PreprintPDF Available
Preprints and early-stage research may not have been peer reviewed yet.


In their important contribution to the free energy principle (FEP) literature, Raja et al. (2021) point out crucial shortcomings and issues for the FEP to meet its ambitious goals, including the provision of a unified science with specific focus on cognitive and biological sciences. and the FEP ambition to establish an operationally defined, objective metaphysics or ontology through the FEP. We want to comment on these two critiques, and explore potential ways forward.
Co-constructing Markov blankets: tricky solutions
Authors: Thomas van Es1*, Inês Hipólito 2, 3
1 Centre for philosophical psychology, Universiteit Antwerpen, Belgium
2 Berlin School of Mind and Brain, Humboldt-Universität zu Berlin, Berlin, Germany
3 Department of Psychology, Amsterdam Brain and Cognition, University of Amsterdam, Amsterdam, Netherlands
*corresponding author
Keywords: co-constructivism, historicity, hierarchical nesting, free energy, dynamical systems
Comment on Raja, V., Valluri, D., Baggs, E., Chemero, A., & Anderson, M. L. (2021) The Markov blanket trick: On
the scope of the free energy principle and active inference.Physics of Life Reviews.
In their important contribution to the free energy principle (FEP) literature, Raja et al. (2021)
point out crucial shortcomings and issues for the FEP to meet its ambitious goals, including the
provision of a unified science with specific focus on cognitive and biological sciences.
Additionally, Raja et al. criticise an FEP ambition to establish an operationally defined, objective
metaphysics or ontology through the FEP. We want to comment on these two critiques, and
explore potential ways forward.
We shall first discuss the issues with the FEP ontology, as they also have implications for
the promise of a unified science. The process of establishing a Markov blanket is fundamentally
co-constructed by the modeller, their history, sociomaterial environment and research interests, and
the real-world system in context that make up the experimental system (Hipólito and van Es
2022). This precludes a claim to objectivity, as different modellers will carve up a system
differently. However, this need not be problematic. As we have said before: “to demand an
ontology over and above what is relevant to our research interests is to demand an ontology that
is epiphenomenal to our investigations” (see Hipólito and van Es, Forthcoming).
This poses issues for realist perspectives on the FEP, which aim to use the formalism as a
proxy for the world to be understood. Under a realist reading, model-based findings would have
a direct relation to the world under study. The computations used in the model are typically
considered equivalent to the computations used by the real-world system it is modelled after; the
boundaries established in the formalism would then also map onto real boundaries in the world.
These derivations from the FEP are problematized by Raja et al.’s criticism.
Yet an ontologically more parsimonious, ‘instrumentalist’ approach of the FEP in which
the FEP is taken to be an interesting tool for investigations rather than a guide to a metaphysical
truth (see e.g. van Es 2020) is not shielded from this criticism either. It seems, following Raja et
al.’s paper, that even as a tool for investigations the FEP may be lacking. The onus is then on
FEP theorists to show how the FEP can contribute to biological and cognitive sciences.
Acknowledging the co-constructive nature of the FEP helps understanding its
contributions. With co-constructive science we mean the following. The various aspects of the
experimental system influence and construct one another and the whole dialectically. Not only
does the experimenter actively intervene in the system to be studied in its experimentation, but
they also construct it by framing it, determining the boundaries of the system of interest,
determining what is and isn’t of interest. The experimenter is also constructed in virtue of its
participation in the experimental environment. Its own behavioural repertoire is directed by the
experiment, the system to be studied, and the sociomaterial environment which includes the
research group’s aims, research conventions and so on (Hesp and Hipólito, 2022).
Co-construction is a historical, temporally thick concept: the current co-constructive
processes are themselves constructed (and continue to be constructed) historically. They also
construct the future situation. In this sense, co-construction is intended as an expansion of the
biological concept of niche construction. By these lights, the current scientific endeavours also
impact the practices of other researchers and others beyond the academy. Think of the way
publications of results open up further investigations or novel questions, spark new
methodological debates or spur on other researchers to incorporate the methodology into their
own investigations and explore the possibilities.
Given this perspective on FEP modelling, there are two fruitful, complementary
responses to the challenge by Raja et al. 1) We can look outside of the FED's ambitions for
applications of the FEP despite these limitations, and simultaneously 2) we can develop the
foundational structures of the FEP so as to overcome the limitations. In the former category,
Northoff et al. (2022) describe how connecting the FEP with a temporospatial dynamic view of
neuro-mental processes could allow us to manufacture ‘adaptive agents’ that could augment
instead of imitate human behaviours, and help patients navigate troublesome situations based on
a database of stored experiences that are flexibly put to use by the free energy minimising
system. Similarly, Da Costa et al. (2022) indicate that challenges in robotics such as robustness
and planning can be alleviated by implementing the FEP. Fields et al. (2022) use the FEP
formalisation of physical interactions as information exchange in conjunction with the
development of quantum theory as a scale-free information theory to break new ground in
quantum biology. Integration of the entropic brain hypothesis in neuropharmacology
(Carhart-Harris 2018) with the FEP has also been helpful in understanding the success of
psychedelics in mental health therapy (Hipólito et al., 2022; Carhart-Harris and Friston 2019).
These are just a few examples in which the tools and methodologies of the FEP are used and
applied beyond its initial intentions to explore novel territory for otherwise unanswered
questions. Even without solving the foundational issues that limit the FEP’s grander ambitions,
then, there remain important contributions to be found in its wider constructive impact.
