Active inference and collective intelligence

15th-17th October 2021, Slideslive and Zoom

Image by Fakurian Design

What is Cognitio all about?

Changing the way collective intelligence is understood

The main topic of the workshop is Active Inference as physics-based models of cultural phenomena, and manners of understanding human behaviour from the social sciences and humanities. This workshop will be important because cross-talk and cross-fertilization between these fields is becoming unavoidable and necessary. Digital technologies are rapidly evolving, and powerful tools have been used in the past for social purposes, without relying on the perspectives and caution of social science experts.


As a consequence of this neglect, critical flaws in the design of such computational tools have led to furthering the oppression of vulnerable groups. One notable instance of this has been the use of artificial intelligences to help in the determination of guilt in trials. The US legal system, based on historically racist outcomes and practices, has only prolonged the “school to prison” pipeline, in which law enforcement largely disproportionately targets Black and Latino men in the states. The systems that perform such legal advisership tend to operate opaquely like black boxes, and as such, realizing that this bias held such way took the community a long time—too long. 

One of the main attributes of active inference in machine learning is that it is transparent: the models at play are fully explicit and specify the causal factors that generate data. This is an incredible advantage over black box networks that can lead to the perpetuation of social problems, when they draw from (e.g.) historical data. Working from the perspective that says “nothing about us without us,” we wish to foster collaboration between the fields of social science, computational modelling, and neuroscience, to further our ability to avoid such pitfalls and to develop the proper tools to benefit such vulnerable communities.

We will reach out to researchers in the field who specifically target social issues, or who make a contribution that focuses on social progress. We will focus on researchers from a variety of fields, and show an interest in an intersectional perspective.

Posters will also be presented, for students who may have smaller contributions or preliminary findings.

This workshop will be held online to promote safety and the international participation of actors across the world, as active inference is still a relatively new concept. Furthermore, we believe that access to such conferences may be rendered easier for participants with low financial means, or physical disabilities.

The different presentations will be collated in a special edition, that will become available for anyone who may want to read the subsequent talks, and continue of the social progress path. 

Image by Matthias Wagner



Garrido 1500px 14.jpg

Dr. Marta Garrido

University of Melbourne, Australia

Associate Professor Marta Garrido leads the Cognitive Neuroscience and Computational Psychiatry Laboratory at the Melbourne School of Psychological Sciences at the University of Melbourne and is Chief Investigator in the Australian Research Council Centre of Excellence
for Integrative Brain Function. Marta initially trained in Engineering Physics at the University of Lisbon and then received her PhD in Neuroscience from University College London in 2008.
She completed postdocs at University California Los Angeles and back at the FIL. In 2013 she moved to the Queensland Brain Institute on a Discovery Early Career Researcher Award and later established her independent laboratory. In mid 2019 the lab moved to the University of
Melbourne. The main goal of the research group is to understand how the brain learns and makes predictions about future events while adapting to the contingencies of novel environments.
Along with our work on typical cognition in typical human individuals, our mission is to contribute to the understanding of mental illness, in particular those conditions where predictive processes and brain circuitry are disrupted such as in schizophrenia and anxiety.
To pursue this endeavour, we use a combination of computational modelling, machine learning and brain imaging techniques such as magnetoencephalography (MEG),
electroencephalography (EEG), and magnetic resonance imaging (MRI).

Dr. Kelsey Perrykkad

Cognition and Philosophy Lab,

School of Philosophical, Historical and International Studies, Faculty of Arts,

Monash University

Kelsey Perrykkad is an interdisciplinary cognitive scientist, working at the intersection of experimental psychology, neuroscience, philosophy and education. Much of Kelsey’s recent work has focused on self-cognition in autism spectrum conditions from the perspective of predictive processing. She is allistic (i.e. not autistic). Kelsey currently works as a Post Doctoral Research Fellow in the Cognition and Philosophy Lab at Monash University, Melbourne, Australia.

2021-06-17 13.53.19.jpg

Dr. Jenny Poulton

Imperial College London

Foundation for Fundamental Research on Matter (FOM) Institute for Atomic and Molecular Physics (AMOLF)

Jenny did her undergraduate degree in theoretical physics at the University of Sheffield, and received her PhD from Imperial College London in the stochastic thermodynamics of biological copying processes in 2021, supervised by Dr Thomas Ouldridge. She has since been working on the “Intelligence in context” FQXI grant with Professor Chris Watkins at Royal Holloway, University of London, and has recently moved to Amsterdam to start working at AMOLF with Professor Pieter Rein Ten Wolde. She has started a regular seminar series and support group for “junior” women (post-docs, PhD students, masters students and undergraduates) who are interested in Stochastic Thermodynamics. This tight knit community of around 25 women get to network directly with senior academics, collaborate with each other and give each other support in countless other ways. If you are interested in attending this group, please email


