Active Inference Research Articles

Объединение теорий прогностической обработ­ки и когнитивного диссонанса в контексте проблемы рационального выбора

A significant problem in constructing the theory of rational choice (and rationality in gen­eral) at the moment is the question of explaining irrationality, or rather the fact that an agent can act irrationally for an external observer, and the models used for it may not work. However, the agent himself “inside”, if you find yourself in his “logic”, acts clearly rationally upon verification. It’s just that his “rationality” and the “rationality” of an ex­ternal observer turn out to be different, hence they do not coincide, and hence this leads to a discrepancy in the language of their description of each other. To resolve this prob­lem and find a theoretical basis for the connection between the physiological states of or­ganisms and testable the theoretical constructs, and thus obtaining the opportunity, with­out changing the optics, to work with all agents, regardless of the types of rationality they have, psychological approaches have increasingly become involved. But for them, the basic problem is always “rootedness” in the “material substrate”. This work largely suggests and shows which mechanism can allow this problem to be solved. The paper draws atten­tion to the fact that the combination of the theory of predictive processing and the theory of cognitive dissonance turns out to be extremely promising for this task. This is due to the fact that both of these theories are engaged in reducing surprise or inconsistency (prediction errors and dissonance), with the former providing the latter with valuable and informative context within the broader perception – prediction – action framework (known as active inference) and also helps us to separate dissonant cognitive elements from dissonant psychological states, while considering them as part of the same process of perception and cognition. All this allows us to set a fundamental theoretical basis for the connection between the physiological states of organisms and testable theoretical con­structs about the degree of rationality of a particular agent, and, thus, gain access to the internal information processes in the brain for study.

Generating meaning: active inference and the scope and limits of passive AI.

Prominent accounts of sentient behavior depict brains as generative models of organismic interaction with the world, evincing intriguing similarities with current advances in generative artificial intelligence (AI). However, because they contend with the control of purposive, life-sustaining sensorimotor interactions, the generative models of living organisms are inextricably anchored to the body and world. Unlike the passive models learned by generative AI systems, they must capture and control the sensory consequences of action. This allows embodied agents to intervene upon their worlds in ways that constantly put their best models to the test, thus providing a solid bedrock that is - we argue - essential to the development of genuine understanding. We review the resulting implications and consider future directions for generative AI.

Representation and Regulation in Emotional Theory

The case of pain asymbolia is a case study that provides evidence of the mechanisms underlying the relationship between bodily experience, affective experience, and self-awareness. On one account pain asymbolia is the result of an affective deficit. Sensory signals of bodily damage are not associated with characteristic negative affect. Cochrane endorses this account as part of his version of a “conceptual act” theory of affective experience. In contrast, I propose an active inference account of affect in general and pain asymbolia in particular. In the active inference framework the self is inferred as the endogenous cause of bodily and affective experience in the process of organismic regulation. This preserves Cochranes ambition to ground affect in bodily regulation but avoids the problem for affective deficit accounts of asymbolia that cannot do justice to the neural correlates.

Agentive Cognitive Construction Grammar: a predictive semiotic theory of mind and language

Abstract This paper introduces a novel perspective on Agentive Cognitive Construction Grammar (AgCCxG) by examining the intricate interplay between mind and language through the lens of both Active Inference and Peircean semiotics. AgCCxG emphasizes the impact of intention and purpose on linguistic choices as a cognitive imperative to balance the symbolic Self (Intelligent Agent) with the dynamics of the environment. Among other things, the paper posits that linguistic constructions, particularly Constructional Attachment Patterns (CAPs), like argument structure constructions, embody experienced interactions with the world through reenactment routines via the integration of multisensory channels. Unlike traditional usage-based approaches (e.g., construction grammars), AgCCxG embraces a robust theory of signs that reveals human representation as a continuous process of semiotic hybridization for the creative prediction of uncertain scenarios. Importantly, the paper challenges the notion of the mind as a unified, rational, uncertainty-reducing machine by asserting that physical processes governing open biological systems profoundly influence the linguistic sign system. Intelligent agents’ adaptability in expressing incongruous realities thus highlights the role of semiotic hybridization in preserving an agent’s autonomy and semiotic boundary.

