Submit to Special Issue Submit Abstract to Special Issue Review for Entropy Propose a Special Issue

Journal Menu

Journal Browser

► Journal Browser

Active Inference in Cognitive Neuroscience

Print Special Issue Flyer
Special Issue Editors
Special Issue Information
Keywords
Benefits of Publishing in a Special Issue
Published Papers

A special issue of Entropy (ISSN 1099-4300). This special issue belongs to the section "Entropy and Biology".

Deadline for manuscript submissions: 30 June 2026 | Viewed by 24341

Share This Special Issue

Special Issue Editor

Dr. Ivilin Stoianov

E-Mail Website
Guest Editor

Institute of Cognitive Science and Technologies (ISTC), National Research Council (CNR) of Italy, Via Martiri della Libertà 2, 35137 Padova, Italy
Interests: neurocomputational modelling; machine learning; neural networks; deep learning; bayesian reinforcement learning; visual perception; cognitive number processing; decision making and planning; spatial navigation
Special Issues, Collections and Topics in MDPI journals

Special Issue Information

Dear Colleagues,

Active inference is a normative computational theory of cognitive processing developed by Karl Friston, inspired by historical insights from Helmholtz and more recent “Bayesian brain” predictive coding perspectives. It entails a generative model that integrates sensory and internal representations of the world, enabling the brain to predict and interpret observations, encode goals (e.g., desires, intentions), predict the consequences of actions, plan, make choices, and exert control.

Perception, control, and learning are unified under the objective of minimizing surprise: perception refines internal representations, control selects actions that fulfill internal predictions, and learning improves the generative model itself.

The theory exists in several forms, including continuous-time active inference, addressing lower-level processes such as perception and motor control; discrete-time active inference, which supports higher-level cognitive phenomena like decision making and planning; and hybrid active inference, enabling full-fledged cognition and, more broadly, intelligent behavior. Building on the increasing attention Entropy has devoted to active inference in recent years—and the growing research interest from phenomenological disciplines—the goal of this Special Issue is to advance active inference as a theory of cognition by gathering novel empirical and computational evidence from a cognitive neuroscience perspective.

We invite original submissions, commentaries, review articles, and highlights of key innovations, focusing on the following:

Novel empirical evidence for active inference (behavioral, neural, clinical);
Interpretations of published data through an active inference lens;
Computational advances with an emphasis of biological plausibility;
Philosophical essays exploring theoretical implications;
Applications of active inference, such as its use as a research tool or for human–machine interaction.

Dr. Ivilin Stoianov
Guest Editor

Manuscript Submission Information

Manuscripts should be submitted online at www.mdpi.com by registering and logging in to this website. Once you are registered, click here to go to the submission form. Manuscripts can be submitted until the deadline. All submissions that pass pre-check are peer-reviewed. Accepted papers will be published continuously in the journal (as soon as accepted) and will be listed together on the special issue website. Research articles, review articles as well as short communications are invited. For planned papers, a title and short abstract (about 250 words) can be sent to the Editorial Office for assessment.

Submitted manuscripts should not have been published previously, nor be under consideration for publication elsewhere (except conference proceedings papers). All manuscripts are thoroughly refereed through a single-blind peer-review process. A guide for authors and other relevant information for submission of manuscripts is available on the Instructions for Authors page. Entropy is an international peer-reviewed open access monthly journal published by MDPI.

Please visit the Instructions for Authors page before submitting a manuscript. The Article Processing Charge (APC) for publication in this open access journal is 2600 CHF (Swiss Francs). Submitted papers should be well formatted and use good English. Authors may use MDPI's English editing service prior to publication or during author revisions.

Keywords

active inference
predictive coding
free energy principle
prediction errors
neurocomputational modeling
perception
motor control
cognitive control
decision making
cognitive neuroscience
neuropsychology
psychiatry

Benefits of Publishing in a Special Issue

Ease of navigation: Grouping papers by topic helps scholars navigate broad scope journals more efficiently.
Greater discoverability: Special Issues support the reach and impact of scientific research. Articles in Special Issues are more discoverable and cited more frequently.
Expansion of research network: Special Issues facilitate connections among authors, fostering scientific collaborations.
External promotion: Articles in Special Issues are often promoted through the journal's social media, increasing their visibility.
Reprint: MDPI Books provides the opportunity to republish successful Special Issues in book format, both online and in print.

