Search

Asa result of the time lag in transmission, the data obtained by the sensor is delayed and does not reflect the state at the current moment. The effects of input delay are often overlooked in active inference (AIF), which may lead to significant deviations in state estimation and increased prediction errors, particularly when the system is subjected to a sudden external stimulus. In this paper, a theoretical framework of delayed feedback active inference (DAIF) is proposed to enhance the applicability of AIF to real systems. The probability model of DAIF is defined by incorporating a control distribution into that of AIF. The free energy of DAIF is defined as the sum of the quadratic state, sense, and control prediction error. A predicted state derived from previous states is defined and introduced as the expectation of the prior distribution of the real-time state. A proportional-integral (PI)-like control based on the predicted state is taken to be the expectation of DAIF preference control, whose gain coefficient is inversely proportional to the measurement accuracy variance. To adaptively compensate for external disturbances, a second-order inverse variance accuracy replaces the fixed sensory accuracy of preference control. The simulation results of the trajectory tracking control of a quadrotor unmanned aerial vehicle (UAV) show that DAIF performs better than AIF in state estimation and disturbance resistance. [ABSTRACT FROM AUTHOR]

Major Depressive Disorder (MDD) is a complex, heterogeneous condition affecting millions worldwide. Computational neuropsychiatry offers potential breakthroughs through the mechanistic modeling of this disorder. Using the Kolmogorov theory (KT) of consciousness, we developed a foundational model where algorithmic agents interact with the world to maximize an Objective Function evaluating affective valence. Depression, defined in this context by a state of persistently low valence, may arise from various factors—including inaccurate world models (cognitive biases), a dysfunctional Objective Function (anhedonia, anxiety), deficient planning (executive deficits), or unfavorable environments. Integrating algorithmic, dynamical systems, and neurobiological concepts, we map the agent model to brain circuits and functional networks, framing potential etiological routes and linking with depression biotypes. Finally, we explore how brain stimulation, psychotherapy, and plasticity-enhancing compounds such as psychedelics can synergistically repair neural circuits and optimize therapies using personalized computational models. [ABSTRACT FROM AUTHOR]

The free energy principle is a formal theory of adaptive self-organising systems that emerged from statistical thermodynamics, machine learning and theoretical neuroscience and has since been translated into biologically plausible 'process theories' of cognition and behaviour, which fall under the banner of 'active inference'. Despite the promise this theory holds for theorising, research and practical applications in psychology and psychiatry, its impact on these disciplines has only now begun to bear fruit. The aim of this treatment is to consider the extent to which active inference has informed theoretical progress in psychology, before exploring its contributions to our understanding and treatment of psychopathology. Despite facing persistent translational obstacles, progress suggests that active inference has the potential to become a new paradigm that promises to unite psychology's subdisciplines, while readily incorporating the traditionally competing paradigms of evolutionary and developmental psychology. To date, however, progress towards this end has been slow. Meanwhile, the main outstanding question is whether this theory will make a positive difference through applications in clinical psychology, and its sister discipline of psychiatry. [ABSTRACT FROM AUTHOR]

Active inference describes (Bayes-optimal) behaviour as being motivated by the minimisation of surprise of one's sensory observations, through the optimisation of a generative model (of the hidden causes of one's sensory data) in the brain. One of active inference's key appeals is its conceptualisation of precision as biasing neuronal communication and, thus, inference within generative models. The importance of precision in perceptual inference is evident—many studies have demonstrated the importance of ensuring precision estimates are correct for normal (healthy) sensation and perception. Here, we highlight the many roles precision plays in action, i.e., the key processes that rely on adequate estimates of precision, from decision making and action planning to the initiation and control of muscle movement itself. Thereby, we focus on the recent development of hierarchical, "mixed" models—generative models spanning multiple levels of discrete and continuous inference. These kinds of models open up new perspectives on the unified description of hierarchical computation, and its implementation, in action. Here, we highlight how these models reflect the many roles of precision in action—from planning to execution—and the associated pathologies if precision estimation goes wrong. We also discuss the potential biological implementation of the associated message passing, focusing on the role of neuromodulatory systems in mediating different kinds of precision. [ABSTRACT FROM AUTHOR]

The representation of intelligence is achieved by patterns of connections among neurons in brains and machines. Brains grow continuously, such that their patterns of connections develop through activity-dependent specification, with the continuing ontogenesis of individual experience. The theory of active inference proposes that the developmental organization of sentient systems reflects general processes of informatic self-evidencing, through the minimization of free energy. We interpret this theory to imply that the mind may be described in information terms that are not dependent on a specific physical substrate. At a certain level of complexity, self-evidencing of living (self-organizing) information systems becomes hierarchical and reentrant, such that effective consciousness emerges as the consequence of a good regulator. We propose that these principles imply that an adequate reconstruction of the computational dynamics of an individual human brain/mind is possible with sufficient neuromorphic computational emulation. [ABSTRACT FROM AUTHOR]

According to active inference, constantly running prediction engines in our brain play a large role in delivering all human experience. These predictions help deliver everything we see, hear, touch, and feel. In this paper, I pursue one apparent consequence of this increasingly well-supported view. Given the constant influence of hidden predictions on human experience, can we leverage the power of prediction in the service of human flourishing? Can we learn to hack our own predictive regimes in ways that better serve our needs and purposes? Asking this question rapidly reveals a landscape that is at once familiar and new. It is also challenging, suggesting important questions about scope and dangers while casting further doubt (as if any was needed) on old assumptions about a firm mind/body divide. I review a range of possible hacks, starting with the careful use of placebos, moving on to look at chronic pain and functional disorders, and ending with some speculations concerning the complex role of genetic influences on the predictive brain. [ABSTRACT FROM AUTHOR]

This paper develops an outline for a hierarchically embedded architecture of an artificial agent that models human translation processes based on principles of active inference (AIF) and predictive processing (PP). AIF and PP posit that the mind constructs a model of the environment which guides behavior by continually generating and integrating predictions and sensory input. The proposed model of the translation agent consists of three processing strata: a sensorimotor layer, a cognitive layer, and a phenomenal layer. Each layer consists of a network of states and transitions that interact on different time scales. Following the AIF framework, states are conditioned on observations which may originate from the environment and/or the embedded processing layer, while transitions between states are conditioned on actions that implement plans to optimize goal-oriented behavior. The AIF agent aims at simulating the variation in translational behavior under various conditions and to facilitate investigating the underlying mental mechanisms. It provides a novel framework for generating and testing new hypotheses of the translating mind. [ABSTRACT FROM AUTHOR]