I like therefore I can, and I can therefore I like: the role of self-efficacy and affect in active inference of allostasis.

Active inference (AIF) is a theory of the behavior of information-processing open dynamic systems. It describes them as generative models (GM) generating inferences on the causes of sensory input they receive from their environment. Based on these inferences, GMs generate predictions about sensory input. The discrepancy between a prediction and the actual input results in prediction error. GMs then execute action policies predicted to minimize the prediction error. The free-energy principle provides a rationale for AIF by stipulating that information-processing open systems must constantly minimize their free energy (through suppressing the cumulative prediction error) to avoid decay. The theory of homeostasis and allostasis has a similar logic. Homeostatic set points are expectations of living organisms. Discrepancies between set points and actual states generate stress. For optimal functioning, organisms avoid stress by preserving homeostasis. Theories of AIF and homeostasis have recently converged, with AIF providing a formal account for homeo- and allostasis. In this paper, we present bacterial chemotaxis as molecular AIF, where mutual constraints by extero- and interoception play an essential role in controlling bacterial behavior supporting homeostasis. Extending this insight to the brain, we propose a conceptual model of the brain homeostatic GM, in which we suggest partition of the brain GM into cognitive and physiological homeostatic GMs. We outline their mutual regulation as well as their integration based on the free-energy principle. From this analysis, affect and self-efficacy emerge as the main regulators of the cognitive homeostatic GM. We suggest fatigue and depression as target neurocognitive phenomena for studying the neural mechanisms of such regulation.

Bayesian brain computing and the free-energy principle: an interview with Karl Friston.

The free-energy principle entails the Bayesian brain hypothesis that can be implemented by many schemes considered in this field. The combination of multimodal brain imaging and free-energy minimization has shown promise in unraveling complex brain dynamics and understanding the interactions among distinct brain regions. The Bayesian mechanics of brain computing gives a unique route to understanding authentic (neuromimetic) intelligence and, more importantly, points towards the development of brain-inspired intelligence. NSR spoke to a leading theoretical neuroscientist and authority on brain imaging-Karl Friston, the inventor of statistical parametric mapping, voxel-based morphometry and dynamic causal modeling. Friston is also known for his contributions to theoretical biology in the form of the free-energy principle and applications such as active inference. Friston is currently the Scientific Director of the Wellcome Trust Centre for Neuroimaging, Professor of Neuroscience at Queen Square Institute of Neurology, University College London and Honorary Consultant at The National Hospital for Neurology and Neurosurgery, UK.

Robustness evaluations of pathway activity inference methods on gene expression data

BackgroundWith the exponential growth of high-throughput technologies, multiple pathway analysis methods have been proposed to estimate pathway activities from gene expression profiles. These pathway activity inference methods can be divided into two main categories: non-Topology-Based (non-TB) and Pathway Topology-Based (PTB) methods. Although some review and survey articles discussed the topic from different aspects, there is a lack of systematic assessment and comparisons on the robustness of these approaches.ResultsThus, this study presents comprehensive robustness evaluations of seven widely used pathway activity inference methods using six cancer datasets based on two assessments. The first assessment seeks to investigate the robustness of pathway activity in pathway activity inference methods, while the second assessment aims to assess the robustness of risk-active pathways and genes predicted by these methods. The mean reproducibility power and total number of identified informative pathways and genes were evaluated. Based on the first assessment, the mean reproducibility power of pathway activity inference methods generally decreased as the number of pathway selections increased. Entropy-based Directed Random Walk (e-DRW) distinctly outperformed other methods in exhibiting the greatest reproducibility power across all cancer datasets. On the other hand, the second assessment shows that no methods provide satisfactory results across datasets.ConclusionHowever, PTB methods generally appear to perform better in producing greater reproducibility power and identifying potential cancer markers compared to non-TB methods.