Further information on MDPI's Special Issue policies can be found here.

Published Papers (10 papers)

Download All Papers

Order results

Result details

Show export options Show export options

Select all

Export citation of selected articles as:

Research

45 pages, 1426 KB

Open AccessArticle

Chaotic Itinerancy in Collective Behaviour Emerging from Active Inference: A Multi-Agent Model of Trust and Empowerment Dynamics in Theatre Workshops

by Shoko Miyano and Takashi Shiono

Entropy 2026, 28(5), 491; https://doi.org/10.3390/e28050491 - 24 Apr 2026

Viewed by 320

Abstract

Chaotic itinerancy—irregular switching among metastable collective states—provides a dynamical substrate for flexible social coordination, yet its mechanistic origin in multi-agent systems remains unclear. We present a multi-agent Active Inference model in which chaotic itinerancy emerges from Expected Free Energy minimisation without outcome-level social priors. Agents select actions to minimise Expected Free Energy while updating preferences through a precision-gated learning mechanism modulated by interpersonal trust. Hill-function nonlinearity in state transitions creates bistable “affordance landscapes” that gate behavioural mode switching. Simulations with small number of agents on an Erdős–Rényi trust network reveal spontaneous alternation among multiple metastable behavioural clusters, heavy-tailed dwell-time distributions, and sign-changing finite-time Lyapunov exponents—three hallmarks of chaotic itinerancy. Crucially, replacing Hill-function dynamics with linear transitions reduces the chaotic-itinerancy detection rate from 80% to 20%, demonstrating that nonlinear affordance structure is necessary for generating metastable switching. We further show that agents with simplified internal models of the world sustain richer itinerant dynamics as a group than “perfect-foresight” agents, suggesting that bounded rationality may be functionally advantageous for maintaining behavioural flexibility. These results establish active inference as a principled framework for modelling chaotic itinerancy in social systems and offer a computational account of trust-mediated collective transitions observed in theatre workshops and group dynamics. Full article

(This article belongs to the Special Issue Active Inference in Cognitive Neuroscience)

► Show Figures

Figure 1

15 pages, 2622 KB

Open AccessArticle

Contextual Modulation of Semantic Coherence in vmPFC Patients’ Mental Constructions

by Debora Stendardi, Matteo Reale, Francesca Dalle Piagge, Elena Garavini, Michela Grasselli and Elisa Ciaramelli

Entropy 2026, 28(5), 488; https://doi.org/10.3390/e28050488 - 24 Apr 2026

Viewed by 378

Abstract

Previous evidence has identified the ventromedial prefrontal cortex (vmPFC) as crucial for implementing high-level semantic memory structures (schemas) during event construction. If this is the case, one would expect reduced semantic coherence in events mentally constructed by vmPFC patients compared to healthy and brain-damaged controls. We tested this prediction by having participants mentally construct events using objects as cues and reanalyzing a published dataset using sentences as cues. In both cases, we measured the semantic coherence of patients’ mental constructions and their semantic coherence with the cue, using transformer-based sentence embeddings (S-BERT), and further corroborated the findings with E5 Multilingual and E5 Italian embedding models. Our results reveal that the hypothesized impairment in semantic coherence following vmPFC damage is, in fact, task-dependent. With minimal (object) cues, vmPFC patients’ reports exhibited reduced local coherence, increased connectedness to the cues, and reduced lexical diversity. In contrast, with extended (sentence) cues, they showed preserved- or even enhanced-local and global coherence. We suggest that vmPFC integrity is necessary to trigger schema activation under minimal cue conditions. Although extended cues may facilitate schema activation, schemas are degraded and essentialized following vmPFC damage, thereby constraining patients’ mental constructions within a narrower—hence overly coherent—semantic space. Full article

(This article belongs to the Special Issue Active Inference in Cognitive Neuroscience)

► Show Figures

Figure 1

22 pages, 307 KB

Open AccessArticle

The Awareness-First Theory: A Coherence Principle Underlying Active Inference and Physical Law

by Jason Clarke

Entropy 2026, 28(3), 306; https://doi.org/10.3390/e28030306 - 9 Mar 2026

Viewed by 1418

Abstract

The Free Energy Principle (FEP) and Active Inference provide a unifying variational framework for modelling perception, action, learning, and self-organisation across biological systems. While highly successful at explaining how systems maintain organisation under uncertainty, these frameworks remain explicitly neutral with respect to a foundational question: why there is experience at all. This paper argues that this limitation reflects not an empirical gap but a misplaced starting point. The Awareness-First Theory (AFT) inverts the usual explanatory order by beginning from the givenness of awareness itself and asking what must be the case for any world to appear coherently. This requirement is formalised as a Coherence Principle, expressed as a variational stationarity condition,

