conditions of existence of human groups leads to the fact that cultural worldviews and the corresponding semantic systems of languages differ. The authors consider the relative and culturally variable nature of colour terms using the example of the adjective ⱪonyr in the Kazakh language and its lexical equivalent in the Polish language. The article shows that the presence of a large number of precedent phenomena of the brown colour among Kazakhs is manifested in a more intense functionality of this colour name in the Kazakh language. This concerns several aspects: high frequency, intensive polysemy, phraseology and connotations. To illustrate this, the authors provide language material from various sources, including national corpora of the Kazakh and Polish languages.

Agentic RPA: Enabling self-driven decision-making workflows in enterprise automation

Modern enterprise environments have become increasingly dynamic, requiring automation tools that move beyond the fixed workflows of traditional Robotic Process Automation (RPA). This study introduces the concept of Agentic RPA an advanced automation framework that redefines conventional bots as intelligent, autonomous agents capable of making real-time decisions based on goals and context. Drawing from agent-based programming principles and cognitive system design, Agentic RPA incorporates technologies such as decision logic engines, real-time data interpretation, and large language models to support adaptive and context-aware automation. Centered on the UiPath ecosystem, the paper outlines a modular structure that integrates AI components with event-driven workflows and optional human oversight. Practical applications across ERP, CRM, and HRIS platforms demonstrate measurable benefits, including reduced process cycle times and significantly fewer manual interventions. These results highlight Agentic RPA’s value in enhancing operational agility and resilience, offering a forward-looking path for enterprises seeking scalable, intelligent automation that aligns with evolving business needs.

Cyber Security of Robots: Integrating Safety and Security in Cognitive Social Computers

For social robots to coexist with people, they must be trustworthy and safe. This study examines the implications and relationships between cybersecurity and safety to develop more intelligent, secure, dependable autonomous robots. The initial findings are presented in this publication. Robots are used extensively in the modern world, not just in automated cars and medicine but also in industry, national security, and defence. Cyberattacks against robots and other security concerns are rising with the number of robots. We propose creating a robotic intrusion prevention system (RIPS) that employs a cutting-edge methodology to identify and stop intrusions in cyber-physical systems, including cognitive social robot systems. To mitigate cyber-physical risks, the RIPS uses system modes to specify which robotic system components limit or decrease its functioning when the system is penetrated. The RIPS detects threats at the robotic communication level. Additionally, we show that enhancements in encryption, authorisation/authentication, and physical security can considerably lower the probability of cybersecurity threats on robotic platforms. A dual regulatory framework that governs cybersecurity and physical product safety separately makes it challenging to regulate cyber-physical systems like care robots consistently and effectively. We conceive and talk about the difficulties in controlling the security of cyber-physical systems using the current dual framework, especially the absence of required certificates. Various robotic systems' security levels are examined in multiple domains to ascertain whether they need an upgrade or correction. Additionally, we outline and highlight open challenges that may emerge over the coming years.

Construction of a Value Evaluation System for Fujian Tubao Architectural Heritage Based on Grounded Theory and the Analytic Hierarchy Process

Scientific evaluation of architectural heritage value constitutes a crucial foundation for advancing effective conservation practices and guiding policy development. Fujian Tubao, a distinctive form of defensive vernacular architecture found in southeastern China, integrates military defense, residential functions, and clan-based social organization into a unified spatial structure, making it an important component of China’s regional cultural heritage. In response to current challenges of inadequate preservation and progressive loss of heritage value associated with Fujian Tubao, there is an urgent need to establish a systematic evaluation framework to support and inform conservation efforts. This study centered on Fujian Tubao and proposed a comprehensive value assessment system that could innovatively integrate Grounded Theory with the Analytic Hierarchy Process (AHP). The Grounded Theory was first adopted to extract value dimensions of architectural heritage and their underlying connotations, thereby forming a value cognition system. Subsequently, the AHP was applied to assign quantitative weights and establish the priority order of each dimension, ultimately constructing a value evaluation system that could identify core objects for heritage conservation and clarify the hierarchy of their relative importance. This study achieved methodological integration in both the value extraction and evaluation stages, overcoming limitations of a single-method approach in dimension identification and weight assignment. The framework ensures logical consistency in value structuring and enhances the scientific validity of results. This study formulated a standardized and replicable evaluation framework tailored to the heritage value of Fujian Tubao. It also provides theoretical support for future conservation planning and offers a methodological reference for value assessment across diverse categories of architectural heritage.