A Systems Biology Approach to Understand the Racial Disparities in Colorectal Cancer.

The purpose of this work is to investigate the racial disparities in colorectal cancer between Black/AA and White patient cohorts using a systems biology and bioinformatic approach. Our study investigates the underlying biology of each patient cohort. Concretely, the findings of this study include disparity-associated genes and pathways, which provide a tangible starting point to guide precision medicine approaches tailored specifically for colorectal cancer racial disparities.

Making proteomics accessible: RokaiXplorer for interactive analysis of phospho-proteomic data.

We present RokaiXplorer, an intuitive web tool designed to address the scarcity of user-friendly solutions for proteomics and phospho-proteomics data analysis and visualization. RokaiXplorer streamlines data processing, analysis, and visualization through an interactive online interface, making it accessible to researchers without specialized training in proteomics or data science. With its comprehensive suite of modules, RokaiXplorer facilitates phospho-proteomic analysis at the level of phosphosites, proteins, kinases, biological processes, and pathways. The tool offers functionalities such as data normalization, statistical testing, activity inference, pathway enrichment, subgroup analysis, automated report generation, and multiple visualizations, including volcano plots, bar plots, heat maps, and network views. As a unique feature, RokaiXplorer allows researchers to effortlessly deploy their own data browsers, enabling interactive sharing of research data and findings. Overall, RokaiXplorer fills an important gap in phospho-proteomic data analysis by providing the ability to comprehensively analyze data at multiple levels within a single application. Access RokaiXplorer at: http://explorer.rokai.io.

The police hunch: the Bayesian brain, active inference, and the free energy principle in action.

In the realm of law enforcement, the "police hunch" has long been a mysterious but crucial aspect of decision-making. Drawing on the developing framework of Active Inference from cognitive science, this theoretical article examines the genesis, mechanics, and implications of the police hunch. It argues that hunches - often vital in high-stakes situations - should not be described as mere intuitions, but as intricate products of our mind's generative models. These models, shaped by observations of the social world and assimilated and enacted through active inference, seek to reduce surprise and make hunches an indispensable tool for officers, in exactly the same way that hypotheses are indispensable for scientists. However, the predictive validity of hunches is influenced by a range of factors, including experience and bias, thus warranting critical examination of their reliability. This article not only explores the formation of police hunches but also provides practical insights for officers and researchers on how to harness the power of active inference to fully understand policing decisions and subsequently explore new avenues for future research.

Designing ecosystems of intelligence from first principles

This white paper lays out a vision of research and development in the field of artificial intelligence for the next decade (and beyond). Its denouement is a cyber-physical ecosystem of natural and synthetic sense-making, in which humans are integral participants—what we call “shared intelligence.” This vision is premised on active inference, a formulation of adaptive behavior that can be read as a physics of intelligence, and which inherits from the physics of self-organization. In this context, we understand intelligence as the capacity to accumulate evidence for a generative model of one’s sensed world—also known as self-evidencing. Formally, this corresponds to maximizing (Bayesian) model evidence, via belief updating over several scales, that is, inference, learning, and model selection. Operationally, this self-evidencing can be realized via (variational) message passing or belief propagation on a factor graph. Crucially, active inference foregrounds an existential imperative of intelligent systems; namely, curiosity or the resolution of uncertainty. This same imperative underwrites belief sharing in ensembles of agents, in which certain aspects (i.e., factors) of each agent’s generative world model provide a common ground or frame of reference. Active inference plays a foundational role in this ecology of belief sharing—leading to a formal account of collective intelligence that rests on shared narratives and goals. We also consider the kinds of communication protocols that must be developed to enable such an ecosystem of intelligences and motivate the development of a shared hyper-spatial modeling language and transaction protocol, as a first—and key—step towards such an ecology.

Blind CT Image Quality Assessment Using DDPM-Derived Content and Transformer-Based Evaluator.