Rising Stars


Mel Andrews

University of Cincinnati, Department of Philosophy

Mel Andrews is an instructor and doctoral student in the department of philosophy, University of Cincinnati. Their work explores the phenomena of cognition (anticipatory dynamics) and life (far from thermodynamic equilibrium steady-state dynamics). They compare and contrast the merits and explanatory scope of conceptual and formal models of life and mind, and explore the implications of these considerations for some of the major outstanding questions in the cognitive and biological sciences and in the philosophy of mind. Their research principally concerns the free energy principle, a mathematical modelling framework for brain, behaviour, and biology which draws from physics and machine learning. 

Shannon Proksch

University of California

Shannon Proksch is a Ph.D. Candidate in Cognitive and Information Sciences at the University of California, Merced. She received her B.A. in Music from Texas A&M University, Corpus Christi in 2015, and her M.Sc. in Mind, Language and Embodied Cognition from the University of Edinburgh in 2017. Her research blends empirical and philosophical methods from the dynamical systems and predictive processing frameworks to examine the neural, behavioral, and social dynamics of music cognition—from lower-level beat processing to higher-level coordination and social interaction—using tools such as transcranial magnetic stimulation (TMS), motion capture, acoustic analysis, and computational techniques. Her work is supported, in part, by a National Science Foundation Traineeship in Intelligent and Adaptive Systems. 


kate nave.png

Kate Nave

University of Edinburgh

Kate is a PhD student on the X-SPECT project, led by Andy Clark, which aims to investigate the nature of conscious experience in the predictive brain. Her research focusses on identifying the principles underlying the self-producing activity distinctive of living organisms, and whether these can be scaled up to explain the origin and structure of conscious experience.


Glass Buildings


Mahault is a PhD student in cognitive computing, and has a masters in Sexology from the university of UQAM. Her project concerns the way active inference can help us model epistemic communities.
She strives to integrate anti-oppressive approaches to her endeavours, technological or social.



Ines Hipolito

Ines works on the intersection between philosophy of mind and computational cognitive neuroscience. 
​More specifically, she aims to elucidate how we understand cognition in everyday life in a way that is consistent with using the technical apparatus of dynamical systems theory.

Pierre Poirier is a research professor in the philosophy of cognitive science at the Université du Québec à Montréal. His areas of interest are the philosophy of science, the philosophy of neuroscience and the philosophy of mind. He has long been interested in enactive, embodied and situated cognition.




Introduction - Pierre Poirier

Kate Nave - What makes free energy minimisation an ideal theory of living systems?

Ines Hipolito. Cognitive biases in data science

Itzel Cadena Alvear and Melina Gastelum Vargas. An integrative view of collective action for human and social development

Troy Weekes and Thomas Eskridge. Personal Flow and Effortless Attention in Knowledge Work using Active Inference

Kelsey Perrykkad


Mark Miller, Daphne Demekas and Ben White. Artificial empathy: active inference and a new collective intelligence

Jenny Poulton

Avel Guénin-Carlut, Daniel Friedman and Virginia Bleu Knight. Thinking like a State : Active inference and the deep roots of complex societies

Alejandra Ciria, Mahault Albarracin, Mark Miller and Bruno Lara. Social Media Platforms: Trading Prediction Error Minimization for your Attention

Panel Discussion: true transparency in modeling and data processing. Panelists: Daniela Cialfi, Maxwell Ramstead, Julio Alcantara

Mel Andrews

Julio Alcántara. On Pure Consciousness of Autopoietic Machines

Mahault Albarracin: Affordance negotiation: The case of gender and its fluidity

I. A. Roland-Rodríguez. A Computational Phenomenology of Enmity and Normativity from Husserl to Fanon


Avel Guénin-Carlut. Intelligence without creativity : can Active Inference ground our understanding of life, cognition and society

Shannon Proksch

Amit Singh. Evolution of Latent Model for Collective Cognition

Rémi Tison. Active inference, mindshaping and the "fanciest" sort of intentionality

Panel Discussion: the role of diversity and inclusion in research design. Panelists: Daniela Cialfi, Jonas Mago, Maxwell Ramstead

Natalie Kastel and Casper Hesp. Ideas Worth Spreading: A Free Energy Proposal For Cumulative Cultural Dynamics

Sergio Rubin. Planetary cognition is autonomous in relation to the collective intelligence of the biosphere

Jonas Mago, George Deane, Michael Lifshitz and Maxwell Ramstead. Dissolving and consolidating selves in an active inference agent: The multiplicity of agency and selfhood in the human brain

