Complex and scalable IoT services increasingly rely on collaborative agreements across multi-cloud and edge-cloud environments. Such cooperation enables the dynamic composition of service workflows by leveraging both local and external resources, thus addressing the variability and surge in data traffic and computational demand typical of IoT scenarios. In this paper, we propose a trust-based model for dynamic IoT service composition that integrates system-level Quality of Service (QoS) metrics, operational costs, and user-generated feedback. The approach combines graph-based QoS prediction with trust-driven feedback aggregation and cost-aware heuristics, enabling more informed and adaptive service selection. A qualitative comparison with existing approaches highlights how the proposed model addresses key limitations in current solutions by jointly considering trust, QoS, and cost in a unified framework. Simulation-based experiments demonstrate that the proposed approach improves the reliability and responsiveness of service orchestration in heterogeneous IoT environments.
Metacognition, the capacity of an intelligent system to reason about its own processes, is widely regarded as absent from current artificial intelligence. This special issue examines how metacognitive principles can be modeled, measured, and applied to predict and improve AI behavior. Drawing on cognitive psychology, we frame metacognition through two related functions. Metacognitive monitoring gathers cues that reveal the state of a problem. Metacognitive control uses those cues to guide action. The six contributions gathered here leverage these ideas and span several branches of artificial intelligence examining metacognitive research relating to language models, uncertainty quantification, robotics, vision, federated learning, and reinforcement learning.
Frontier agentic systems powered by large language models (LLMs) exhibit human-like patterns of cognition. As these systems become deeply integrated across different domains, their cognitive engagement raises critical concerns for human society that remain insufficiently studied. To address this gap, we systematically analyze risks induced by expanding cognitive capabilities, following a three-level framework defined by their cognitive scope, from physical cognition to social cognition, and finally to self-referential cognition. We study their potential risks to human agency, autonomy, and control capability, corresponding to each cognitive level. We finally propose strategies to mitigate these risks and enhance the controllability of agentic AI systems, ensuring their long-term safe development.
As artificial intelligence (AI) continues to evolve, it has transitioned from a research-focused discipline to a widely adopted technology, enabling intelligent solutions across various sectors. In security, AI's role in strengthening organizational resilience has been studied for more than two decades. Although much attention has focused on AI's constructive applications, the increasing maturity and integration of AI have also exposed its darker potentials. This article explores two emerging AI-related threats and the interplay between them: AI as a target of attacks ("adversarial AI") and AI as a means to launch attacks on any target ("offensive AI")-potentially even on another AI. By cutting through the confusion and explaining these threats in plain terms, we introduce the complex and often misunderstood interplay between adversarial AI and offensive AI, offering a clear and accessible introduction to the challenges posed by these threats.
With artificial intelligence- (AI) and machine learning (ML)-based tools being discussed nowadays in academia, industry, and in the media, interesting questions also arise regarding their applications. The title of this article expresses that question. One reason for writing this forward-looking perspective/position article is because this author served in various editorial roles in certain publications, and knowing the challenges in finding reviewer support, this author believes that it will be highly beneficial if an AI/ML-based tool could be utilized for such review support. The scope of this article does not intend to provide demonstration of such a tool, since the development of such a tool will need significant collaboration between multiple entities involving academia, libraries, publishers, and industry. The article also discusses the huge energy consumption of AI/ML-based tools.
Spiking neural networks (SNNs) are gaining attention for biological plausibility and energy efficiency. Advances in neuromorphic systems-integrating hardware and software tools-accelerate SNN implementation. Yet, deploying SNNs on such platforms remains challenging due to model complexity and system heterogeneity, requiring flexible frameworks. Existing tools (e.g., PyNN and Brian2) show limited expressiveness for neuromorphic applications or poor cross-platform support. This article proposes SNNL, a flexible, domain-specific language for SNN development and deployment on neuromorphic hardware. SNNL decouples neuronal dynamics modeling from network topology specification: equation-based representations handle diverse neuron/synapse models, while hierarchical constructs define complex connectivity patterns. We present a Darwin3-targeted compiler with efficient code generation. Evaluations confirm that SNNL achieves precise neuronal dynamic descriptions and flexible network configurations. This work bridges algorithm-hardware gaps in neuromorphic computing by enhancing programmability. Experimental results have demonstrated the feasibility of SNNL in developing SNNs for neuromorphic systems.
Precise stability criteria are essential for vehicle handling control, but conventional methods based on tire adhesion limits or linear models often lack robustness across diverse scenarios. To address this issue, this article proposes a novel lateral stability criterion fusing phase plane analysis and radial basis function (RBF) neural networks. The approach begins with an analysis of the vehicle's stable state using the phase plane, followed by a division of the vehicle's stability region employing the diamond method to generate a phase plane stability region database. Subsequently, the proposed phase plane-RBF stability criterion is constructed by leveraging the RBF neural network for nonlinear fitting of the stability region data, which are further refined through multiple rounds of optimization. Compared to traditional tire force and linear single-track model criteria, the proposed criterion demonstrates superior accuracy in identifying extreme conditions and enhanced adaptability across operational scenarios.