δ A = 0

, which specifies the invariance of coherent awareness across changing appearances. I argue that familiar variational principles-most notably free-energy minimisation (

δ F = 0

) and stationary-action physics (

δ S = 0

)-can be understood as restricted projections of this parent constraint under specific abstractions. Active Inference therefore does not generate awareness but describes how locally bounded systems maintain coherence within awareness under uncertainty. Making this projection structure explicit dissolves the explanatory gap between physical process and phenomenal presence, revealing the gap itself as a category error. Although the Coherence Principle itself is transcendental rather than empirical, the AFT generates testable consequences at the level of its projections, including predicted dissociations between inferential optimisation and phenomenological coherence in dreaming, altered states, meditation, and psychopathology. Full article

(This article belongs to the Special Issue Active Inference in Cognitive Neuroscience)

28 pages, 2533 KB

Open AccessArticle

Intermittent Active Inference

by Markus Klar, Sebastian Stein, Fraser Paterson, John H. Williamson, Henrik Gollee and Roderick Murray-Smith

Entropy 2026, 28(3), 269; https://doi.org/10.3390/e28030269 - 28 Feb 2026

Viewed by 1487

Abstract

Active inference provides a unified framework for perception and action as processes of minimizing prediction error given a generative model of the environment. Whilst standard formulations assume continuous inference and control, empirical evidence indicates that humans update their control strategies intermittently, which reduces computational demands and mitigates propagation of correlated noise in closed feedback loops. To address this, we introduce Intermittent Active Inference (IAIF), a novel variant in which sensing, inference, planning, or acting can occur intermittently. This paper investigates intermittent planning, where IAIF agents follow their current plan and only re-plan when the prediction error exceeds a predefined threshold or the Expected Free Energy associated with the current plan surpasses prior estimates. We evaluate intermittent planning in a mouse pointing task, comparing against continuous planning while examining the impact of different threshold parameters on performance and efficiency. The findings indicate that IAIF reduces computation time whilst maintaining task performance, particularly when the number of plans sampled during planning is increased. In case of the proposed trigger based on Expected Free Energy, no additional calibration is required for this. The straightforward integration of IAIF makes it valuable in practical modelling workflows. Full article

(This article belongs to the Special Issue Active Inference in Cognitive Neuroscience)

► Show Figures

Figure 1

33 pages, 3062 KB

Open AccessEditor’s ChoiceArticle

Gradient-Free De Novo Learning

by Karl Friston, Thomas Parr, Conor Heins, Lancelot Da Costa, Tommaso Salvatori, Alexander Tschantz, Magnus Koudahl, Toon Van de Maele, Christopher Buckley and Tim Verbelen

Entropy 2025, 27(9), 992; https://doi.org/10.3390/e27090992 - 22 Sep 2025

Cited by 1 | Viewed by 3231

Abstract

This technical note applies active inference to the problem of learning goal-directed behaviour from scratch, namely, de novo learning. By de novo learning, we mean discovering, directly from observations, the structure and parameters of a discrete generative model for sequential policy optimisation. Concretely, our procedure grows and then reduces a model until it discovers a pullback attractor over (generalised) states; this attracting set supplies paths of least action among goal states while avoiding costly states. The implicit efficiency rests upon reframing the learning problem through the lens of the free energy principle, under which it is sufficient to learn a generative model whose dynamics feature such an attracting set. For context, we briefly relate this perspective to value-based formulations (e.g., Bellman optimality) and then apply the active inference formulation to a small arcade game to illustrate de novo structure learning and ensuing agency. Full article

(This article belongs to the Special Issue Active Inference in Cognitive Neuroscience)

