Embodied and Extended Cognition

Predictive Coding Explains Asymmetric Connectivity in the Brain: A Neural Network Study

The paper uses neural network models to study how asymmetric feedback connections improve grasp detection in noisy visual conditions.

Planning-while-acting: addressing the continuous dynamics of planning and action in a virtually embodied task

This study investigates how people balance immediate actions with long-term planning in a virtual task, finding that participants plan ahead and adjust decisions based on spatial and embodied constraints.

Desegregation of neuronal predictive processing

This study investigates how neural circuits generate predictive representations for complex sensorimotor signals using recurrent networks and experimental validation.

Fragmentation and Multithreading of Experience in the Default-Mode Network

The paper shows that three midline prefrontal cortical regions perform distinct predictive operations during naturalistic experiences, with results generalizing across sensory modalities and content.

Dynamic Inference by Model Reduction

This study proposes an approach based on active inference with dynamic priors to infer dynamic intentions and perform tasks like trajectory prediction and object grasping.

Future shapes present: autonomous goal-directed and sensory-focused mode switching in a Bayesian allostatic network model

This paper introduces a Bayesian recurrent neural network model that autonomously switches between goal-directed and sensory-focused behaviors to adapt to environmental changes.

Embodied decisions as active inference

This paper introduces a modeling approach to study embodied decision-making, showing that decision and action processes interact through feedback loops in active inference.

The Neural Correlates of Novelty and Variability in Human Decision-Making under an Active Inference Framework

This study shows active inference better explains human decision-making under novelty and variability, with specific neural correlates identified using EEG.

PrediRep: Modelling hierarchical predictive coding with an unsupervised deep learning network

PrediRep is a deep learning model designed to align closely with hierarchical predictive coding principles while maintaining computational performance in next-frame prediction tasks.

Learning and embodied decisions in active inference

This study investigates how embodied decision-making models adapt to environmental contingencies using active inference in a two-alternative forced choice task.

Slow but flexible or fast but rigid? Discrete and continuous processes compared

This study compares discrete and continuous models in a dynamic task, finding continuous models perform better with less resource use but less flexibility.

Human Brain Dynamics Dissociate Early Perceptual and Late Motor-Related Stages of Affordance Processing

The paper examines affordance perception by comparing early perceptual and late motor-related ERP components in different movement contexts, finding that affordance perception shifts from automated to context-dependent over time.

Cerebellar state estimation enables resilient coupling across behavioural domains

This paper proposes a cerebellar model for state estimation that links neuronal dynamics with behavioral coordination, using the free energy principle to explain contextual behavior.

Needing: an active inference process for physiological motivation

This paper proposes a model of needing as an active inference process that reduces pervasive surprise, demonstrating its influence on reward processing and behavior through simulations.

Deep kinematic inference affords efficient and scalable control of bodily movements

The paper introduces an Active Inference architecture for efficient control of complex kinematic chains by mapping extrinsic to intrinsic coordinates through inference.

Body size as a metric for the affordable world

This study explores how human body size influences perception of environmental affordances, finding that body size defines a boundary for action possibilities and that this boundary is not exclusive to embodied organisms.

A stochastic world model on gravity for stability inference

The paper investigates how humans infer object stability using a stochastic world model of gravity, finding that gravity's direction is represented as a Gaussian distribution, which explains perceptual illusions and supports efficient stability inference.

Predictive neural representations of naturalistic dynamic input

This paper introduces a dynamic extension to representational similarity analysis to study predictive neural representations in naturalistic dynamic input using MEG data.

Flexible Intentions: An Active Inference Theory

The paper presents a computational theory using Active Inference to model goal-directed actions, showing how the PPC and DVS support dynamic behavior through flexible intentions.

ACTION AFFORDANCE AFFECTS PROXIMAL AND DISTAL GOAL-ORIENTED PLANNING

This study examines how action affordance realism affects eye movement patterns during goal-oriented tool interaction in virtual environments.

Predictive coding is a consequence of energy efficiency in recurrent neural networks

This paper shows that predictive coding in neural networks arises from energy efficiency rather than requiring hard-wired hierarchical connectivity.

In the Body’s Eye: The Computational Anatomy of Interoceptive Inference

This paper introduces a formal model of cardiac active inference to explain how cardiac signals influence exteroceptive perception and confidence, and demonstrates its implications for fear perception and physiological reactivity.

Prediction-error neurons in circuits with multiple neuron types: Formation, refinement and functional implications

The paper develops a computational model to study how prediction-error neurons form and function in networks with multiple inhibitory interneurons, showing their robustness and functional roles.

Simulating homeostatic, allostatic and goal-directed forms of interoceptive control using Active Inference

This paper uses Active Inference to simulate and model different forms of interoceptive control, including homeostatic, allostatic, and goal-directed, and their dysfunctions.

Canonical neural networks perform active inference

This paper shows canonical neural networks perform active inference by minimizing variational free energy through delayed plasticity and threshold adaptation.

Architectural Affordance Impacts Human Sensorimotor Brain Dynamics

This study shows that alpha-band oscillations in specific brain regions correlate with architectural affordances, influencing sensorimotor dynamics during interaction with the environment.

The Effect of Uncertainty on Prediction Error in the Action-Perception Loop

The paper shows that uncertainty affects prediction error minimization and agency judgments in the action-perception loop, with differences observed in autism traits.

Deep Active Inference and Scene Construction

This paper develops a hierarchical active inference model for visual scene construction, showing it generates empirical evidence accumulation phenomena and novel behaviors in uncertain environments.

Sources of predictive information in dynamical neural networks

This paper investigates how predictive information in dynamical neural networks originates, showing it can come from the environment or the agent's internal dynamics, and that it does not reliably indicate behavioral performance.

The Predictive Global Neuronal Workspace: A Formal Active Inference Model of Visual Consciousness

This paper presents a neurocomputational model based on Active Inference to address limitations in the global neuronal workspace theory, demonstrating its ability to reproduce experimental findings and extend the GNW taxonomy.