Modern networked systems are becoming increasingly complex and demand high levels of reliability. Digital Twin Networks (DTNs) have been proposed to improve network management by creating virtual representations of physical networks. However, most existing DTNs lack formal semantics and mainly function as passive monitoring tools, which limits their ability to support automation and autonomous decision-making. This paper proposes a knowledge graph (KG)-based framework that enables DTNs to move beyond passive observation toward automated, autonomous, and interpretable operation. In the proposed framework, the KG provides a semantic foundation that integrates heterogeneous network data with domain knowledge in a consistent and machine-readable manner. This semantic foundation further supports reasoning, decision-making, and explainable human–AI interaction. We structure the framework into four phases, Construction, Maintenance, Prediction and Reasoning, and Decision Intelligence, covering the full pipeline from data integration to actionable insights. A proof-of-concept implementation based on real-world networking is presented to demonstrate feasibility.
Future artificial general intelligence (AGI)-native 6G networks will widen research from a narrow computer science perspective to a broader neuroscience as well as robotics perspective, given that key cognitive abilities of AGI such as lifelong learning (LL) via environmental coupling have been studied in both fields for decades. After highlighting key lessons from neuroscience and robotics with regard to more advanced brain models and LL capabilities, this paper focuses on active inference, an ideal methodology for advancing AGI, with its recurrent dark-room problem and how to resolve it by applying biomimetic psychoactives in the future 6G World Brain. We show that active inference points in the exactly opposite direction of today’s envisioned AGI-native 6G trajectory. The journey beyond 6G might not be outward via AI embodiment, but inward into what neuroscientists coined the brain’s dark energy, thereby resolving the dark-room problem and helping us penetrate ever deeper into cognitive space.
As AI becomes a ubiquitous and partially autonomous layer of everyday activity, it increases information volume, interaction tempo, delegated decision-making, and supervisory requirements. This paper introduces "AI overload," defined as a persistent mismatch between AI-enhanced demands and human and institutional capacity for attention, deliberation, validation, and accountability. We propose a multi-level taxonomy of AI overload across individual, organizational, and societal contexts, comprising seven types: cognitive, informational, interactional, coordination, control, normative, and affective. Exploiting research on cognitive load, automation bias, technostress, and algorithmic mediation, we show how increasing AI agency and human-AI co-adaptation produce new overload pressures. At the societal level, we link AI-driven recommendations and microtargeting with reduced cultural diversity, polarization, and increased susceptibility to manipulation under constrained attention. Finally, we highlight mitigation strategies focused on limited AI autonomy, maintaining control, AI literacy, information access, and complexity-aware modeling to preserve human agency and well-being.
Fairness in machine learning (ML)-based software has gained increasing attention as these systems are deployed in critical decision-making. Since ML software learns from labeled data, biased labels can substantially affect prediction outcomes, potentially resulting in discriminatory behavior. This study presents FairLabeler, a model-agnostic preprocessing approach that detects and massages biased labels in training datasets, thereby improving both individual-and group-level fairness while preserving predictive performance. We evaluate FairLabeler on nine real-world datasets using standard ML classifiers and a deep neural network. Experimental results show that FairLabeler surpasses the state-of-the-art bias mitigation method, as assessed with the fairness–performance tradeoff tool Fairea, achieving fairness with improved performance in 11% more cases. To facilitate replication and further research, we publicly release the code and datasets at https://github.com/fairlabeler/anonymous.
Building intelligent agents capable for autonomously perceiving, reasoning, and acting to achieve goals has been a central pursuit of artificial intelligence (AI) since its inception. For decades, the notion of agency was dominated by symbolic architectures that represent information as formal knowledge and use deliberative reasoning to derive rational actions, offering reliability but at the cost of brittleness and limited generalizability. The recent advancements in large language models (LLMs) and their integration into tool-using, environment-interacting "agentic" systems have reignited interest in AI agents. However, while LLM-based agents provide the flexibility that symbolic systems lacked, they introduce new challenges in reliability and control. We posit that the future of AI agents lies not in indefinitely scaling the model size, but in synthesizing the methods and theories developed by the autonomous agents and multiagent systems community with modern neural architectures to create neuro-symbolic agents capable of trustworthy autonomy.