Taking inspiration from humans can help catalyse embodied AI solutions for important real-world applications. Current human-inspired tools include neuromorphic systems and the developmental approach to learning. However, this developmental neurorobotics approach is currently lacking important frameworks for human-like computation and learning. We propose that human-like computation is inherently embodied, with its interface to the world being neuromorphic, and its learning processes operating across different timescales. These constraints necessitate a unified framework: active inference, underpinned by the free energy principle (FEP). Herein, we describe theoretical and empirical support for leveraging this framework in embodied neuromorphic agents with autonomous mental development. We additionally outline current implementation approaches (including toolboxes) and challenges, and we provide suggestions for next steps to catalyse this important field. [ABSTRACT FROM AUTHOR]

Understandings of how visual hallucinations appear have been highly influenced by generative approaches, in particular Friston's Active Inference conceptualization. Their core proposition is that these phenomena occur when hallucinatory expectations outweigh actual sensory data. This imbalance occurs as the brain seeks to minimize informational free energy, a measure of the distance between predicted and actual sensory data in a stationary open system. We review this approach in the light of old and new information on the role of environmental factors in episodic hallucinations. In particular, we highlight the possible relationship of specific visual triggers to the onset and offset of some episodes. We use an analogy from phase transitions in physics to explore factors which might account for intermittent shifts between veridical and hallucinatory vision. In these triggered forms of hallucinations, we suggest that there is a transient disturbance in the normal one-to-one correspondence between a real object and the counterpart perception such that this correspondence becomes between the real object and a hallucination. Generative models propose that a lack of information transfer from the environment to the brain is one of the key features of hallucinations. In contrast, we submit that specific information transfer is required at onset and offset in these cases. We propose that this transient one-to-one correspondence between environment and hallucination is mediated more by aberrant discriminative than by generative inference. Discriminative inference can be conceptualized as a process for maximizing shared information between the environment and perception within a self-organizing nonstationary system. We suggest that generative inference plays the greater role in established hallucinations and in the persistence of individual hallucinatory episodes. We further explore whether thermodynamic free energy may be an additional factor in why hallucinations are temporary. Future empirical research could productively concentrate on three areas. Firstly, subjective perceptual changes and parallel variations in brain function during specific transitions between veridical and hallucinatory vision to inform models of how episodes occur. Secondly, systematic investigation of the links between environment and hallucination episodes to probe the role of information transfer in triggering transitions between veridical and hallucinatory vision. Finally, changes in hallucinatory episodes over time to elucidate the role of learning on phenomenology. These empirical data will allow the potential roles of different forms of inference in the stages of hallucinatory episodes to be elucidated. [ABSTRACT FROM AUTHOR]

Karl Friston's free-energy principle casts agents as self-evidencing through active inference. This implies that decision-making, planning and information-seeking are, in a generic sense, 'wishful'. We take an interdisciplinary perspective on this perplexing aspect of the free-energy principle and unpack the epistemological implications of wishful thinking under the free-energy principle. We use this epistemic framing to discuss the emergence of biases for self-evidencing agents. In particular, we argue that this elucidates an optimism bias as a foundational tenet of self-evidencing. We allude to a historical precursor to some of these themes, interestingly found in Machiavelli's oeuvre, to contextualise the universal optimism of the free-energy principle. [ABSTRACT FROM AUTHOR]

Given the rapid advancement of artificial intelligence, understanding the foundations of intelligent behaviour is increasingly important. Active inference, regarded as a general theory of behaviour, offers a principled approach to probing the basis of sophistication in planning and decision-making. This paper examines two decision-making schemes in active inference based on "planning" and "learning from experience". Furthermore, we also introduce a mixed model that navigates the data complexity trade-off between these strategies, leveraging the strengths of both to facilitate balanced decision-making. We evaluate our proposed model in a challenging grid-world scenario that requires adaptability from the agent. Additionally, our model provides the opportunity to analyse the evolution of various parameters, offering valuable insights and contributing to an explainable framework for intelligent decision-making. [ABSTRACT FROM AUTHOR]

New insights into intractable industrial challenges can be revealed by framing them in terms of natural science. One intractable industrial challenge is that creative production can be much more financially expensive and time consuming than standardized production. Creative products include a wide range of goods that have one or more original characteristics. The scaling up of creative production is hindered by high financial production costs and long production durations. In this paper, creative production is framed in terms of interactions between entropy and complexity during progressions from emergent pragmatics to action semantics. An analysis of interactions between entropy and complexity is provided that relates established practice in creative production to organizational survival in changing environments. The analysis in this paper is related to assembly theory, which is a recent theoretical development in natural science that addresses how open-ended generation of complex physical objects can emerge from selection in biology. Parallels between assembly practice in industrial production and assembly theory in natural science are explained through constructs that are common to both, such as assembly index. Overall, analyses reported in the paper reveal that interactions between entropy and complexity underlie intractable challenges in creative production, from the production of individual products to the survival of companies. [ABSTRACT FROM AUTHOR]

Asa result of the time lag in transmission, the data obtained by the sensor is delayed and does not reflect the state at the current moment. The effects of input delay are often overlooked in active inference (AIF), which may lead to significant deviations in state estimation and increased prediction errors, particularly when the system is subjected to a sudden external stimulus. In this paper, a theoretical framework of delayed feedback active inference (DAIF) is proposed to enhance the applicability of AIF to real systems. The probability model of DAIF is defined by incorporating a control distribution into that of AIF. The free energy of DAIF is defined as the sum of the quadratic state, sense, and control prediction error. A predicted state derived from previous states is defined and introduced as the expectation of the prior distribution of the real-time state. A proportional-integral (PI)-like control based on the predicted state is taken to be the expectation of DAIF preference control, whose gain coefficient is inversely proportional to the measurement accuracy variance. To adaptively compensate for external disturbances, a second-order inverse variance accuracy replaces the fixed sensory accuracy of preference control. The simulation results of the trajectory tracking control of a quadrotor unmanned aerial vehicle (UAV) show that DAIF performs better than AIF in state estimation and disturbance resistance.