Lowering radiation dose per view and utilizing sparse views per scan are two common CT scan modes, albeit often leading to distorted images characterized by noise and streak artifacts. Blind image quality assessment (BIQA) strives to evaluate perceptual quality in alignment with what radiologists perceive, which plays an important role in advancing low-dose CT reconstruction techniques. An intriguing direction involves developing BIQA methods that mimic the operational characteristic of the human visual system (HVS). The internal generative mechanism (IGM) theory reveals that the HVS actively deduces primary content to enhance comprehension. In this study, we introduce an innovative BIQA metric that emulates the active inference process of IGM. Initially, an active inference module, implemented as a denoising diffusion probabilistic model (DDPM), is constructed to anticipate the primary content. Then, the dissimilarity map is derived by assessing the interrelation between the distorted image and its primary content. Subsequently, the distorted image and dissimilarity map are combined into a multi-channel image, which is inputted into a transformer-based image quality evaluator. By leveraging the DDPM-derived primary content, our approach achieves competitive performance on a low-dose CT dataset.

Active inference and deep generative modeling for cognitive ultrasound.

Ultrasound has the unique potential to offer access to medical imaging to anyone, everywhere. Devices have become ultra-portable and cost-effective, akin to the stethoscope. Nevertheless, and despite many advances, ultrasound image quality and diagnostic efficacy are still highly operator- and patient-dependent. In difficult-to-image patients, image quality is often insufficient for reliable diagnosis. In this paper, we put forth the idea that ultrasound imaging systems can be recast as information-seeking agents that engage in reciprocal interactions with their anatomical environment. Such agents autonomously adapt their transmit-receive sequences to fully personalize imaging and actively maximize information gain in-situ. To that end, we will show that the sequence of pulse-echo experiments that an ultrasound system performs can be interpreted as a perception-action loop: the action is the data acquisition, probing tissue with acoustic waves and recording reflections at the detection array, and perception is the inference of the anatomical and or functional state, potentially including associated diagnostic quantities. We then equip systems with a mechanism to actively reduce uncertainty and maximize diagnostic value across a sequence of experiments, treating action and perception jointly using Bayesian inference given generative models of the environment and action-conditional pulse-echo observations. Since the representation capacity of the generative models dictates both the quality of inferred anatomical states and the effectiveness of inferred sequences of future imaging actions, we will be greatly leveraging the enormous advances in deep generative modelling (generative AI), that are currently disrupting many fields and society at large. Finally, we show some examples of cognitive, closed-loop, ultrasound systems that perform active beamsteering and adaptive scanline selection, based on deep generative models that track anatomical belief states.

Meta-learning in active inference

Abstract Binz et al. propose meta-learning as a promising avenue for modelling human cognition. They provide an in-depth reflection on the advantages of meta-learning over other computational models of cognition, including a sound discussion on how their proposal can accommodate neuroscientific insights. We argue that active inference presents similar computational advantages while offering greater mechanistic explanatory power and biological plausibility.

A Unified Dynamic Model for the Decomposition of Skin Conductance and the Inference of Sudomotor Nerve Activities.

Electrodermal activity (EDA), commonly measured as skin conductance (SC), is a widely used physiological signal in psychological research and behavioral health applications. EDA is considered an indicator of arousal, a key aspect of emotion and stress. This work proposes a data-driven dynamic system model that characterizes the temporal dynamics of skin conductance and infers the latent arousal signal, utilizing techniques from system identification and sparse optimization. It introduces a fourth-order, linear time-invariant model for the overall skin conductance signal, including both the tonic and phasic components. The model was applied to a large dataset of over 200 participants to evaluate model fit. Furthermore, a three-component decomposition of skin conductance is introduced, based on mathematical definitions derived from the model, which provides key insights into the temporal dynamics of skin conductance. Comparative evaluation shows that the estimated latent neural signal effectively differentiates between high and low arousal states, while maintaining expected physiological properties. This work lays the foundation for numerous behavioral health applications and paves the road for designing physiology-based interventions aimed at regulating arousal.

