Cognitive Neurosymbolic Artificial Intelligence for Complex Decision-Making: Integrating Foundation Models, Cognitive Architectures, and Knowledge

Building a human-like integrative artificial cognitive system, that is, an artificial general intelligence (AGI), is the holy grail of the artificial intelligence (AI) field. Furthermore, a computational model that enables an artificial system to achieve cognitive development will be an excellent reference for brain and cognitive science. This paper describes an approach to develop a cognitive architecture by integrating elemental cognitive modules to enable the training of the modules as a whole. This approach is based on two ideas: (1) brain-inspired AI, learning human brain architecture to build human-level intelligence, and (2) a probabilistic generative model (PGM)-based cognitive architecture to develop a cognitive system for developmental robots by integrating PGMs. The proposed development framework is called a whole brain PGM (WB-PGM), which differs fundamentally from existing cognitive architectures in that it can learn continuously through a system based on sensory-motor information.In this paper, we describe the rationale for WB-PGM, the current status of PGM-based elemental cognitive modules, their relationship with the human brain, the approach to the integration of the cognitive modules, and future challenges. Our findings can serve as a reference for brain studies. As PGMs describe explicit informational relationships between variables, WB-PGM provides interpretable guidance from computational sciences to brain science. By providing such information, researchers in neuroscience can provide feedback to researchers in AI and robotics on what the current models lack with reference to the brain. Further, it can facilitate collaboration among researchers in neuro-cognitive sciences as well as AI and robotics.

The ADAPT project is a collaboration of researchers in robotics, linguistics and artificial intelligence at three universities to create a cognitive architecture specifically designed to be embodied in a mobile robot. There are major respects in which existing cognitive architectures are inadequate for robot cognition. In particular, they lack support for true concurrency and for active perception. ADAPT addresses these deficiencies by modeling the world as a network of concurrent schemas, and modeling perception as problem solving. Schemas are represented using the RS (Robot Schemas) language, and are activated by spreading activation. RS provides a powerful language for distributed control of concurrent processes. Also, The formal semantics of RS provides the basis for the semantics of ADAPT's use of natural language. We have implemented the RS language in Soar, a mature cognitive architecture originally developed at CMU and used at a number of universities and companies. Soar's subgoaling and learning capabilities enable ADAPT to manage the complexity of its environment and to learn new schemas from experience. We describe the issues faced in developing an embodied cognitive architecture, and our implementation choices.

Cognitive architectures aim to provide blueprints for achieving artificial general intelligence AGI. Cognitive models that can provide crucial clues to the working of human mind are typically built on top of cognitive architectures. This paper reviews contemporary cognitive architectures including Soar, ACT-R, CLARION, OpenCog Prime and ICARUS in terms of architecture and components, knowledge and representation, learning and adaptability, and memory models. It then evaluates these architectures based on certain evaluation criteria. In addition, this paper reviews recent research focus in these architectures and their usage in real-world applications. Finally, this paper discusses the trends and challenges ahead in building effective cognitive architectures.

Cognitive architectures evolved at the intersection of artificial intelligence (AI) and cognitive science and derive many concepts and methods from both disciplines. This chapter provides a brief overview of the notable events in AI and cognitive science from their inception in the 1950s and onward that helps better understand the roots of cognitive architectures. History of AI so far alternated between “summers” full of excitement and flush with funding and “winters” that followed critical assessment of the results. In parallel, two computational paradigms, symbolism and connectionism, competed for dominance. Cognitive science initially followed some of the trends in AI but started to diverge in the 1990s when new approaches to cognition as embodied and situated emerged. Interaction between AI and cognitive science throughout decades resulted in many algorithms and theories of cognition. In this chapter we list the most important of these and trace their influences on cognitive architectures.