In this paper, we unite concepts from Husserlian phenomenology, the active inference framework in theoretical biology, and category theory in mathematics to develop a comprehensive framework for understanding social action premised on shared goals. We begin with an overview of Husserlian phenomenology, focusing on aspects of inner time-consciousness, namely, retention, primal impression, and protention. We then review active inference as a formal approach to modeling agent behavior based on variational (approximate Bayesian) inference. Expanding upon Husserl's model of time consciousness, we consider collective goal-directed behavior, emphasizing shared protentions among agents and their connection to the shared generative models of active inference. This integrated framework aims to formalize shared goals in terms of shared protentions, and thereby shed light on the emergence of group intentionality. Building on this foundation, we incorporate mathematical tools from category theory, in particular, sheaf and topos theory, to furnish a mathematical image of individual and group interactions within a stochastic environment. Specifically, we employ morphisms between polynomial representations of individual agent models, allowing predictions not only of their own behaviors but also those of other agents and environmental responses. Sheaf and topos theory facilitates the construction of coherent agent worldviews and provides a way of representing consensus or shared understanding. We explore the emergence of shared protentions, bridging the phenomenology of temporal structure, multi-agent active inference systems, and category theory. Shared protentions are highlighted as pivotal for coordination and achieving common objectives. We conclude by acknowledging the intricacies stemming from stochastic systems and uncertainties in realizing shared goals. [ABSTRACT FROM AUTHOR]

A theoretical account of development in mesocortical anatomy is derived from the free energy principle, operating in a neural field with both Hebbian and anti-Hebbian neural plasticity. An elementary structural unit is proposed, in which synaptic connections at mesoscale are arranged in paired patterns with mirror symmetry. Exchanges of synaptic flux in each pattern form coupled spatial eigenmodes, and the line of mirror reflection between the paired patterns operates as a Markov blanket, so that prediction errors in exchanges between the pairs are minimized. The theoretical analysis is then compared to the outcomes from a biological model of neocortical development, in which neuron precursors are selected by apoptosis for cell body and synaptic connections maximizing synchrony and also minimizing axonal length. It is shown that this model results in patterns of connection with the anticipated mirror symmetries, at micro-, meso- and inter-arial scales, among lateral connections, and in cortical depth. This explains the spatial organization and functional significance of neuron response preferences, and is compatible with the structural form of both columnar and noncolumnar cortex. Multi-way interactions of mirrored representations can provide a preliminary anatomically realistic model of cortical information processing. [ABSTRACT FROM AUTHOR]

The representation of intelligence is achieved by patterns of connections among neurons in brains and machines. Brains grow continuously, such that their patterns of connections develop through activity-dependent specification, with the continuing ontogenesis of individual experience. The theory of active inference proposes that the developmental organization of sentient systems reflects general processes of informatic self-evidencing, through the minimization of free energy. We interpret this theory to imply that the mind may be described in information terms that are not dependent on a specific physical substrate. At a certain level of complexity, self-evidencing of living (self-organizing) information systems becomes hierarchical and reentrant, such that effective consciousness emerges as the consequence of a good regulator. We propose that these principles imply that an adequate reconstruction of the computational dynamics of an individual human brain/mind is possible with sufficient neuromorphic computational emulation.

Active inference describes (Bayes-optimal) behaviour as being motivated by the minimisation of surprise of one’s sensory observations, through the optimisation of a generative model (of the hidden causes of one’s sensory data) in the brain. One of active inference’s key appeals is its conceptualisation of precision as biasing neuronal communication and, thus, inference within generative models. The importance of precision in perceptual inference is evident—many studies have demonstrated the importance of ensuring precision estimates are correct for normal (healthy) sensation and perception. Here, we highlight the many roles precision plays in action, i.e., the key processes that rely on adequate estimates of precision, from decision making and action planning to the initiation and control of muscle movement itself. Thereby, we focus on the recent development of hierarchical, “mixed” models—generative models spanning multiple levels of discrete and continuous inference. These kinds of models open up new perspectives on the unified description of hierarchical computation, and its implementation, in action. Here, we highlight how these models reflect the many roles of precision in action—from planning to execution—and the associated pathologies if precision estimation goes wrong. We also discuss the potential biological implementation of the associated message passing, focusing on the role of neuromodulatory systems in mediating different kinds of precision.

The free energy principle is a formal theory of adaptive self-organising systems that emerged from statistical thermodynamics, machine learning and theoretical neuroscience and has since been translated into biologically plausible ‘process theories’ of cognition and behaviour, which fall under the banner of ‘active inference’. Despite the promise this theory holds for theorising, research and practical applications in psychology and psychiatry, its impact on these disciplines has only now begun to bear fruit. The aim of this treatment is to consider the extent to which active inference has informed theoretical progress in psychology, before exploring its contributions to our understanding and treatment of psychopathology. Despite facing persistent translational obstacles, progress suggests that active inference has the potential to become a new paradigm that promises to unite psychology’s subdisciplines, while readily incorporating the traditionally competing paradigms of evolutionary and developmental psychology. To date, however, progress towards this end has been slow. Meanwhile, the main outstanding question is whether this theory will make a positive difference through applications in clinical psychology, and its sister discipline of psychiatry.

According to active inference, constantly running prediction engines in our brain play a large role in delivering all human experience. These predictions help deliver everything we see, hear, touch, and feel. In this paper, I pursue one apparent consequence of this increasingly well-supported view. Given the constant influence of hidden predictions on human experience, can we leverage the power of prediction in the service of human flourishing? Can we learn to hack our own predictive regimes in ways that better serve our needs and purposes? Asking this question rapidly reveals a landscape that is at once familiar and new. It is also challenging, suggesting important questions about scope and dangers while casting further doubt (as if any was needed) on old assumptions about a firm mind/body divide. I review a range of possible hacks, starting with the careful use of placebos, moving on to look at chronic pain and functional disorders, and ending with some speculations concerning the complex role of genetic influences on the predictive brain.

Robust evidence suggests that humans explore their environment using a combination of topological landmarks and coarse-grained path integration. This approach relies on identifiable environmental features (topological landmarks) in tandem with estimations of distance and direction (coarse-grained path integration) to construct cognitive maps of the surroundings. This cognitive map is believed to exhibit a hierarchical structure, allowing efficient planning when solving complex navigation tasks. Inspired by human behaviour, this paper presents a scalable hierarchical active inference model for autonomous navigation, exploration, and goal-oriented behaviour. The model uses visual observation and motion perception to combine curiosity-driven exploration with goal-oriented behaviour. Motion is planned using different levels of reasoning, i.e., from context to place to motion. This allows for efficient navigation in new spaces and rapid progress toward a target. By incorporating these human navigational strategies and their hierarchical representation of the environment, this model proposes a new solution for autonomous navigation and exploration. The approach is validated through simulations in a mini-grid environment. [ABSTRACT FROM AUTHOR]

This study investigated how a physical robot can adapt goal-directed actions in dynamically changing environments, in real-time, using an active inference-based approach with incremental learning from human tutoring examples. Using our active inference-based model, while good generalization can be achieved with appropriate parameters, when faced with sudden, large changes in the environment, a human may have to intervene to correct actions of the robot in order to reach the goal, as a caregiver might guide the hands of a child performing an unfamiliar task. In order for the robot to learn from the human tutor, we propose a new scheme to accomplish incremental learning from these proprioceptive–exteroceptive experiences combined with mental rehearsal of past experiences. Our experimental results demonstrate that using only a few tutoring examples, the robot using our model was able to significantly improve its performance on new tasks without catastrophic forgetting of previously learned tasks. [ABSTRACT FROM AUTHOR]

Taking inspiration from humans can help catalyse embodied AI solutions for important real-world applications. Current human-inspired tools include neuromorphic systems and the developmental approach to learning. However, this developmental neurorobotics approach is currently lacking important frameworks for human-like computation and learning. We propose that human-like computation is inherently embodied, with its interface to the world being neuromorphic, and its learning processes operating across different timescales. These constraints necessitate a unified framework: active inference, underpinned by the free energy principle (FEP). Herein, we describe theoretical and empirical support for leveraging this framework in embodied neuromorphic agents with autonomous mental development. We additionally outline current implementation approaches (including toolboxes) and challenges, and we provide suggestions for next steps to catalyse this important field.