Marta Garrido



Image by Brands&People

Naturalizing Folk Psychology: an Active Inference Account

Folk psychology” refers to the systems of language and inferential strategies used by laypersons to understand, explain, and predict the mental states and behaviour of each other. A position known as “eliminative materialism” suggests that such everyday concepts have no role to play in scientific research due to their lack of grounding in the material reality of neuro-cognitive processes. We argue on the contrary that integrating folk intuition and academic knowledge is a key element of research, and that Active Inference provides both grounding for this argument and a practical way to build such integration. This framework indeed accentuates the role of active attempts by organisms to model and navigate their environment in cognition. Those attempts must themselves be grounded in tentative models underlying action and integration of surprising observation, and cannot therefore map on passive and faithful representations. The prevalence of folk language in research, rather than indicating a failure to naturalize scientific discourse, may therefore demonstrate the scaffolding role of everyday intuition in building scientific knowledge. The language of Active Inference also maps well with the intuitive concepts of belief, desire, and intention. By giving them a very specific formal sense amenable to testing and prediction, it provides an opportunity to ground and sharpen our intuitive comprehension of the world rather than suppressing it. Its highly interdisciplinary and parsimonious approach makes it an excellent a priori ontology for living systems, both in the context of scientific enquiry and of everyday interaction. Because of its depth and transversality, Active Inference can become the keystone of an unprecedented integration of theory and practice in our understanding of the world. What is lacking at the moment is a pragmatic approach to leverage its ontology to couple the concepts of everyday life, the scientific and philosophical realms. We propose here a first attempt to articulate this mapping by drawing an Active Inference account of the role folk psychology has in the scientific study of cognition.

Image by Alexander Krivitskiy

Personal Flow and Effortless Attention in Knowledge Work using Active Inference

Knowledge work often involves unfamiliar experiences with goal-relevant percepts and rules to learn new information and create novel concepts. Attention is a prerequisite for practical knowledge work and generally requires significant cognitive effort to focus and sustain over time. The challenge of knowledge work is driven by the novelty and complexity of the task and correlates with the subjective enjoyment of the activity. Humans perceive flow as the optimal state of awareness during which task performance maximizes and self-awareness minimizes. Research studies on flow to date have a significant disagreement regarding what flow is and how to measure it. This paper proposes a formal human-task-context performance model of flow that integrates attention, surprise, and enjoyment to measure flow using active inference and Markov Decision Processes. We administered a cross-sectional questionnaire with knowledge workers to obtain priors for our Bayesian model, capturing evidence about the flow components to make inferences about knowledge work performance. Our hypothesis states that when the human knowledge worker experiences flow at work, their ability to focus and sustain attention on the task is maximized, which minimizes their perceived ambiguity and stress, thereby resulting in effortless attention, which perpetuates flow.

Image by Olenka Kotyk

The physical grounding of cognition : criticality, enactivism, and active inference

Integrative principles provide formal ontologies for specific classes of target systems, allowing both intuitive legibility and the generation and assessment o f possible mechanical explanations of their behavior. As cognitive science developed around the causal-mechanical characterization of the physical architecture underlying cognition, it naturally elicited proposals to ground its study in some underlying physical principle. Here we discuss here the value of Free Energy Principle / Active Inference (FEP/AI) as an integrative principle for the study of the multiscale dynamics underlying cognition.

Indeed, the brain’s capacity to process information relies on connectivity patterns that generate endogenous scale-invariant fluctuations in neuronal activity. These multiscale oscillations allow for non-trivial stimuli susceptibility and coordinated organismal reaction. The “Critical Brain Hypothesis” (CBH) claims such organisation is explained by the brain’s self-organization near a so-called “critical” transition in mean effective connectivity. Altough the brain does show phenomenological similarity to the critical transition in condensed matter, there is no widely-accepted model of mechanisms of brain criticality, and current accounts do not convey how the relevant phase space for system behavior is generated.

We argue FEP/AI significantly improves on this approach by explicitly integrating semantical considerations (what meaning the system embodies) and ecological principles with the syntactic study of system organisation which CBH emphasize. Indeed, FEP/AI highlights the enactive conception of cognition as skilled agency/structural autonomy, and accentuates the role of niche co-construction in the multiscale structures of living systems. Consequently, FEP/AI provides both a non-reductive grounding of intelligent behavior in its underlying physical processes, and a meaningful conceptualization of collective action in ecological systems.

We propose here to articulate a discussion of the status of FEP/AI both as an integrative principle and a grounding principle for the study of cognition and life. Our discussion will be complemented by a computational analysis based on the Inferants model aiming to investigate the conditions in which multiscale active inference reproduces the phenomenology of criticality in swarm systems, and what is its role in the emergence of collective intelligence.

