A taxonomy-driven survey of the major cognitive capability gaps that continue to constrain the development of Cognitive AI and outlines a conceptual Adaptive Cognitive Intelligence Architecture (ACIA), providing a foundation for future progress toward Cognitive AI and, ultimately, Artificial General Intelligence (AGI).
Abstract
Cognitive AI seeks to move beyond language generation and autonomous task execution toward systems capable of sustained reasoning, adaptive behavior, persistent memory, and self-regulation. While generative and agentic AI have demonstrated impressive capabilities across a wide range of tasks, many fundamental cognitive functions remain fragmented or weakly developed, limiting reliable operation over extended time horizons. This paper presents a taxonomy-driven survey of the major cognitive capability gaps that continue to constrain the development of Cognitive AI. The literature is organized around five dimensions: persistent state modeling, goal-directed autonomy, self-monitoring and control, environment interaction, and learning and adaptation. For each dimension, we review recent advances, identify recurring limitations, and discuss open research challenges. Building on these insights, we outline a conceptual Adaptive Cognitive Intelligence Architecture (ACIA) and examine emerging directions in cognition-centric evaluation. The proposed taxonomy provides a unified framework for organizing existing research, identifying unresolved challenges, and guiding the design of future cognitively capable systems. Together, the taxonomy, architectural perspective, and evaluation framework offer a roadmap for advancing AI systems that exhibit more reliable long-term reasoning, adaptive decision-making, and continual learning. The survey highlights key research opportunities toward more adaptive, reliable, and cognitively capable AI systems, providing a foundation for future progress toward Cognitive AI and, ultimately, Artificial General Intelligence (AGI).
This work synthesizes perspectives from philosophy, cognitive science, and AI to define agency, outline its key properties, and situate it in relation to existing paradigms such as reinforcement learning, symbolic reasoning, Belief–Desire–Intention (BDI) architectures, and embodied cognition.
Pera describes a persistent agent organized around perception and control components that continually perceive service-relevant signals from episodic task executions, internal context, and changes in the surrounding environment, and use these signals to construct lifecycle tasks.
Shihan Dou, Haoxiang Jia, Shichun Liu et al.· 0 citations
Embodied Artificial Intelligence (Embodied AI) has emerged as a promising paradigm for developing more general and adaptive intelligent systems, emphasizing that intelligence emerges from continuous interaction among perception, cognition, and action in real-world environments. Recent advances increasingly integrate large language models and multimodal learning into embodied agents; however, most existing approaches remain correlation-driven, relying on implicit objectives, task-specific rewards, or prompt-level instructions. As a consequence, intent is rarely represented explicitly, limiting causal coherence, long-horizon consistency, and robust value alignment in open-world settings. In this Review, we synthesize recent progress in Embodied AI and articulate Intent-Driven Embodied Artificial Intelligence (IDEAI) as a system-level organizing framework in which intent functions as an explicit, revisable, and verifiable mediating construct between human goals, environmental constraints, and agent behavior. Building on this synthesis, we propose a four-layer conceptual organization-semantic grounding, concept generation and learning, intent modeling, and value alignment-that clarifies how explicit intent mediates perception, cognition, and action in embodied systems. We analyze how existing techniques address recurring failure modes along the intent-to-execution pipeline and highlight the limitations that arise when intent remains implicit. By making intent explicit, revisable, and value-constrained where such structure is needed, IDEAI supports interpretable decision-making, adaptive task decomposition, and value-consistent behavior in open-ended, human-interactive, and safety-critical embodied domains, providing a unifying perspective for advancing Embodied AI toward robust, socially deployable intelligent systems.
Nanning Zheng· National Science Review· 0 citations
This survey synthesizes recent developments in self-improving agent architectures from the 2020 to 2025 period, focusing on how generative AI and cognitive architectures can be integrated for practical applications.
AbdElKader Seif El Islem Rahmani, Yasser Yahiaoui, Abdelghani Bouziane· SN Computer Science· 0 citations
The increasing capabilities of AI in automation and integration within existing methodology present a new opportunity for disciplines to address long-standing theoretical challenges. This is evident in archaeology, where persistent theoretical debates and interpretive variability present prime opportunities for applying AI-based techniques to address uncertainty. Using the phenomenology of agent-based modeling (ABM), we propose the incorporation of agentic AI methods that facilitate the automation and minimal human intervention in the application of social behavior, where perceptions, values, interaction, and cognition evolve within applied systems. We argue this represents the next challenge and stage of AI in understanding the past, one where applications can move beyond the case-specific and narrow technical results that generative and discriminative AI have been mostly applied toward. For AI to better understand potential pasts, new emergent behavioral outcomes that are better able to address long-standing theoretical debates are needed. Practical applications related to a variety of sub-areas within archeology, including
chaîne opératoire
, lived experience, logistics, governance, and other applications, are just some examples where closer ABM–AI integration using agentic techniques could better address.
M. Altaweel, Maurizio Forte· AI & SOCIETY· 0 citations
Agentic AI systems that reason, plan, and act on complex goals have advanced rapidly across software engineering, scientific discovery, drug development, healthcare, finance, and social simulation. Across these domains a single failure pattern recurs: current systems can execute tasks competently but often struggle to determine when to act, when to pause, when to change strategy, and when to involve a human. Existing reviews catalog agentic architectures, taxonomies, and limitations, but none specify what capabilities these systems must acquire to support dynamic human-AI collaboration. We address that gap. We define collaborative AI as a class of systems that combine generative exploration with autonomous action and calibrate between them based on context, uncertainty, and task demands. We identify four required capabilities: metacognition, contextual mode-switching, uncertainty-aware action, and adaptive human collaboration. We relate these capabilities to established multi-agent systems foundations, including belief-desire-intention architectures, adjustable autonomy, mixed-initiative interaction, and decentralized decision-theoretic control, while specifying the distinct challenges that LLM-based agents introduce. Across the six domains reviewed here, these gaps appear repeatedly and are not solved by current architectures, which positions collaborative AI as a concrete near-term research objective.
Nalan Karunanayake, Savindu Nanayakkara, Kasun Gayashan Hettihewa et al.· International Journal of Net...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