Text processing using LLM for automatic creation of agricultural crops knowledge bases

The complexity of subject areas in which intelligent information systems operate is steadily increasing. Tasks assigned to smart agriculture systems are increasingly focused on automating and robotizing areas of human activity. Solving such tasks requires adaptive and flexible methods capable of accommodating dynamic changes in the environment in real-time. The mivar approach to creating intelligent decision-making systems enables working with adaptive discrete structures and provides methods for managerial decision-making based on adaptive active logical inference from the mivar rule knowledge base. The mivar logical inference machine forms the core of expert systems based on the mivar approach. As a result of the development of the mivar approach across various subject areas, different versions of mivar logical inference machines with their algorithms for rule traversal in the knowledge base have been created. Recent advancements in artificial intelligence and machine learning have opened new opportunities for enhancing the mivar approach. The integration of large language models for automating text processing in mivar systems significantly enhances the accuracy and efficiency of decision-making processes based on expert systems for sustainable agriculture. This paper demonstrates the feasibility of using automated text processing, intended for human training, through large language models, and its subsequent application in action planning tasks within technical systems. The proposed methodology is aimed at creating extensive knowledge bases based on textual information for real-time monitoring and decision-making in smart agriculture systems.

Quantum Markov blankets for meta-learned classical inferential paradoxes with suboptimal free energy

Abstract Quantum active Bayesian inference and quantum Markov blankets enable robust modeling and simulation of difficult-to-render natural agent-based classical inferential paradoxes interfaced with task-specific environments. Within a non-realist cognitive completeness regime, quantum Markov blankets ensure meta-learned irrational decision making is fitted to explainable manifolds at optimal free energy, where acceptable incompatible observations or temporal Bell-inequality violations represent important verifiable real-world outcomes.

Reasons for using parallel activation of logical rules in solving management tasks

The complexity of the subject areas in which intelligent information systems operate is steadily increasing. The tasks set for such systems are increasingly aimed at automating and robotizing spheres of human activity. The solution of such problems requires adaptive and flexible methods capable of taking into account dynamic changes in the environment in real time. The mivar approach to creating intelligent decision-making systems allows working with adaptive discrete structures and provides methods for making management decisions based on adaptive active logical inference based on the knowledge base of mivar rules. The mivar machine of logical inference is the core of expert systems based on the mivar approach. As a result of the historical development of the mivar approach when working on different subject areas, various versions of mivar machines of logical inference with their algorithms for bypassing rules in the knowledge base were obtained. This paper discusses the reasons for the emergence of such options and demonstrates the need to use a parallel algorithm for activating rules in an adaptive network of logical rules when solving problems of action planning in technical systems.

Design and evaluation of brain-inspired predictive coding networks based on the free-energy principle for novel neuromorphic hardware

Design and evaluation of brain-inspired predictive coding networks based on the free-energy principle for novel neuromorphic hardware

Infants infer and predict coherent event interactions: Modeling cognitive development.

Mental representations of the environment in infants are sparse and grow richer during their development. Anticipatory eye fixation studies show that infants aged around 7 months start to predict the goal of an observed action, e.g., an object targeted by a reaching hand. Interestingly, goal-predictive gaze shifts occur at an earlier age when the hand subsequently manipulates an object and later when an action is performed by an inanimate actor, e.g., a mechanical claw. We introduce CAPRI2 (Cognitive Action PRediction and Inference in Infants), a computational model that explains this development from a functional, algorithmic perspective. It is based on the theory that infants learn object files and events as they develop a physical reasoning system. In particular, CAPRI2 learns a generative event-predictive model, which it uses to both interpret sensory information and infer goal-directed behavior. When observing object interactions, CAPRI2 (i) interprets the unfolding interactions in terms of event-segmented dynamics, (ii) maximizes the coherence of its event interpretations, updating its internal estimates and (iii) chooses gaze behavior to minimize expected uncertainty. As a result, CAPRI2 mimics the developmental pathway of infants' goal-predictive gaze behavior. Our modeling work suggests that the involved event-predictive representations, longer-term generative model learning, and shorter-term retrospective and active inference principles constitute fundamental building blocks for the effective development of goal-predictive capacities.