Image by Johannes Plenio

Active inference, evolutionary transition, and the deep roots of complex societies

We will hereby discuss the evolutionary transition underlying the Neolithic transition in terms of the development of a cultural ecosystem of complementary enactive niches organised at nested scales. The goal of this discussion is to demonstrate FEP/AI’s ability to conceptually integrate established frameworks in life and social sciences, while maintaining intuitive tractability across scales.

We suggest the progressive complexification (in the institutional sense) of human societies throughout deep history was driven by the development of an ecosystem of complementary cultural niche, defined by niche modification and coevolving regimes of expectations. Our multiscale FEP/AI account integrates the emergence of the structure underlying City-States at both inter-personal (economic specialisation, class division) and structural (infrastructural niche construction, centralised administration) scales, and therefore grounds the interpretation of existing social scientific accounts as an evolutionary transition stricto sensu.

Codependency patterns between City metabolism and State agency incite us to understand the emerging organism as an enactive system. This is possible through the FEP/AI account of niche construction trough active inference (building affordances into the world so as to maximise one’s own embodied model evidence), which provides a natural framework to formalise State’s attempts at legibilising social life both as a cognitive process and as a prolongation of social metabolism. We hereby address recent objections to the FEP/AI account of cultural evolution by demonstrating it integrates aspects of cultural evolution evoked yet unformalised by mainstream cultural evolution literature.

The bulk of our work is to account for the dynamic interaction of cultural niches across scales, i.e. for the way socio-professional classes and institutions shape each other's niches and expectations about the world. Although State top-down legalisation may be the most obvious example of niche canalisation in human history, bottom-up growth of modern capitalist ethos and institutions uncontroversially account for the most radical changes in recent human history. We argue an explicit cross-scales FEP/AI analysis of such niche construction dynamics would map the way for a transition beyond the political domination (and resulting extractive growth and instability) characterising the early Anthropocene.

Image by GR Stocks

A Computational Phenomenology of Enmity and Normativity from Husserl to Fanon

Common clichés of the Black thief lingering in the night, preconditioned by great alarm and surprise, actually normatize a systemic enmity against Black mobility, preconditioning actual bodily threat. This actual threat mimes a similar predictive system not of a preemptive enmity but rather of expecting enmity. Fanon articulates how the Black psyche, confronted with the white world, comes to anticipate being perceived as threatening.3 The Bayesian nature of Fanon’s phenomenology becomes explicit when he writes that black bodies are constituted as entropic within a white world: “the body is surrounded by an atmosphere of certain uncertainty” . Computational phenomenology can formalize this normative enmity. Returning us to phenomenology, especially computationally, allows us to account for the systems or ecologies of chains of deferrals of given objects within an already constituted world of use and meaning. ssssssssSoon Extending the Husserlian parsing of social scripts from Albarracin (2021) to Ramstead (2021) allows us to read the passive syntheses of the Husserlian constitution of objects of experience as involving a process of weak scripts that are internalist and externalist, and weak in that they pre-condition perception. Internalist scripts synthesize well with phenomenologist Mahendran who theorizes the facticity of Blackness, not in terms of a social construction of racial meaning, but as a phenomenological constitution of racial meaning (via a white world’s anti-Black “passive syntheses,” or “ scripts”).

Image by David Matos

Dissolving and consolidating selves in an active inference agent: The multiplicity of agency and selfhood in the human brain

The first-person experience of humans is often cast as being accompanied and underwritten by a unified sense of agency, as if there were a singular, distinct agent (our “self”) directing all our thoughts and actions. However, the self can seem to dissolve during practices such as meditation or under the influence of psychedelic substances, and there are also experiences, such as voice-hearing or possession, where multiple agents seem to inhabit a single body. Such experiences of multiple agency are more prevalent than widely assumed, even among healthy people with no need for clinical care, and some practices even deliberately aim to cultivate such experiences. This spectrum of self-experience, from unified to absent or multiple, is reflected in cultural traditions around the world. For example, through charismatic prayer practice, healthy (non-psychotic) people can learn to hear voices, which are experienced as internal mental events (thoughts) that feel as if they are coming from an external agentic source (God). In this paper, we develop and test a novel computational model to explain such experiences of multiple agency in the human brain, using generative modelling and the active inference framework. In active inference, selves are modelled as inferred causes: they are the best explanation for some of the variance in our sensory data or subjective experience. We will develop a generative model that includes a variable number of self-agents as possible causes for sensory impressions. This provides a mechanistic understanding of how a single brain system can support multiple parallel agentic trajectories which may be experienced as a plurality of selves inhabiting a single body/mind. Furthermore, we will deploy this model to shed light on experiences involving an absence of a sense of self. Finally, we will discuss how these basic embodied experiences of the multiplicity or absence of selfhood may experientially support specific ontologies of agency in terms of persons, spirits, or invisible forces, which can be culturally elaborated through narrative accounts of self and others. Studying these a-typical planes of phenomenological selfhood will help us to better understand the computational mechanisms that give rise to both the absence and multiplicity, as well as the unified sense of agency.