Chatbots, powered by large language models (LLMs), were initially part of the effort aimed at developing systems that could provide human-like conversations. However, LLMs continue to add capabilities to perform various additional activities. This article introduces the notion of "LLM commerce" and discusses the evolving nature of LLM capabilities. Search and knowledge management capabilities are turning LLMs into potential e-commerce machines. Google Trends and Wikipedia page views are used empirically to analyze emerging implications of that evolution.
Autonomous aerial vehicle (AAV) object detection is essential for autonomous AAV operations. However, current AAV object-detection algorithms ignore leveraging scene features to assist in object detection, which hinders the improvement of the algorithm's generalization across multiple scenarios. To address this, we propose the AAV scene-driven object-detection network. First, we design the global characteristics modeling module to extract scene features and the region-aware attention module to capture the global integrity features of the scene. The scene classification task is leveraged to assist in optimizing the module's parameters. Subsequently, we design the scene feature guidance module to guide the image feature extraction and object-detection process using scene features, encouraging the network to effectively mine scene-related clues and query the targets of the current scene. Finally, we construct a multiscene AAV object-detection dataset, and experiments show that our algorithm effectively improves performance in multi-scenario AAV object-detection tasks.
AI-enabled systems deployed in dynamic, open-world, safety- and mission-critical environments often lack mechanisms for reliable self-monitoring and self-regulation under shift and perturbation. Artificial Metacognition (AM) addresses this by enabling systems to assess and adapt their own inference and learning processes. We show that relevance structures, which are order-theoretic substrates for comparing generalization and specialization and for quantifying under-coverage versus overreach, are structurally unavoidable for computable self-regulation. In turn, relevance structures provide an operational control theory for metacognitive tasks. We introduce exploratory, parsimonious, and balanced relevance scores and embed them into version-space algebra to regulate hypothesis search and inductive strategy selection. The resulting framework yields competence-aware learning systems that can both correct and explain their behavior by explicitly balancing exploration and parsimony. We illustrate relevance structures and their application with a medical visual question answering study on the SLAKE dataset.
With AI (artificial intelligence) and ML (machine learning) based tools being discussed nowadays in the academia, industry, and in the media, interesting questions also arise regarding their applications. The title of this article expresses that question. One reason for writing this forward looking perspective/position paper is because this author served in various editorial roles in certain publications, and knowing the challenges in finding reviewer support, this author believes that it will be highly beneficial if an AI/ML based tool could be utilized for such review support. The scope of this article does not intend to provide demonstration of such a tool, since the development of such a tool will need significant collaboration between multiple entities, involving the academia, libraries, publishers, and industry. The paper also discusses about the huge energy consumption in AI /ML based tools.
We introduce context conditioning, a phenomenon analogous to Pavlovian learning, in which large language models (LLMs) display heightened sensitivity to small amounts of novel contextual signals. This conditioning is double-edged. Carefully curated contexts can quickly steer models toward trustworthy, inclusive behavior, while minor malicious or biased signals can provoke unsafe, toxic, or privacy-compromising responses. We reveal this double-edged behavior with two studies that collectively highlight the underlying associative amplification mechanism through which novel or low-frequency contextual cues exert outsized influence on model attention and response distributions. Trust in context-based artificial intelligence (AI) thus depends not only on model design but also on how context governs behavior at inference time. We outline five research directions for building trustworthy context-based LLM systems and argue that the future of responsible AI lies not only in safer models but in safer contexts, meaning systems that understand, audit, and adapt to the stimuli that condition them.
We introduce the Microsoft-Northwestern-WITNESS (MNW) deepfake detection benchmark, a dataset designed to evaluate and improve artificial intelligence (AI)-generated content detection algorithms. The dataset contains more than 50,000 artifacts (images, videos, and audio files) generated by us. It also includes real-world examples of AI-manipulated or suspicious media encountered by journalists and human rights defenders globally, annotated by experts to reflect practical, high-stakes detection scenarios. The MNW dataset will be periodically updated to cover emerging generators and includes adversarial examples created with state-of-the-art attacks. This is a collaborative effort, and we encourage generative AI model developers to help maintain the dataset’s currency. This dataset is intended solely for evaluation purposes and cannot be used for training or commercial purposes. We recommend that entities purchasing detection solutions avoid using our dataset to evaluate commercial tools. Our goal is to establish high standards for developers and enhance the reliability of detection systems.
To commemorate the 70th anniversary of artificial intelligence (AI), IEEE Intelligent Systems has identified the 10 most impactful articles in AI published since 2000, marking the field’s evolution from its 1956 origins into a foundational pillar of modern science. The selection process combined rigorous quantitative indicators with the informed judgment of a panel of distinguished experts. The resulting list reflects a broad consensus on the milestones that have redefined AI, spanning the diverse and multifaceted landscape of the discipline. Together, these landmark articles laid the foundations for modern AI and continue to influence its evolution.