This paper develops an outline for a hierarchically embedded architecture of an artificial agent that models human translation processes based on principles of active inference (AIF) and predictive processing (PP). AIF and PP posit that the mind constructs a model of the environment which guides behavior by continually generating and integrating predictions and sensory input. The proposed model of the translation agent consists of three processing strata: a sensorimotor layer, a cognitive layer, and a phenomenal layer. Each layer consists of a network of states and transitions that interact on different time scales. Following the AIF framework, states are conditioned on observations which may originate from the environment and/or the embedded processing layer, while transitions between states are conditioned on actions that implement plans to optimize goal-oriented behavior. The AIF agent aims at simulating the variation in translational behavior under various conditions and to facilitate investigating the underlying mental mechanisms. It provides a novel framework for generating and testing new hypotheses of the translating mind.

Karl Friston’s free-energy principle casts agents as self-evidencing through active inference. This implies that decision-making, planning and information-seeking are, in a generic sense, ‘wishful’. We take an interdisciplinary perspective on this perplexing aspect of the free-energy principle and unpack the epistemological implications of wishful thinking under the free-energy principle. We use this epistemic framing to discuss the emergence of biases for self-evidencing agents. In particular, we argue that this elucidates an optimism bias as a foundational tenet of self-evidencing. We allude to a historical precursor to some of these themes, interestingly found in Machiavelli’s oeuvre, to contextualise the universal optimism of the free-energy principle.

Understandings of how visual hallucinations appear have been highly influenced by generative approaches, in particular Friston’s Active Inference conceptualization. Their core proposition is that these phenomena occur when hallucinatory expectations outweigh actual sensory data. This imbalance occurs as the brain seeks to minimize informational free energy, a measure of the distance between predicted and actual sensory data in a stationary open system. We review this approach in the light of old and new information on the role of environmental factors in episodic hallucinations. In particular, we highlight the possible relationship of specific visual triggers to the onset and offset of some episodes. We use an analogy from phase transitions in physics to explore factors which might account for intermittent shifts between veridical and hallucinatory vision. In these triggered forms of hallucinations, we suggest that there is a transient disturbance in the normal one-to-one correspondence between a real object and the counterpart perception such that this correspondence becomes between the real object and a hallucination. Generative models propose that a lack of information transfer from the environment to the brain is one of the key features of hallucinations. In contrast, we submit that specific information transfer is required at onset and offset in these cases. We propose that this transient one-to-one correspondence between environment and hallucination is mediated more by aberrant discriminative than by generative inference. Discriminative inference can be conceptualized as a process for maximizing shared information between the environment and perception within a self-organizing nonstationary system. We suggest that generative inference plays the greater role in established hallucinations and in the persistence of individual hallucinatory episodes. We further explore whether thermodynamic free energy may be an additional factor in why hallucinations are temporary. Future empirical research could productively concentrate on three areas. Firstly, subjective perceptual changes and parallel variations in brain function during specific transitions between veridical and hallucinatory vision to inform models of how episodes occur. Secondly, systematic investigation of the links between environment and hallucination episodes to probe the role of information transfer in triggering transitions between veridical and hallucinatory vision. Finally, changes in hallucinatory episodes over time to elucidate the role of learning on phenomenology. These empirical data will allow the potential roles of different forms of inference in the stages of hallucinatory episodes to be elucidated.

Given the rapid advancement of artificial intelligence, understanding the foundations of intelligent behaviour is increasingly important. Active inference, regarded as a general theory of behaviour, offers a principled approach to probing the basis of sophistication in planning and decision-making. This paper examines two decision-making schemes in active inference based on “planning” and “learning from experience”. Furthermore, we also introduce a mixed model that navigates the data complexity trade-off between these strategies, leveraging the strengths of both to facilitate balanced decision-making. We evaluate our proposed model in a challenging grid-world scenario that requires adaptability from the agent. Additionally, our model provides the opportunity to analyse the evolution of various parameters, offering valuable insights and contributing to an explainable framework for intelligent decision-making.

Translation process research (TPR) has generated a large number of models that aim at explaining human translation processes. In this paper, I suggest an extension of the monitor model to incorporate aspects of relevance theory (RT) and to adopt the free energy principle (FEP) as a generative model to elucidate translational behaviour. The FEP—and its corollary, active inference—provide a general, mathematical framework to explain how organisms resist entropic erosion so as to remain within their phenotypic bounds. It posits that organisms reduce the gap between their expectations and observations by minimising a quantity called free energy. I map these concepts on the translation process and exemplify them with behavioural data. The analysis is based on the notion of translation units (TUs) which exhibit observable traces of the translator's epistemic and pragmatic engagement with their translation environment, (i.e., the text) that can be measured in terms of translation effort and effects. Sequences of TUs cluster into translation states (steady state, orientation, and hesitation). Drawing on active inference, sequences of translation states combine into translation policies that reduce expected free energy. I show how the notion of free energy is compatible with the concept of relevance, as developed in RT, and how essential concepts of the monitor model and RT can be formalised as deep temporal generative models that can be interpreted under a representationalist view, but also support a non-representationalist account. [ABSTRACT FROM AUTHOR]

New insights into intractable industrial challenges can be revealed by framing them in terms of natural science. One intractable industrial challenge is that creative production can be much more financially expensive and time consuming than standardized production. Creative products include a wide range of goods that have one or more original characteristics. The scaling up of creative production is hindered by high financial production costs and long production durations. In this paper, creative production is framed in terms of interactions between entropy and complexity during progressions from emergent pragmatics to action semantics. An analysis of interactions between entropy and complexity is provided that relates established practice in creative production to organizational survival in changing environments. The analysis in this paper is related to assembly theory, which is a recent theoretical development in natural science that addresses how open-ended generation of complex physical objects can emerge from selection in biology. Parallels between assembly practice in industrial production and assembly theory in natural science are explained through constructs that are common to both, such as assembly index. Overall, analyses reported in the paper reveal that interactions between entropy and complexity underlie intractable challenges in creative production, from the production of individual products to the survival of companies.