Image by Melissa Askew

Evolution of Latent Model for Collective Cognition

Humans are known to share commonalities for collectively intelligent behaviour to emerge. Intelligence is fundamentally the ability for an agent to infer causal dependencies in its environment and change its World Model. However, the precise conceptualization across systems and scales is a polemical question. The concept of “Intelligence” may as well refer to a quantitative measure of formal cognitive ability than to a qualitative property of the skilled agency. This difficulty in understanding the concept compounds when we try to scale to descriptive and predictive models of collective behavior. While it is self-evident that groups may leverage pairwise interactions or their collective resources to tackle complex problems, is that process only the sum of individual intelligence, or is the group intelligent in its own right? If the latter, what does it mean for the classical internalist conception of intelligence and agency? If the former, then what is the proper scale of analysis in systems of nested organization, such as human societies? This question can be approached rigorously through a non-reductive account of the physical processes underlying intelligence. Here I propose that the latent model framework (with active inference as intrinsic reward mechanism) framework is a promising approach that could live up to the multiple dimensions of adeptness required by any framework that would attempt to generalize cognition across scales. A statistical state model for mathematical state transitions can be built and can be used to further define cognitive model.


Social Media Platforms: Trading Prediction Error Minimization for your Attention

Culture exploits the acquisition of meaningful content by crafting regimes of shared attention, determining what is relevant, valuable, and salient. Culture changes the field of relevant social affordances worthy of being acted upon in a context-sensitive manner. When relevant affordances are highly weighted, their salience increases the probability of it being enacted, and as a consequence, their associated prediction error is minimized. This process is known as active inference. In the digital era, individuals need to infer the action-related attributes of digital cues, here characterized as digital affordances. The digital affordances of social media platforms are of particular interest here. By their own nature, these are salient because they are related to social interactions and relevant social cues. However, the problem around social network platforms is that they are not equivalent to situated social interactions because their structure is built, mediated, and defined by third-parties with diverse interests. The third-parties behind the social media platforms are using the same mechanism exploited by culture to manipulate the shared patterns of attention. Moreover, these social media platforms are deliberately designed to be hyper-stimulating, making them dangerously rewarding and increasingly addictive. As we will show, the outcome is a growing risk of our online social interactions disrupting our long-term adaptivity. This appropriation, for economic purposes, is an issue of greater importance, especially as the COVID-19 pandemic brought deep global changes pushing societies to an online digital way of life. In this paper, we examine digital instrumental actions as well as digital epistemic actions afforded by social media in light of the prediction error dynamics they might elicit to their users. This paper aims to analyze, under the active inference framework, how the field of relevant affordances is changing as a product of the use of social network platforms. Specifically, how social network platforms are changing the patterns of attention, affecting the way beliefs are updated, how social norms are learned, and how self-identity is built. Changes in the field of relevant affordances as a product of economic purposes may be putting at risk our context-sensitive grip on a rich, dynamic and varied field of relevant affordances.

Image by Annie Spratt

Artificial empathy: active inference and a new collective intelligence

Affective computing is an exciting new research program aimed at developing computer and robotic systems capable of recognizing and responding intelligently to human emotions (Demekas et al. 2020). So far, the program has focused on refining forms of passive pattern recognition - deep-learning trained on emotional image data. SO, while many of these systems have bodies they are not meaningfully embodied - their bodies play no role in the computational work of understanding emotions.

In this paper, we propose a more embodied starting place for the affective computing program: the active inference framework (AIF). According to the AIF, the embodied brain's dynamics can be functionally expressed as an internal (generative) model making inferences about sensory information. Perception and action are both explainable as attempts to reconcile the discrepancies between the generative model and these signals - either by updating the model (perception) or by changing the world (action) (Clark 2016). According to this framework, to understand what someone is feeling is to have a generative model that is attuned to the other in ways where useful predictions can be made about the underlying causes (e.g., feeling happy) of the sensory evidence (e.g., smiling). This feat is accomplished not only by passively perceiving the other’s emotional expressions, but by actively probing, eliciting and confirming one’s predictions as a form of embodied epistemic foraging (Parr & Friston 2017).