Explainable AI for Cybersecurity Applications: A Review Article on Techniques, Deployments, and Usability Challenges

The growing reliance on Artificial Intelligence (AI) in cybersecurity has elevated concerns about the interpretability and transparency of automated decision-making systems. In environments where trust, accountability, and real-time responsiveness are critical, the "black box" nature of many AI models poses significant barriers to their adoption and operational effectiveness. This systematic literature review examines recent developments in Explainable Artificial Intelligence (XAI) within the cybersecurity domain, focusing on its role in enhancing transparency, trust, and human-AI collaboration. A structured search was conducted across six major academic databases and preprint repositories, yielding nine peer-reviewed studies that met rigorous inclusion criteria. These studies were analyzed across five quality dimensions: relevance, clarity of XAI methods, empirical grounding, human factors consideration, and deployment realism. Findings reveal that while technical innovations—such as SHAP, LIME, Grad-CAM, and lightweight edge-based models—offer substantial gains in model transparency, these advances often fail to translate into actionable insights for end-users due to limitations in cognitive usability and system integration. The review identifies a recurring gap between the theoretical promise of XAI and its practical implementation in real-world security infrastructures. Studies highlight issues such as user disengagement, underutilization of explanation tools, and inadequate alignment with operational workflows. Emerging directions emphasize the need for user-centered design, co-explainability frameworks, and interdisciplinary approaches that incorporate cognitive science and human-computer interaction. In conclusion, the future of XAI in cybersecurity hinges on its ability to go beyond algorithmic transparency and embed interpretability within the social, cognitive, and organizational contexts in which security professionals operate. Bridging these gaps will be essential for realizing the full potential of explainable AI systems as trustworthy and effective tools in modern cybersecurity operations.

Phasic Attractors for Flexible and Adaptive Working Memory in Spiking Recurrent Neural Networks

Abstract Working memory serves as a crucial building block in cognitive systems due to its fundamental role in information processing and decision-making. Acting as a temporary storage and manipulation mechanism, working memory enables individuals to actively hold and manipulate relevant information essential for ongoing tasks. This cognitive function is pivotal for various complex processes, including problem-solving, language comprehension, and decision-making. This paper introduces a novel working memory model designed as a Spiking Recurrent Neural Network of excitatory and inhibitory neurons. The proposed model incorporates biological mechanisms that allows attractors to be active phasically, thereby reducing the energy budget associated with maintaining attractor states. The core innovation of the model lies in its ability to leverage phasic attractors for state-dependent computation in a probabilistic manner. By modulating the activity of attractors via synaptic delays, the model demonstrates context-sensitive information processing. The results showcase the efficiency gains achieved by the proposed phasic attractor mechanism and highlight the model's capacity for flexible and adaptive information processing.

Comparision of digital and printed text reading process

Summary Reading is a cognitive activity that involves engaging with text and drawing on various information sources to achieve comprehension. To fully understand the nature of the reading process, it is essential to explore how the cognitive system functions. This study examines the effects of reading printed versus digital texts on the reading process and performance. Specifically, it compares reading time (speed) for both mediums by measuring response time and the speed of answering questions related to the deeper meaning of the text. Furthermore, the participants’ performance in answering these questions was assessed. The study also investigates whether factors such as age, gender, reading preference, and reading frequency (habit) influence reading speed, response time, and overall performance. The results indicate that participants reading printed texts performed better and were more successful in comprehending the material. Additionally, the findings reveal no significant differences based on age or gender, although reading preference and frequency did impact reading performance.

Clustering Analysis of Cognitive Profiles of Clinical Groups Using the CAS: An Examination of Japanese Clinical Populations

This study examined the distribution characteristics of the standard scores on the Japanese version of the Cognitive Assessment System (CAS)’s Planning, Attention, Simultaneous Processing, and Successive Processing (PASS) scale by clustering the scores using the k-means method, focusing on clinical groups. In Study 1, 140 clinical cases evaluated using the CAS at University A’s educational counseling service were analyzed. The k-means clustering method was applied based on the full-scale standard scores, PASS scale scores, score discrepancies, and subtest scaled scores. Study 2 applied the same clustering method to a clinical group of 91 cases with ADHD, ASD, or comorbid ADHD–ASD, excluding those with intellectual developmental disorders or other disorders. In Study 1, a group with lower full-scale standard scores indicating general intellectual development was identified. Study 2 identified a cluster of cases with ADHD, ASD, or comorbid ADHD–ASD that showed distinct discrepancies among the four standard scores. In addition, there were no significant differences in the diagnoses across clusters. The Japanese version of the CAS provides valid cognitive profile insights in clinical settings, which can aid in planning support interventions beyond clinical diagnosis.