► Show Figures

Figure 1

21 pages, 2034 KB

Open AccessArticle

Brain Oscillations and Autonomic Synthonization via Comodulation in Collaborative Negotiation

by Katia Rovelli, Carlotta Acconito, Laura Angioletti and Michela Balconi

Entropy 2025, 27(8), 873; https://doi.org/10.3390/e27080873 - 18 Aug 2025

Viewed by 1460

Abstract

This study investigates the relationship between neural and physiological synthonization via comodulation (Synth) in dyadic exchanges centered on negotiation processes. In total, 13 dyads participated in a negotiation task with three phases: Initiation (IP), Negotiation Core (NCP), and Resolution (RP). Electroencephalographic (EEG) frequency bands (i.e., delta, theta, alpha) and autonomic responses (heart rate variability, HRV) were recorded. Synth was analyzed using Euclidean distance (EuDist) for EEG and autonomic indices. Significant Synth in delta, theta, and alpha bands in temporo-central and parieto-occipital regions was observed, indicating social cognitive alignment. HRV Synth was higher during the NCP than IP, suggesting better coordination. Based on this result, a cluster analysis was performed on HRV EuDist to identify distinct groups based on HRV, and eventually personality patterns, that revealed one cluster with higher Synth and reward sensitivity, and another with lower Synth and reward sensitivity. These findings show how neural and autonomic Synth enhances social cognition and emotional regulation. Full article

(This article belongs to the Special Issue Active Inference in Cognitive Neuroscience)

► Show Figures

Figure 1

23 pages, 8311 KB

Open AccessFeature PaperEditor’s ChoiceArticle

Active Inference with Dynamic Planning and Information Gain in Continuous Space by Inferring Low-Dimensional Latent States

by Takazumi Matsumoto, Kentaro Fujii, Shingo Murata and Jun Tani

Entropy 2025, 27(8), 846; https://doi.org/10.3390/e27080846 - 9 Aug 2025

Cited by 1 | Viewed by 3444

Abstract

Active inference offers a unified framework in which agents can exhibit both goal-directed and epistemic behaviors. However, implementing policy search in high-dimensional continuous action spaces presents challenges in terms of scalability and stability. Our previously proposed model, T-GLean, addressed this issue by enabling efficient goal-directed planning through low-dimensional latent space search, further reduced by conditioning on prior habituated behavior. However, the lack of an epistemic term in minimizing expected free energy limited the agent’s ability to engage in information-seeking behavior that can be critical for attaining preferred outcomes. In this study, we present EFE-GLean, an extended version of T-GLean that overcomes this limitation by integrating epistemic value into the planning process. EFE-GLean generates goal-directed policies by inferring low-dimensional future posterior trajectories while maximizing expected information gain. Simulation experiments using an extended T-maze task—implemented in both discrete and continuous domains—demonstrate that the agent can successfully achieve its goals by exploiting hidden environmental information. Furthermore, we show that the agent is capable of adapting to abrupt environmental changes by dynamically revising plans through simultaneous minimization of past variational free energy and future expected free energy. Finally, analytical evaluations detail the underlying mechanisms and computational properties of the model. Full article

(This article belongs to the Special Issue Active Inference in Cognitive Neuroscience)

► Show Figures

Figure 1

42 pages, 3822 KB

Open AccessArticle

The Criticality of Consciousness: Excitatory–Inhibitory Balance and Dual Memory Systems in Active Inference

by Don M. Tucker, Phan Luu and Karl J. Friston

Entropy 2025, 27(8), 829; https://doi.org/10.3390/e27080829 - 4 Aug 2025

Cited by 2 | Viewed by 5191

Abstract

The organization of consciousness is described through increasingly rich theoretical models. We review evidence that working memory capacity—essential to generating consciousness in the cerebral cortex—is supported by dual limbic memory systems. These dorsal (Papez) and ventral (Yakovlev) limbic networks provide the basis for mnemonic processing and prediction in the dorsal and ventral divisions of the human neocortex. Empirical evidence suggests that the dorsal limbic division is (i) regulated preferentially by excitatory feedforward control, (ii) consolidated by REM sleep, and (iii) controlled in waking by phasic arousal through lemnothalamic projections from the pontine brainstem reticular activating system. The ventral limbic division and striatum, (i) organizes the inhibitory neurophysiology of NREM to (ii) consolidate explicit memory in sleep, (iii) operating in waking cognition under the same inhibitory feedback control supported by collothalamic tonic activation from the midbrain. We propose that (i) these dual (excitatory and inhibitory) systems alternate in the stages of sleep, and (ii) in waking they must be balanced—at criticality—to optimize the active inference that generates conscious experiences. Optimal Bayesian belief updating rests on balanced feedforward (excitatory predictive) and feedback (inhibitory corrective) control biases that play the role of prior and likelihood (i.e., sensory) precision. Because the excitatory (E) phasic arousal and inhibitory (I) tonic activation systems that regulate these dual limbic divisions have distinct affective properties, varying levels of elation for phasic arousal (E) and anxiety for tonic activation (I), the dual control systems regulate sleep and consciousness in ways that are adaptively balanced—around the entropic nadir of E–I criticality—for optimal self-regulation of consciousness and psychological health. Because they are emotive as well as motive control systems, these dual systems have unique qualities of feeling that may be registered as subjective experience. Full article