We will argue that the AIF offers researchers from neighboring fields interested in affective computing a shared computational currency capable of both modelling the complex dynamics underlying human emotional understanding and social interaction and providing a solid basis for simulating those dynamics in computer systems (Linson et al. 2018). The outcome, we suggest, will be artifacts capable of constructing for themselves useful emotional lexicons (a major challenge in affective computing) by actively engaging, sampling and updating predictions about emotional responses against their user’s behaviour in real time (Friston & Frith, 2015). We will conclude by discussing the possibility that uncertainty-minimizing artifacts and their human users may converge overtime upon a shared generative model of the world (see Friston et al. 2020), which would in turn allow for new forms of human-computer collective intelligence to emerge, and new approaches to transparency and value alignments between humans and artifacts to begin.

Image by Drew Graham

On Pure Consciousness of Autopoietic Machines

This research project will experiment with the simplest form of conscious experience humans are capable of or minimal phenomenal experience (MPE), firstly, to prove its empirical existence, secondly, to disentangle its core causal factors, and thirdly, to attest its content-properties. Deleting and maximizing the unit of identification (UI) are the two channels to analyze empirically the MPE within the model of the Ascending Reticular Activation System (ARAS), the minimal form of phenomenal experience and the minimal phenomenal selfhood (MPS), mainly the thresholds of electrical frequencies of the brain during the analyzed states, their correlations and simulations, and subjective reports. Does it exist an all pervading form of conscious experience? Which phenomenal characteristics does pure consciousness (PC) entail? And based on this, can a minimal model for conscious experience as such be developed? This research tries to overcome ideological contaminations by knowing the metaphysical implications of its propositions and their relation to a wider scientific context based on intuition-free experimental philosophy, in which two experiments are proposed in order to study consciousness as such, firstly, the confirmation of MPE during lucid deep sleep through listening to different musical patterns, and secondly, a controlled hallucination in Virtual Reality (VR) with the maximization of the UI.

Image by Uday Mittal

Active inference, mindshaping and the "fanciest" sort of intentionality

In this communication, I sketch an active inference account of content based on Brandom’s pragmatist approach. It is commonplace to note that ecological, enactive and broadly dynamical approaches to cognition reject the idea that the notion of content should play a prominent or foundational role in our explanation of cognitive processes. Accordingly, ecological-enactive interpretations of the active inference framework (Bruineberg and Rietveld, 2014; Bruineberg et al, 2018) also aim to eschew any appeal to the notion of content. However, it seems that this notion constitutes an important part of our folk-psychological and linguistic practices, that is, of the way we understand each other in our daily social interactions. When we say that somebody believes that so and so is the case, or that somebody’s utterance means so and so, we specify the content of this belief or this utterance. What do we do when we do such a thing? Even ecological-enactive interpretations of active inference must eventually provide an explanation of such content-involving practices, which Brandom refers to as the “fanciest” sort of intentionality. According to an important tradition (Sellars, 1956; Kripke, 1982), accounting for the notion of content begins with the recognition that content is normative: the state or performance of an individual has a content insofar as it has a normative status, which specifies a pattern of entitlements and commitments attributed to the individual (Brandom, 1994). Such normative statuses are often taken to be instituted by the normative attitudes of the members of the community to which belongs the interpreted individual. The suggestion developed in this communication is that we can understand these normative attitudes in the active inference framework as a particular type of regime of expectations (Ramstead et al., 2016; Constant et al., 2019) having the function of regulating the behavior of community members, in turn facilitating the coordination of joint projects. Attributing a content to a mental state or a meaning to an utterance is therefore a form of “mindshaping” (Zawidzki, 2013; Fenici and Zawidzki, 2020), constraining behavior according to social norms. Accordingly, the function of content attribution is not to predict behavior, but rather to make behavior predictable.

Image by mostafa meraji

An integrative view of collective action for human and social development

In order to create a unified theory of cognition and collective action, we propose the convergence between enactive and ecological approaches by drawing together the relational understanding of affordances 2.0, ecological resonance, coordination dynamics and the sensorimotor approach. We conceive cognitive agents as sensorimotor systems whose perception-action occurs in terms of affordances in continuous constitutive interaction with the environment. This framework provides an integrative view of cognition considering affectivity, exploration and interaction within a complex and dynamical dimension, creating different affordance spaces. We argue that interacting behaviors in collective action involve sensorimotor patterns that generate particular affective discharges and relations with the material environment and others. We propose that the particular intentionality of skilful and meaningful engagement with the environment will be characterized given the agents’ sensorimotor system, its development and individual history (umwelt) and their set of possible affordances given a shared form of life (habitat). We argue that this skillful engagement with affordances will be influenced by the normativity imposed by the sociomateriality of the situation, and so will the networks of sensorimotor schemes. A physical model to explain this proposal could be found in ecological resonance. Accordingly, we propose that different approaches to collective action can be reconceptualized in real time social interactions within this approach, arguing that some assumptions on active inference contradict in particular the historicity of agency and sense-making that is explicit in the enactive-ecological approach. The critical view of theoretical and methodological paradigms within cognitive science could help us to create an integrative comprehension of different ways of understanding the physical and social world, such as the case of autism. Thus, by creating an interdisciplinary bridge exploring the dynamics of intersubjective and ecological resonance we can elaborate more comprehensive, integrative and ethical applied interventions leading to rethink human and social development in general.