A theoretical account of development in mesocortical anatomy is derived from the free energy principle, operating in a neural field with both Hebbian and anti-Hebbian neural plasticity. An elementary structural unit is proposed, in which synaptic connections at mesoscale are arranged in paired patterns with mirror symmetry. Exchanges of synaptic flux in each pattern form coupled spatial eigenmodes, and the line of mirror reflection between the paired patterns operates as a Markov blanket, so that prediction errors in exchanges between the pairs are minimized. The theoretical analysis is then compared to the outcomes from a biological model of neocortical development, in which neuron precursors are selected by apoptosis for cell body and synaptic connections maximizing synchrony and also minimizing axonal length. It is shown that this model results in patterns of connection with the anticipated mirror symmetries, at micro-, meso- and inter-arial scales, among lateral connections, and in cortical depth. This explains the spatial organization and functional significance of neuron response preferences, and is compatible with the structural form of both columnar and noncolumnar cortex. Multi-way interactions of mirrored representations can provide a preliminary anatomically realistic model of cortical information processing.

In this paper, we unite concepts from Husserlian phenomenology, the active inference framework in theoretical biology, and category theory in mathematics to develop a comprehensive framework for understanding social action premised on shared goals. We begin with an overview of Husserlian phenomenology, focusing on aspects of inner time-consciousness, namely, retention, primal impression, and protention. We then review active inference as a formal approach to modeling agent behavior based on variational (approximate Bayesian) inference. Expanding upon Husserl’s model of time consciousness, we consider collective goal-directed behavior, emphasizing shared protentions among agents and their connection to the shared generative models of active inference. This integrated framework aims to formalize shared goals in terms of shared protentions, and thereby shed light on the emergence of group intentionality. Building on this foundation, we incorporate mathematical tools from category theory, in particular, sheaf and topos theory, to furnish a mathematical image of individual and group interactions within a stochastic environment. Specifically, we employ morphisms between polynomial representations of individual agent models, allowing predictions not only of their own behaviors but also those of other agents and environmental responses. Sheaf and topos theory facilitates the construction of coherent agent worldviews and provides a way of representing consensus or shared understanding. We explore the emergence of shared protentions, bridging the phenomenology of temporal structure, multi-agent active inference systems, and category theory. Shared protentions are highlighted as pivotal for coordination and achieving common objectives. We conclude by acknowledging the intricacies stemming from stochastic systems and uncertainties in realizing shared goals.

This study explains how the leader-follower relationship and turn-taking could develop in a dyadic imitative interaction by conducting robotic simulation experiments based on the free energy principle. Our prior study showed that introducing a parameter during the model training phase can determine leader and follower roles for subsequent imitative interactions. The parameter is defined as w , the so-called meta-prior, and is a weighting factor used to regulate the complexity term versus the accuracy term when minimizing the free energy. This can be read as sensory attenuation, in which the robot's prior beliefs about action are less sensitive to sensory evidence. The current extended study examines the possibility that the leader-follower relationship shifts depending on changes in w during the interaction phase. We identified a phase space structure with three distinct types of behavioral coordination using comprehensive simulation experiments with sweeps of w of both robots during the interaction. Ignoring behavior in which the robots follow their own intention was observed in the region in which both w s were set to large values. One robot leading, followed by the other robot was observed when one w was set larger and the other was set smaller. Spontaneous, random turn-taking between the leader and the follower was observed when both w s were set at smaller or intermediate values. Finally, we examined a case of slowly oscillating w in anti-phase between the two agents during the interaction. The simulation experiment resulted in turn-taking in which the leader-follower relationship switched during determined sequences, accompanied by periodic shifts of w s. An analysis using transfer entropy found that the direction of information flow between the two agents also shifted along with turn-taking. Herein, we discuss qualitative differences between random/spontaneous turn-taking and agreed-upon sequential turn-taking by reviewing both synthetic and empirical studies. [ABSTRACT FROM AUTHOR]

Robust evidence suggests that humans explore their environment using a combination of topological landmarks and coarse-grained path integration. This approach relies on identifiable environmental features (topological landmarks) in tandem with estimations of distance and direction (coarse-grained path integration) to construct cognitive maps of the surroundings. This cognitive map is believed to exhibit a hierarchical structure, allowing efficient planning when solving complex navigation tasks. Inspired by human behaviour, this paper presents a scalable hierarchical active inference model for autonomous navigation, exploration, and goal-oriented behaviour. The model uses visual observation and motion perception to combine curiosity-driven exploration with goal-oriented behaviour. Motion is planned using different levels of reasoning, i.e., from context to place to motion. This allows for efficient navigation in new spaces and rapid progress toward a target. By incorporating these human navigational strategies and their hierarchical representation of the environment, this model proposes a new solution for autonomous navigation and exploration. The approach is validated through simulations in a mini-grid environment.

This study investigated how a physical robot can adapt goal-directed actions in dynamically changing environments, in real-time, using an active inference-based approach with incremental learning from human tutoring examples. Using our active inference-based model, while good generalization can be achieved with appropriate parameters, when faced with sudden, large changes in the environment, a human may have to intervene to correct actions of the robot in order to reach the goal, as a caregiver might guide the hands of a child performing an unfamiliar task. In order for the robot to learn from the human tutor, we propose a new scheme to accomplish incremental learning from these proprioceptive–exteroceptive experiences combined with mental rehearsal of past experiences. Our experimental results demonstrate that using only a few tutoring examples, the robot using our model was able to significantly improve its performance on new tasks without catastrophic forgetting of previously learned tasks.

Translation process research (TPR) has generated a large number of models that aim at explaining human translation processes. In this paper, I suggest an extension of the monitor model to incorporate aspects of relevance theory (RT) and to adopt the free energy principle (FEP) as a generative model to elucidate translational behaviour. The FEP—and its corollary, active inference—provide a general, mathematical framework to explain how organisms resist entropic erosion so as to remain within their phenotypic bounds. It posits that organisms reduce the gap between their expectations and observations by minimising a quantity called free energy. I map these concepts on the translation process and exemplify them with behavioural data. The analysis is based on the notion of translation units (TUs) which exhibit observable traces of the translator’s epistemic and pragmatic engagement with their translation environment, (i.e., the text) that can be measured in terms of translation effort and effects. Sequences of TUs cluster into translation states (steady state, orientation, and hesitation). Drawing on active inference, sequences of translation states combine into translation policies that reduce expected free energy. I show how the notion of free energy is compatible with the concept of relevance, as developed in RT, and how essential concepts of the monitor model and RT can be formalised as deep temporal generative models that can be interpreted under a representationalist view, but also support a non-representationalist account.