The purpose—explain identify the features of General Artificial Intelligence (AGI), from the standpoint of scientific rationality stages (classic, non-classical, post-non-classical), and show its difference from traditional Artificial Intelligence (AI). The latter is currently implemented mainly with digital computers and implements the functions of recognition, forecasting and preparation of answers to simple questions. New conditions force us to consider AGI from the standpoint of functionalism, as a man–machine system, purposefully functioning in a self-developing poly-subject (reflexive-active) environment.Design/Methodology/Approach—AI development paradigm should consider aspects of its immersion in the social and humanitarian environment and the innovative atmosphere. The new paradigm of AI development should reflect the unformalized cognitive dynamics of AI models and support the self-development of AI systems under pressure from the external environment. The methodology of creating AGI is based on the ideas of the subject-oriented and ontological approach, functionalism, the phenomenology of subjective reality, the convergent cognitive architectures, as well as the methods of creating a self-developing poly-subject (reflexive-active) environment. AGI becomes a hybrid, purposefully integrating the capabilities of a machine and a person.Findings—General properties, value-semantic and intentional-volitional operational structures of the phenomenon of subjective reality do not lean itself on direct formalized and algorithmic representation in discrete computer systems of von Neumann architecture. The study of consciousness in the context of subjective reality made it possible to formulate the main systemic, structural, functional, and operational characteristics of human cognitive activity, which allows a new approach to the modelling of cognitive architectures that meet the tasks of building AGI. The characteristics of subjective reality cannot be fully represented in the paradigm of physicalism; that is, it cannot be represented only with the help of physical devices. The chapter proposes a non-reductionist way of taking this characteristic into account by considering the problem of consciousness in an ontological and epistemological context, which allows representing the processes of consciousness and cognitive activity of a person and a group of people indirectly and inversely.Originality/Value—State-of-the-art cognitive architectures and traditional AI approaches practically ignore solving the problems of AGI. They are more focused on the formalized construction of a thinking model, identifying physical blocks and processes of mental activity. At the same time, for AGI, ontological, subjective and hybrid reality issues are of the most importance, especially in explanations of the activity of consciousness, unconsciousness, and causeless processes, which can act purposefully in conditions of goal uncertainty. AGI must help to describe the phenomena of subjective reality, which causes physical changes, explain the ability of goal setting, free will, the ability of self-management by the physical actions of an individual in a team, etc.Research/Practical/Social/Environment implications—the chapter give rise to a new type of control, which differs from the traditional control in digital reality. The chapter’s results made it possible to uniquely find the optimal measure of centralization and autonomy of control loops that can ensure the preservation and strengthening of the integrity of a complex poly-subject system functioning in a reflexive-active environment, the interpretation of which does not fit into the narrow framework of digital and algorithmic reality, and traditional AI.Research limitations—the AGI approach based on exceptional methods of constructing subjective reality also has its limitations. For example, the approach we propose to explain the connection between the human brain, consciousness, thought processes and environment does not yet allow us to explain the information and cognitive processes generated by the effect of subjective reality nonlocality, which arises, e.g., at the atomic level of the human brain and should be considered when studying cognitive processes.KeywordsArtificial general intelligenceScientific rationalityCognitive semanticsSelf-developing poly-subject (reflexive-active) environmentSubjective reality

Biochemical origin of the schizophrenia and of the behavioral pathologies of schizoid type

This paper traces the evolutionary trajectory of artificial intelligence from narrow artificial intelligence (ANI) through artificial general intelligence (AGI) to potential artificial superintelligence (ASI), critically examining the technological singularity as a transformative inflection point in technological and human development. The analysis covers key technical milestones required for this progression, including advances in machine learning architectures, knowledge representation, and cognitive architectures. Drawing from recent developments in neuroscience and deep learning, the paper proposes a framework for understanding the qualitative differences between these three stages of artificial intelligence development. The paper addresses critical questions about intelligence, consciousness and the challenges of maintaining human agency in a post-singularity world. Novel perspectives on safety mechanisms and technical guidelines necessary for managing the transition between these developmental stages are presented. The analysis acknowledges the uncertainty surrounding the path to AGI and ASI, highlighting the complex and unpredictable nature of advanced artificial intelligence development. The paper concludes by exploring potential trajectories of recursive self-improvement in artificial systems and their theoretical implications for the future of machine cognition.