Image by Riccardo Annandale

Ideas Worth Spreading: A Free Energy Proposal For Cumulative Cultural Dynamics

The dynamics underlying cultural evolution include the introduction of novel beliefs and practices to a culture (i.e., innovation), the transmission of established beliefs and practices within a population (i.e., innovation diffusion), and its change in prevalence (i.e., global cultural dynamics). While there is a fast growing body of theoretical and empirical work on characterising these dynamics, mathematical models able to integrate this data into quantifiable models are scarce. An emerging and prominent theory of cultural evolution is that cultural traits are slightly modified with every transmission such that over time these modifications accumulate to bring about an adaptive cumulative culture. Though cumulative culture is a powerful theory in that it faithfully represents the complex nature of societal change, this complexity is exceptionally challenging to formalise in quantitative models. This paper provides an active-inference formalisation of cumulative culture in three steps that correspond to the stages of cultural evolution: (1) Firstly, we cast cultural transmission as a bi-directional process of communication. When active inference agents communicate, they are able to understand each other by referring to their own generative model and inferring the internal state of the other from their behaviour. This couples communicating agents in an action perception cycle of prediction and inference that induces a generalised synchrony between their internal states. We operationalise generalised synchrony as a particular convergence between the internal states of interlocutors and provide accompanying simulations of dyadic communication between active inference agents. When we simulate these local dynamics, generalised synchrony is largely modulated by agents’ sensitivity to model evidence. (2) Secondly, we cast cumulative culture as the emergence of accumulated modifications to cultural beliefs from the local efforts of agents to converge on a shared narrative. As a proof of principle for this hypothesis, we simulate a population of agents that interchangeably engage in dialogue with each other over time. When we introduce a divergent belief state to a uniform population holding (variations of) a status quo belief, it spreads through it and brings about cumulative collective behaviour of separation and isolation between groups holding distinct beliefs. (3) Thirdly, we discuss a possible formalisation of innovation as an emergent property from the dynamics of cumulative culture described here, thus closing the circular dynamics underlying complex cultural evolution.

Image by The New York Public Library

Planetary cognition is autonomous in relation to the collective intelligence of the biosphere

From the recent formal proposal of the Gaia hypothesis in which the Earth system possess a Markov blanket and therefore active inference (Rubin et al, 2020), here I propose that any trajectory that the Earth system (hothouse, coldhouse, etc.) is autonomous, anticipative and goes beyond the collective intelligence of its biosphere and therefore the anthroposphere. This is based on the distinction of discrete biological unities as autopoietic and cognitive systems, which, despite being constituted by collectivities, do not depend on them for their decisions and choose.

Prepped to be Shaven

Affordance negotiation: The case of gender and its fluidity

The aim of this talk is to provide an enactive-ecological account of gender and its fluidity. Gender is often viewed as static binary state for people to embody, based on the sex they were assigned at birth. However, cultural studies increasingly understand gender as neither binary nor static, a view supported both in psychology and sociology. On this view, gender is negotiated between individuals, and highly dependent on context. Specifically, individuals are thought to be offered culturally gendered social scripts that allow them and those they interact with the ability to predict future actions, and to understand the scene being set, establishing roles and expectations. We propose to understand scripts in the framework of enactive-ecological predictivism, which integrates aspects of ecological enactivism, notably the importance of dynamical sensorimotor interaction with an environment conceived as a field of affordances, and predictive processing, which views the brain as a predictive engine that builds its probabilistic models to reduce prediction error. Under this view, script-based negotiation can be linked to the enactive neuroscience concept of a cultural niche, understood as a landscape of cultural affordances. Affordances are possibilities for action that constrain what actions are pre-reflectively felt possible based on biological, experiential and cultural multisensorial cues. Given these theoretical developments, we propose to understand those aspects of gender that reduce to scripts as culturally set-up fields of affordance and action-affordance loops within these fields. With the shifting landscapes of cultural affordances brought about by a number of recent social, technological and epistemic developments, the gender scripts offered to individuals are less culturally rigid, which translates in an increase in the variety of affordance fields each individual can negotiate. This entails that any individual has an increased possibility for gender fluidity, which may in part explain the increasing number of people currently identifying outside the binary.