Evolution is full of coevolving systems characterized by complex spatio-temporal interactions that lead to intertwined processes of adaptation. Yet, how adaptation across multiple levels of temporal scales and biological complexity is achieved remains unclear. Here, we formalize how evolutionary multi-scale processing underlying adaptation constitutes a form of metacognition flowing from definitions of metaprocessing in machine learning. We show (1) how the evolution of metacognitive systems can be expected when fitness landscapes vary on multiple time scales, and (2) how multiple time scales emerge during coevolutionary processes of sufficiently complex interactions. After defining a metaprocessor as a regulator with local memory, we prove that metacognition is more energetically efficient than purely object-level cognition when selection operates at multiple timescales in evolution. Furthermore, we show that existing modeling approaches to coadaptation and coevolution—here active inference networks, predator–prey interactions, coupled genetic algorithms, and generative adversarial networks—lead to multiple emergent timescales underlying forms of metacognition. Lastly, we show how coarse-grained structures emerge naturally in any resource-limited system, providing sufficient evidence for metacognitive systems to be a prevalent and vital component of (co-)evolution. Therefore, multi-scale processing is a necessary requirement for many evolutionary scenarios, leading to de facto metacognitive evolutionary outcomes. [ABSTRACT FROM AUTHOR]

The spread of ideas is a fundamental concern of today's news ecology. Understanding the dynamics of the spread of information and its co-option by interested parties is of critical importance. Research on this topic has shown that individuals tend to cluster in echo-chambers and are driven by confirmation bias. In this paper, we leverage the active inference framework to provide an in silico model of confirmation bias and its effect on echo-chamber formation. We build a model based on active inference, where agents tend to sample information in order to justify their own view of reality, which eventually leads to them to have a high degree of certainty about their own beliefs. We show that, once agents have reached a certain level of certainty about their beliefs, it becomes very difficult to get them to change their views. This system of self-confirming beliefs is upheld and reinforced by the evolving relationship between an agent's beliefs and observations, which over time will continue to provide evidence for their ingrained ideas about the world. The epistemic communities that are consolidated by these shared beliefs, in turn, tend to produce perceptions of reality that reinforce those shared beliefs. We provide an active inference account of this community formation mechanism. We postulate that agents are driven by the epistemic value that they obtain from sampling or observing the behaviours of other agents. Inspired by digital social networks like Twitter, we build a generative model in which agents generate observable social claims or posts (e.g., 'tweets') while reading the socially observable claims of other agents that lend support to one of two mutually exclusive abstract topics. Agents can choose which other agent they pay attention to at each timestep, and crucially who they attend to and what they choose to read influences their beliefs about the world. Agents also assess their local network's perspective, influencing which kinds of posts they expect to see other agents making. The model was built and simulated using the freely available Python package pymdp. The proposed active inference model can reproduce the formation of echo-chambers over social networks, and gives us insight into the cognitive processes that lead to this phenomenon. [ABSTRACT FROM AUTHOR]

We show that goal-directed action planning and generation in a teleological framework can be formulated by extending the active inference framework. The proposed model, which is built on a variational recurrent neural network model, is characterized by three essential features. These are that (1) goals can be specified for both static sensory states, e.g., for goal images to be reached and dynamic processes, e.g., for moving around an object, (2) the model cannot only generate goal-directed action plans, but can also understand goals through sensory observation, and (3) the model generates future action plans for given goals based on the best estimate of the current state, inferred from past sensory observations. The proposed model is evaluated by conducting experiments on a simulated mobile agent as well as on a real humanoid robot performing object manipulation. [ABSTRACT FROM AUTHOR]

This study explains how the leader-follower relationship and turn-taking could develop in a dyadic imitative interaction by conducting robotic simulation experiments based on the free energy principle. Our prior study showed that introducing a parameter during the model training phase can determine leader and follower roles for subsequent imitative interactions. The parameter is defined as w, the so-called meta-prior, and is a weighting factor used to regulate the complexity term versus the accuracy term when minimizing the free energy. This can be read as sensory attenuation, in which the robot’s prior beliefs about action are less sensitive to sensory evidence. The current extended study examines the possibility that the leader-follower relationship shifts depending on changes in w during the interaction phase. We identified a phase space structure with three distinct types of behavioral coordination using comprehensive simulation experiments with sweeps of w of both robots during the interaction. Ignoring behavior in which the robots follow their own intention was observed in the region in which both ws were set to large values. One robot leading, followed by the other robot was observed when one w was set larger and the other was set smaller. Spontaneous, random turn-taking between the leader and the follower was observed when both ws were set at smaller or intermediate values. Finally, we examined a case of slowly oscillating w in anti-phase between the two agents during the interaction. The simulation experiment resulted in turn-taking in which the leader-follower relationship switched during determined sequences, accompanied by periodic shifts of ws. An analysis using transfer entropy found that the direction of information flow between the two agents also shifted along with turn-taking. Herein, we discuss qualitative differences between random/spontaneous turn-taking and agreed-upon sequential turn-taking by reviewing both synthetic and empirical studies.

Evolution is full of coevolving systems characterized by complex spatio-temporal interactions that lead to intertwined processes of adaptation. Yet, how adaptation across multiple levels of temporal scales and biological complexity is achieved remains unclear. Here, we formalize how evolutionary multi-scale processing underlying adaptation constitutes a form of metacognition flowing from definitions of metaprocessing in machine learning. We show (1) how the evolution of metacognitive systems can be expected when fitness landscapes vary on multiple time scales, and (2) how multiple time scales emerge during coevolutionary processes of sufficiently complex interactions. After defining a metaprocessor as a regulator with local memory, we prove that metacognition is more energetically efficient than purely object-level cognition when selection operates at multiple timescales in evolution. Furthermore, we show that existing modeling approaches to coadaptation and coevolution—here active inference networks, predator–prey interactions, coupled genetic algorithms, and generative adversarial networks—lead to multiple emergent timescales underlying forms of metacognition. Lastly, we show how coarse-grained structures emerge naturally in any resource-limited system, providing sufficient evidence for metacognitive systems to be a prevalent and vital component of (co-)evolution. Therefore, multi-scale processing is a necessary requirement for many evolutionary scenarios, leading to de facto metacognitive evolutionary outcomes.

We show that goal-directed action planning and generation in a teleological framework can be formulated by extending the active inference framework. The proposed model, which is built on a variational recurrent neural network model, is characterized by three essential features. These are that (1) goals can be specified for both static sensory states, e.g., for goal images to be reached and dynamic processes, e.g., for moving around an object, (2) the model cannot only generate goal-directed action plans, but can also understand goals through sensory observation, and (3) the model generates future action plans for given goals based on the best estimate of the current state, inferred from past sensory observations. The proposed model is evaluated by conducting experiments on a simulated mobile agent as well as on a real humanoid robot performing object manipulation.