(This article belongs to the Special Issue Active Inference in Cognitive Neuroscience)

► Show Figures

Figure 1

33 pages, 3843 KB

Open AccessArticle

Deep Hybrid Models: Infer and Plan in a Dynamic World

by Matteo Priorelli and Ivilin Peev Stoianov

Entropy 2025, 27(6), 570; https://doi.org/10.3390/e27060570 - 27 May 2025

Cited by 2 | Viewed by 1718

Abstract

To determine an optimal plan for complex tasks, one often deals with dynamic and hierarchical relationships between several entities. Traditionally, such problems are tackled with optimal control, which relies on the optimization of cost functions; instead, a recent biologically motivated proposal casts planning and control as an inference process. Active inference assumes that action and perception are two complementary aspects of life whereby the role of the former is to fulfill the predictions inferred by the latter. Here, we present an active inference approach that exploits discrete and continuous processing, based on three features: the representation of potential body configurations in relation to the objects of interest; the use of hierarchical relationships that enable the agent to easily interpret and flexibly expand its body schema for tool use; the definition of potential trajectories related to the agent’s intentions, used to infer and plan with dynamic elements at different temporal scales. We evaluate this deep hybrid model on a habitual task: reaching a moving object after having picked a moving tool. We show that the model can tackle the presented task under different conditions. This study extends past work on planning as inference and advances an alternative direction to optimal control. Full article

(This article belongs to the Special Issue Active Inference in Cognitive Neuroscience)

► Show Figures

Figure 1

18 pages, 286 KB

Open AccessArticle

The Physics and Metaphysics of Social Powers: Bridging Cognitive Processing and Social Dynamics, a New Perspective on Power Through Active Inference

by Mahault Albarracin, Sonia de Jager and David Hyland

Entropy 2025, 27(5), 522; https://doi.org/10.3390/e27050522 - 14 May 2025

Cited by 1 | Viewed by 2883

Abstract

Power operates across multiple scales, from physical action to complex social dynamics, and is constrained by fundamental principles. In the social realm, power is shaped by interactions and cognitive capacity: socially-facilitated empowerment enhances an agent’s information-processing ability, either by delegating tasks or leveraging collective resources. This computational advantage expands access to policies and buffers against vulnerabilities, amplifying an individual’s or group’s influence. In AIF, social power emerges from the capacity to attract attention and process information effectively. Our semantic habitat—narratives, ideologies, representations, etc.—functions through attentional scripts that coordinate social behavior. Shared scripts shape power dynamics by structuring collective attention. Speculative scripts serve as cognitive tools for low-risk learning, allowing agents to explore counterfactuals and refine predictive models. However, dominant scripts can reinforce misinformation, echo chambers, and power imbalances by directing collective attention toward self-reinforcing policies. We argue that power through scripts stems not only from associations with influential agents but also from the ability to efficiently process information, creating a feedback loop of increasing influence. This reframes power beyond traditional material and cultural dimensions, towards an informational and computational paradigm—what we term possibilistic power, i.e., the capacity to explore and shape future trajectories. Understanding these mechanisms has critical implications for political organization and technological foresight. Full article

(This article belongs to the Special Issue Active Inference in Cognitive Neuroscience)

Show export options Show export options

Select all

Export citation of selected articles as:

Displaying articles 1-10

Journal Menu

Journal Browser

Active Inference in Cognitive Neuroscience

Share This Special Issue

Special Issue Editor

Special Issue Information

Keywords

Benefits of Publishing in a Special Issue

Published Papers (10 papers)

Research

Further Information

Guidelines

MDPI Initiatives

Follow MDPI