A novel cognitive diagnosis system with attention residual mechanism and broad learning

A novel cognitive diagnosis system with attention residual mechanism and broad learning

AI in Neurology: Everything, Everywhere, All at Once Part 1: Principles and Practice.

Artificial intelligence (AI) is rapidly transforming healthcare, yet it often remains opaque to clinicians, scientists, and patients alike. This review, part 1 of a 3-part series, provides neurologists and neuroscientists with a foundational understanding of AI's key concepts, terminology, and applications. We begin by tracing AI's origins in mathematics, human logic, and brain-inspired neural networks to establish a context for its development. The review highlights AI's growing role in neurological diagnostics and treatment, emphasizing machine learning applications, such as computer vision, brain-machine interfaces, and precision care. By mapping the evolution of AI tools and linking them to neuroscience and human reasoning, we illustrate how AI is reshaping neurological practice and research. We end the review with an overview of model selection in AI and a case scenario illustrating how AI may drive precision neurological care. Part 1 sets the stage for part 2, which will focus on practical applications of AI in real-world scenarios where humans and AI collaborate as joint cognitive systems. Part 3 will examine AI's integration with extensive healthcare and neurology networks, innovative clinical trials, and massive datasets, expanding our vision of AI's global impact on neurology, healthcare systems, and society. ANN NEUROL 2025.

Sharpening the relationship between cognitive complexity and moral choice

Abstract Kant-based tradition suggests that complex moral reasoning improves moral choices, whereas intuitionists consider complex reasoning as an indicator of a desire to rationalize the immoral. The findings of this study support intuitionists, with one exception. In two studies conducted with a total of 195 subjects, two cognitive systems framing moral decisions came to the forefront. One is more complex than the other in terms of having multiple perspectives. It was observed that individuals who chose any of these systems and made relatively selfish decisions justified them in a more complex way. Among these selfish subjects, those who preferred the simpler system, supported their single perspectives with multiple reasons; whereas those who preferred the complex system, assumed useful perspectives to produce justifications that suited their interests. Complex reasoning contributes to the morality of the decision when it causes the decision maker to shift from the simple system to the complex one. However, according to our observations, such a shift during the decision period is not natural to human behavior, but it can still be triggered.

Are ambient smart environments sufficient for developing good habits?

Ambient smart environments (ASEs) have been proposed as technologically enriched spaces capable of supporting adaptive behaviour and healthier habits Aydin (Philosophy and Technology 32(2):321–338, 2019); Hipólito et al. (Synthese 199(11):4457–4481, 2023); White & Hipólito, (Cognitive Systems Research, 84, 101199, 2024) White & Miller, (Synthese, 204(2), 1–24, 2024) White et al. Synthese, 205(1), 1–18, 2024); Verbeek (NanoEthics 3(3):231–242, 2009). Proponents suggest that ASEs effectively disrupt maladaptive habitual behaviours by introducing optimal levels of uncertainty, implicitly guiding users towards healthier patterns of thinking and acting without requiring deliberate reflection. However, we argue that these interventionist approaches risk overlooking the conditions necessary for the consolidation of more sustainable and beneficial habits. In particular, we argue that such habits require the engagement of motivational states, effortful interaction, and opportunities for further reinforcement. To address this gap, we propose that ASE interventions must be complemented by more stable, deliberately structured, design-centred environments (DEs), which provide consistent affordances that explicitly invite reflective, intrinsically motivated interactions. We emphasise how these hybrid environments scaffold learning and feedback processes, and enable agents to progressively refine their engagement with affordances through experience and reflection. Finally, we discuss the broader philosophical and ethical implications, highlighting how hybrid environments mitigate the risks of techno-paternalism, and actively support adaptive autonomy, reflective control, and meaningful behavioural change.

Cognitive Robotics: Where Brain Science Meets Automation

Cognitive robotics is an interdisciplinary field that fuses artificial intelligence, neuroscience, robotics, and cognitive science to create intelligent robots capable of adaptive, goal-driven behavior. This paper presents a comprehensive analysis of cognitive robotics systems, emphasizing biologically inspired models, sensory integration, and learning mechanisms. By integrating brain-like processes, cognitive robots can learn from interaction, adapt to complex environments, and perform tasks traditionally reserved for humans. We present experimental results from a robotic testbed implementing cognitive architectures, evaluate performance improvements in dynamic tasks, and outline future implications for human-robot collaboration.