The spread of ideas is a fundamental concern of today’s news ecology. Understanding the dynamics of the spread of information and its co-option by interested parties is of critical importance. Research on this topic has shown that individuals tend to cluster in echo-chambers and are driven by confirmation bias. In this paper, we leverage the active inference framework to provide an in silico model of confirmation bias and its effect on echo-chamber formation. We build a model based on active inference, where agents tend to sample information in order to justify their own view of reality, which eventually leads to them to have a high degree of certainty about their own beliefs. We show that, once agents have reached a certain level of certainty about their beliefs, it becomes very difficult to get them to change their views. This system of self-confirming beliefs is upheld and reinforced by the evolving relationship between an agent’s beliefs and observations, which over time will continue to provide evidence for their ingrained ideas about the world. The epistemic communities that are consolidated by these shared beliefs, in turn, tend to produce perceptions of reality that reinforce those shared beliefs. We provide an active inference account of this community formation mechanism. We postulate that agents are driven by the epistemic value that they obtain from sampling or observing the behaviours of other agents. Inspired by digital social networks like Twitter, we build a generative model in which agents generate observable social claims or posts (e.g., ‘tweets’) while reading the socially observable claims of other agents that lend support to one of two mutually exclusive abstract topics. Agents can choose which other agent they pay attention to at each timestep, and crucially who they attend to and what they choose to read influences their beliefs about the world. Agents also assess their local network’s perspective, influencing which kinds of posts they expect to see other agents making. The model was built and simulated using the freely available Python package pymdp. The proposed active inference model can reproduce the formation of echo-chambers over social networks, and gives us insight into the cognitive processes that lead to this phenomenon.

The free energy principle, and its corollary active inference, constitute a bio-inspired theory that assumes biological agents act to remain in a restricted set of preferred states of the world, i.e., they minimize their free energy. Under this principle, biological agents learn a generative model of the world and plan actions in the future that will maintain the agent in an homeostatic state that satisfies its preferences. This framework lends itself to being realized in silico, as it comprehends important aspects that make it computationally affordable, such as variational inference and amortized planning. In this work, we investigate the tool of deep learning to design and realize artificial agents based on active inference, presenting a deep-learning oriented presentation of the free energy principle, surveying works that are relevant in both machine learning and active inference areas, and discussing the design choices that are involved in the implementation process. This manuscript probes newer perspectives for the active inference framework, grounding its theoretical aspects into more pragmatic affairs, offering a practical guide to active inference newcomers and a starting point for deep learning practitioners that would like to investigate implementations of the free energy principle.

The Sustainable Development Goals have been criticized for not providing sufficient balance between human well-being and environmental well-being. By contrast, joint agent-environment systems theory is focused on reciprocal synchronous generative development. The purpose of this paper is to extend this theory towards practical application in sustainable development projects. This purpose is fulfilled through three interrelated contributions. First, a practitioner description of the theory is provided. Then, the theory is extended through reference to research concerned with multilevel pragmatics, competing signals, commitment processes, technological mediation, and psychomotor functioning. In addition, the theory is related to human-driven biosocial-technical innovation through the example of digital twins for agroecological urban farming. Digital twins being digital models that mirror physical processes; that are connected to physical processes through, for example, sensors and actuators; and which carry out analyses of physical processes in order to improve their performance. Together, these contributions extend extant theory towards application for synchronous generative development that balances human well-being and environmental well-being. However, the practical examples in the paper indicate that counterproductive complexity can arise from situated entropy amidst biosocial-technical innovations: even when those innovations are compatible with synchronous generative development.

Active inference theory (AIT) is a corollary of the free-energy principle, which formalizes cognition of living system’s autopoietic organization. AIT comprises specialist terminology and mathematics used in theoretical neurobiology. Yet, active inference is common practice in human organizations, such as private companies, public institutions, and not-for-profits. Active inference encompasses three interrelated types of actions, which are carried out to minimize uncertainty about how organizations will survive. The three types of action are updating work beliefs, shifting work attention, and/or changing how work is performed. Accordingly, an alternative starting point for grasping active inference, rather than trying to understand AIT specialist terminology and mathematics, is to reflect upon lived experience. In other words, grasping active inference through autoethnographic research. In this short communication paper, accessing AIT through autoethnography is explained in terms of active inference in existing organizational practice (implicit active inference), new organizational methodologies that are informed by AIT (deliberative active inference), and combining implicit and deliberative active inference. In addition, these autoethnographic options for grasping AIT are related to generative learning.

Active Inference (AIF) is a framework that can be used both to describe information processing in naturally intelligent systems, such as the human brain, and to design synthetic intelligent systems (agents). In this paper we show that Expected Free Energy (EFE) minimisation, a core feature of the framework, does not lead to purposeful explorative behaviour in linear Gaussian dynamical systems. We provide a simple proof that, due to the specific construction used for the EFE, the terms responsible for the exploratory (epistemic) drive become constant in the case of linear Gaussian systems. This renders AIF equivalent to KL control. From a theoretical point of view this is an interesting result since it is generally assumed that EFE minimisation will always introduce an exploratory drive in AIF agents. While the full EFE objective does not lead to exploration in linear Gaussian dynamical systems, the principles of its construction can still be used to design objectives that include an epistemic drive. We provide an in-depth analysis of the mechanics behind the epistemic drive of AIF agents and show how to design objectives for linear Gaussian dynamical systems that do include an epistemic drive. Concretely, we show that focusing solely on epistemics and dispensing with goal-directed terms leads to a form of maximum entropy exploration that is heavily dependent on the type of control signals driving the system. Additive controls do not permit such exploration. From a practical point of view this is an important result since linear Gaussian dynamical systems with additive controls are an extensively used model class, encompassing for instance Linear Quadratic Gaussian controllers. On the other hand, linear Gaussian dynamical systems driven by multiplicative controls such as switching transition matrices do permit an exploratory drive.