Schema formalism for the common model of cognition

Origins and evolution of enactive cognitive science: Toward an enactive cognitive architecture

The rapid evolution of Artificial Intelligence (AI) has significantly influenced business decision-making across industries. This study conducts a systematic literature review to examine how elements of human consciousness are being integrated into AI frameworks to enhance decision-making quality and ethical awareness, with a specific focus on the Fast-Moving Consumer Goods (FMCG) sector. Using the PRISMA methodology, 40 peer-reviewed studies from domains including AI, cognitive science, and management were analyzed to synthesize theoretical and empirical insights. The review identifies four key thematic areas: (1) foundational theories and consciousness lenses in AI, (2) decision-making augmented by consciousness-inspired attributes, (3) cognitive architectures that emulate self-reflection and contextual reasoning, and (4) current AI adoption practices in FMCG. Findings reveal that, although true machine consciousness remains unrealized, integrating conscious-like attributes – such as introspection, situational awareness, and ethical deliberation – enhances the transparency and reliability of AI-assisted decisions. The paper concludes with a structured research agenda outlining future directions for developing consciousness-oriented AI systems for managerial decision-making.

This paper introduces CAAI, a novel cognitive architecture for artificial intelligence in cyber-physical production systems. The goal of the architecture is to reduce the implementation effort for the usage of artificial intelligence algorithms. The core of the CAAI is a cognitive module that processes the user’s declarative goals, selects suitable models and algorithms, and creates a configuration for the execution of a processing pipeline on a big data platform. Constant observation and evaluation against performance criteria assess the performance of pipelines for many and different use cases. Based on these evaluations, the pipelines are automatically adapted if necessary. The modular design with well-defined interfaces enables the reusability and extensibility of pipeline components. A big data platform implements this modular design supported by technologies such as Docker, Kubernetes, and Kafka for virtualization and orchestration of the individual components and their communication. The implementation of the architecture is evaluated using a real-world use case. The prototypic implementation is accessible on GitHub and contains a demonstration.

The key abilities of the cognitive radar system include learning new knowledge about its surroundings, critical decision-making, and adaptation to a constantly changing environment. Among others, cognitive radar aims to achieve optimal radar resource management and maximize target detection performance. Radar systems often have to operate in strongly interfered environments. This work investigates the novel research direction of cognitive radar system design based on artificial intelligence (AI) cognitive architectures. Namely, the integration of a coherent radar model with Soar cognitive architecture has been developed. This work describes in detail the sensorimotor subsystem, which represents radar signal processing algorithms and waveform design. The AI subsystem comprising of the long-term and short-term memories of the Soar architecture was also characterized. Cognitive memories are designed to represent radar environment state and solve signal-to-noise-ratio (SNR) maximization problem. Using computer simulations it has been shown that the proposed system improved target detection performance in comparison to traditional radar system.

Artificial General Intelligence (AGI) represents a significant leap in the field of artificial intelligence, defined by its ability to perform any intellectual task that a human can. Unlike narrow AI, which is task-specific, AGI is characterized by versatility, adaptability, autonomy, and reasoning capabilities. This comprehensive review explores the defining features of AGI, including its cognitive flexibility and capacity for autonomous decision-making and problem-solving. The current state of AGI research is examined, highlighting key areas such as cognitive architectures, neuroscience-inspired AI, and machine learning and deep learning advancements. The review also addresses the technical challenges and ethical considerations associated with AGI development, including potential impacts on employment, privacy, and security, as well as the necessity for robust safety and control measures. Looking ahead, the potential benefits of AGI in various domains, such as healthcare, climate change, education, and economic growth, are discussed. Finally, the importance of establishing ethical frameworks and governance structures to guide AGI development and usage is emphasized. By addressing these challenges and fostering collaboration among scientists, policymakers, and society, AGI can be developed and deployed to benefit humanity as a whole.

One area of intense focus in Artificial Intelligence (AI) research is to implement intelligent agents and machines that can think, reason, and solve problems with similar if not better proficiency than humans. The advancements in our understanding of intelligence and its governing principles have lead researchers to explore vastly different and passionately debated approaches to building intelligent systems. However, in recent years a consensus appears to be emerging, as observed by the 2013 AAAI Fall Symposium on Integrated Cognition, where an initial proposal is for a standard model of human-like minds. This model is largely based on current state-of-the-art cognitive architectures, such as ACT-R, Sigma and Soar. While we do not disagree with the proposed model, we believe more computer science based theories for modeling human-like minds should be considered for modeling memory and recognition functions. In this paper, we present a survey of the current literature that supports our position.

From dancing robots to action aesthetics: Re-examining mirror system activity as a function of the observer's experience