Call for Papers and Posters on Active inference and collective intelligence - CLOSED

You are invited to submit an abstract for a presentation at the Cognitio 2021 – Active inference and collective intelligence, organized by the Cognitive Sciences Institute (ISC) at UQAM, Montreal.

Related to deep learning and predictive coding, Active inference has had many practical applications in recent years. It has been used to gain insights and elucidate the dynamics underlying things (Friston, 2019), the multiscale nervous system (Friston,  Zeidman, Fagerholm,  Zarghami, Parr,  Hipólito,  Magrou,  & Razi, 2020 ; Hipólito, Ramstead, Convertino,  Bhat,  Friston,  Parr,  2021), organisms interacting with the environment (Veissière, Constant, Ramstead, Friston,  & Kirmayer, 2020), all the way up to forms of collective intelligence.

The aim of this colloquium is to collect and discuss insights upon forms of collective intelligence using the increasingly insightful tool of active inference under dynamical causal modelling. Specifically, this event focuses on using the insights on collective intelligence for action. The end goal of the colloquium is to (1) gain insights on active inference as a tool; (2) gain insights of collective intelligence of societal organization, or “epistemic communities”, through the lens of active inference; (3) collect insights on real-world ways of implementing real change towards human and social development, where this include climate change, sustainability, as well as societal segregation, discrimination and minority issues.

Graduates, postgraduates and postdoctoral students will gather to discuss how to use active inference as a means to promote social justice, and improve communication across multiple perspectives and disciplines, including psychology, philosophy, linguistics, education, neuroscience, anthropology, cognitive computing, but also biological sciences, chemistry and physics. Talks and poster presentations from young researchers affiliated with a university or a research centre and working on active inference in any of these fields are welcome. Collaborative submissions (more than one discipline and more than two young researchers) will be prioritized. We especially welcome proposals from scholars with backgrounds that are underrepresented in publishing and academia (including women, ethnic minority scientists, scientists with disabilities, researchers of the global south, and other underrepresented groups).


The topics of the event include (but are not limited to):

  1. Active inference (as well as deep learning and predictive coding)

  2. Active inference for collective intelligence

  3. Collective action for human and social development

Encouraged sub-themes for original presentations are:

  • Active inference and transparency

  • Designing active inference models that promote social justice

  • Ensuring representativity in active inference research

  • Social issues that can be modeled via active inference

  • Neo-materialist and feminist studies of active inference

  • Contributions of post/decolonial perspectives on active inference research

  • Active inference models of epistemic injustice

  • Focusing on bridging gaps and forming a common language through active inference

  • Learning, Teaching and Communicating about Active inference

  • The state of the active inference field as it pertains to diversity

  • Appropriation of the active inference tool for capitalistic purposes

  • The agency of the researcher in ensuring such tools are used for social good

  • Similar themes involving deep learning and predictive coding



Friston, K. (2019). A free energy principle for a particular physics. arXiv preprint arXiv:1906.10184.

Friston, K., Zeidman, P., Fagerholm, E., Zarghami, T., Parr, T., Hipólito, I., Magrou, L. & Razi, A. (2020) Parcels and particles: Markov blankets in the brain. Network Neuroscience

Hipólito, I. Ramstead, M., Convertino, L., Bhat, A. Friston, K. Parr, T. (2021). Markov Blankets in the Brain. Neuroscience & Biobehavioral Reviews.

Veissière, S. P., Constant, A., Ramstead, M. J., Friston, K. J., & Kirmayer, L. J. (2020).

Guidelines for submission

All submissions must be made through the EasyChair system -


If you don’t have an EasyChair account, click on the "create an account " tab and follow the instructions. When you are connected, click the "New Submission" tab.

Your submission must include:

  • A title

  • An abstract of about 350 words (2500 characters, spaces included)

  • Full name, affiliation and email address of the author and co-authors.

  • A minimum of 3 keywords

  • The theme relevant to your poster/talk/paper and the discipline(s) studied

Deadline - CLOSED

The submission deadline is now June 28th, 2021. Notifications of acceptance will be sent to authors by July 16th, 2021 with guidelines.

Capture d’écran 2021-02-11 à 14.00.45.png
Get your ticket and get the link to our conference

This conference will be broadcasted online. Get your tickets to receive further information in order to participate. 

No upcoming events at the moment
Where Great Minds Reunite

Lectures. Networking. Workshops.