In this paper, I investigate a connection between a common characterisation of freedom and how uncertainty is managed in a Bayesian hierarchical model. To do this, I consider a distributed factorization of a group’s optimization of free energy, in which each agent is attempting to align with the group and with its own model. I show how this can lead to equilibria for groups, defined by the capacity of the model being used, essentially how many different datasets it can handle. In particular, I show that there is a “sweet spot” in the capacity of a normal model in each agent’s decentralized optimization, and that this “sweet spot” corresponds to minimal free energy for the group. At the sweet spot, an agent can predict what the group will do and the group is not surprised by the agent. However, there is an asymmetry. A higher capacity model for an agent makes it harder for the individual to learn, as there are more parameters. Simultaneously, a higher capacity model for the group, implemented as a higher capacity model for each member agent, makes it easier for a group to integrate a new member. To optimize for a group of agents then requires one to make a trade-off in capacity, as each individual agent seeks to decrease capacity, but there is pressure from the group to increase capacity of all members. This pressure exists because as individual agent’s capacities are reduced, so too are their abilities to model other agents, and thereby to establish pro-social behavioural patterns. I then consider a basic two-level (dual process) Bayesian model of social reasoning and a set of three parameters of capacity that are required to implement such a model. Considering these three capacities as dependent elements in a free energy minimization for a group leads to a “sweet surface” in a three-dimensional space defining the triplet of parameters that each agent must use should they hope to minimize free energy as a group. Finally, I relate these three parameters to three notions of freedom and equality in human social organization, and postulate a correspondence between freedom and model capacity. That is, models with higher capacity, have more freedom as they can interact with more datasets.

Drawing from both enactivist and cognitivist perspectives on mind, I propose that explaining teleological phenomena may require reappraising both “Cartesian theaters” and mental homunculi in terms of embodied self-models (ESMs), understood as body maps with agentic properties, functioning as predictive-memory systems and cybernetic controllers. Quasi-homuncular ESMs are suggested to constitute a major organizing principle for neural architectures due to their initial and ongoing significance for solutions to inference problems in cognitive (and affective) development. Embodied experiences provide foundational lessons in learning curriculums in which agents explore increasingly challenging problem spaces, so answering an unresolved question in Bayesian cognitive science: what are biologically plausible mechanisms for equipping learners with sufficiently powerful inductive biases to adequately constrain inference spaces? Drawing on models from neurophysiology, psychology, and developmental robotics, I describe how embodiment provides fundamental sources of empirical priors (as reliably learnable posterior expectations). If ESMs play this kind of foundational role in cognitive development, then bidirectional linkages will be found between all sensory modalities and frontal-parietal control hierarchies, so infusing all senses with somatic-motoric properties, thereby structuring all perception by relevant affordances, so solving frame problems for embodied agents. Drawing upon the Free Energy Principle and Active Inference framework, I describe a particular mechanism for intentional action selection via consciously imagined (and explicitly represented) goal realization, where contrasts between desired and present states influence ongoing policy selection via predictive coding mechanisms and backward-chained imaginings (as self-realizing predictions). This embodied developmental legacy suggests a mechanism by which imaginings can be intentionally shaped by (internalized) partially-expressed motor acts, so providing means of agentic control for attention, working memory, imagination, and behavior. I further describe the nature(s) of mental causation and self-control, and also provide an account of readiness potentials in Libet paradigms wherein conscious intentions shape causal streams leading to enaction. Finally, I provide neurophenomenological handlings of prototypical qualia including pleasure, pain, and desire in terms of self-annihilating free energy gradients via quasi-synesthetic interoceptive active inference. In brief, this manuscript is intended to illustrate how radically embodied minds may create foundations for intelligence (as capacity for learning and inference), consciousness (as somatically-grounded self-world modeling), and will (as deployment of predictive models for enacting valued goals).

Psychomotor experience can be based on what people predict they will experience, rather than on sensory inputs. It has been argued that disconnects between human experience and sensory inputs can be addressed better through further development of predictive processing theory. In this paper, the scope of predictive processing theory is extended through three developments. First, by going beyond previous studies that have encompassed embodied cognition but have not addressed some fundamental aspects of psychomotor functioning. Second, by proposing a scientific basis for explaining predictive processing that spans objective neuroscience and subjective experience. Third, by providing an explanation of predictive processing that can be incorporated into the planning and operation of systems involving robots and other new technologies. This is necessary because such systems are becoming increasingly common and move us farther away from the hunter-gatherer lifestyles within which our psychomotor functioning evolved. For example, beliefs that workplace robots are threatening can generate anxiety, while wearing hardware, such as augmented reality headsets and exoskeletons, can impede the natural functioning of psychomotor systems. The primary contribution of the paper is the introduction of a new formulation of hierarchical predictive processing that is focused on psychomotor functioning.

Collective intelligence, an emergent phenomenon in which a composite system of multiple interacting agents performs at levels greater than the sum of its parts, has long compelled research efforts in social and behavioral sciences. To date, however, formal models of collective intelligence have lacked a plausible mathematical description of the relationship between local-scale interactions between autonomous sub-system components (individuals) and global-scale behavior of the composite system (the collective). In this paper we use the Active Inference Formulation (AIF), a framework for explaining the behavior of any non-equilibrium steady state system at any scale, to posit a minimal agent-based model that simulates the relationship between local individual-level interaction and collective intelligence. We explore the effects of providing baseline AIF agents (Model 1) with specific cognitive capabilities: Theory of Mind (Model 2), Goal Alignment (Model 3), and Theory of Mind with Goal Alignment (Model 4). These stepwise transitions in sophistication of cognitive ability are motivated by the types of advancements plausibly required for an AIF agent to persist and flourish in an environment populated by other highly autonomous AIF agents, and have also recently been shown to map naturally to canonical steps in human cognitive ability. Illustrative results show that stepwise cognitive transitions increase system performance by providing complementary mechanisms for alignment between agents’ local and global optima. Alignment emerges endogenously from the dynamics of interacting AIF agents themselves, rather than being imposed exogenously by incentives to agents’ behaviors (contra existing computational models of collective intelligence) or top-down priors for collective behavior (contra existing multiscale simulations of AIF). These results shed light on the types of generic information-theoretic patterns conducive to collective intelligence in human and other complex adaptive systems.

Active inference is a normative framework for explaining behaviour under the free energy principle—a theory of self-organisation originating in neuroscience. It specifies neuronal dynamics for state-estimation in terms of a descent on (variational) free energy—a measure of the fit between an internal (generative) model and sensory observations. The free energy gradient is a prediction error—plausibly encoded in the average membrane potentials of neuronal populations. Conversely, the expected probability of a state can be expressed in terms of neuronal firing rates. We show that this is consistent with current models of neuronal dynamics and establish face validity by synthesising plausible electrophysiological responses. We then show that these neuronal dynamics approximate natural gradient descent, a well-known optimisation algorithm from information geometry that follows the steepest descent of the objective in information space. We compare the information length of belief updating in both schemes, a measure of the distance travelled in information space that has a direct interpretation in terms of metabolic cost. We show that neural dynamics under active inference are metabolically efficient and suggest that neural representations in biological agents may evolve by approximating steepest descent in information space towards the point of optimal inference.