There is also work that rises up to the latter, foundational issue. Unfortunately, modelling
the entire development of the course and dynamics of a living being during their lifespan remains
an impossible task due to the high level of complexity. After all, complexity is proportional to the
tractability for modelling. This is generally known as the intractability of the posterior in Bayesian
inference. This problem calls for dimensionality reduction.1As a methodological concession, this
common technique in modelling science treats a nonlinear system (i.e. a system whose output is
not proportional to the change of the input, therefore chaotic, unpredictable, or counterintuitive)
as if it were linear (i.e.the output is proportional to the input) (Hipólito and van Es, 2022).
Yet methods that better encompass complexity can be found in dynamical and complex
systems theory. Using the FEP, we can apply Markov blankets in a dynamical setting. For this, we
can restrict ourselves and look into a particular moment in time. Using nested Markov blankets,
one starts with the acknowledgment of the situatedness of variational dynamics of a behaviour
(i.e. that a system’s behaviour is situated in a history of continuous flux of interactions with their
environment). Any state in time then depends on a set of previous states: those without which
the present state would not emerge or exist in the way that it does. The present state density of a
system is thus determined by the system dynamics at a previous time (Parr et al. 2021).
Simultaneously, it relates to the future in a probabilistic manner (Parr et al. 2020; Hipólito et al.
2021). Thus, while limited to specific, well-defined situations, the FEP researchers are making
headway into accommodating historicity. Nonetheless, we concede to Raja et al. that it remains
difficult to account for larger scale dynamics and drastic changes over time that we find in many
living systems. In this comment, however, we have shown that there remains plenty for the FEP
to address, tackle, and influence within and outside of its usual boundaries.
Inês Hipólito gratefully acknowledges support from the Berlin School of Mind and Brain and the
Institute of Philosophy at the Humboldt-Universität zu Berlin. We also would like to thank
Guilherme Sanches de Oliveira for fruitful discussions that helped shape some of these ideas.
1For a survey of dimensionality reduction techniques see Sorzano (2014).
Carhart-Harris R. L. (2018). The entropic brain - revisited. Neuropharmacology,142, 167–178.
Carhart-Harris, R. L., & Friston, K. (2019). REBUS and the anarchic brain: toward a unified
model of the brain action of psychedelics.Pharmacological reviews,71(3), 316-344.
Da Costa, L., Lanillos, P., Sajid, N., Friston, K., & Khan, S. (2022). How active inference could
help revolutionise robotics. Entropy,24(3), 361.
Fields, C., Friston, K., Glazebrook, J. F., & Levin, M. (2022). A free energy principle for generic
quantum systems. Progress in Biophysics and Molecular Biology.
Friston, K. J., Fagerholm, E. D., Zarghami, T. S., Parr, T., Hipólito, I., Magrou, L., & Razi, A.
(2021). Parcels and particles: Markov blankets in the brain. Network Neuroscience,5(1), 211-251.
Hesp, C., & Hipólito, I. (2022). Living on the edge-practical information geometry for studying
the emergence and propagation of life forms: Comment on" How particular is the physics of the
free-energy principle?" by Aguilera et al. Physics of life reviews, 42, 52-55.
Hipólito, ., Mago, J., Rosas, F., & Carhart-Harris, R. (2022). Pattern Breaking: A Complex
Systems Approach to Psychedelic Medicine. Retrieved from
Hipólito, I. & van Es, T. (Forthcoming) Free Energy Pragmatics: Markov blankets don’t
prescribe objective ontology, and that’s okay. Behavioural Brain Sciences
Hipólito, I., & van Es, T. (2022). Enactive-Dynamic Social Cognition and Active Inference.
Frontiers in psychology,13, 855074.
Hipólito, I., Ramstead, M. J., Convertino, L., Bhat, A., Friston, K., & Parr, T. (2021). Markov
blankets the brain. Neuroscience & Biobehavioral Reviews,125, 88-97.
Northoff, G., Fraser, M., Griffiths, J., Pinotsis, D. A., Panangaden, P., Moran, R., & Friston, K.
(2022). Augmenting Human Selves Through Artificial Agents - Lessons From the Brain. Frontiers
in computational neuroscience,16, 892354.
Parr, T., Da Costa, L., & Friston, K. (2020). Markov blankets, information geometry and
stochastic thermodynamics. Philosophical Transactions of the Royal Society A,378(2164), 20190159.
Parr, T., Da Costa, L., Heins, C., Ramstead, M. J. D., & Friston, K. J. (2021). Memory and markov
blankets. Entropy,23(9), 1105.
Raja, V., Valluri, D., Baggs, E., Chemero, A., & Anderson, M. L. (2021). The Markov blanket
trick: On the scope of the free energy principle and active inference. Physics of Life Reviews, 39,
Sorzano, C. O. S., Vargas, J., & Montano, A. P. (2014). A survey of dimensionality reduction
techniques. arXiv preprint arXiv:1403.2877.
van Es, T. (2020). Minimising prediction errors in predictive processing: from inconsistency to
non-representationalism. Phenomenology and the Cognitive Sciences,19(5), 997-1017.
ResearchGate has not been able to resolve any citations for this publication.
Full-text available
Much of current artificial intelligence (AI) and the drive toward artificial general intelligence (AGI) focuses on developing machines for functional tasks that humans accomplish. These may be narrowly specified tasks as in AI, or more general tasks as in AGI – but typically these tasks do not target higher-level human cognitive abilities, such as consciousness or morality; these are left to the realm of so-called “strong AI” or “artificial consciousness.” In this paper, we focus on how a machine can augment humans rather than do what they do, and we extend this beyond AGI-style tasks to augmenting peculiarly personal human capacities, such as wellbeing and morality. We base this proposal on associating such capacities with the “self,” which we define as the “environment-agent nexus”; namely, a fine-tuned interaction of brain with environment in all its relevant variables. We consider richly adaptive architectures that have the potential to implement this interaction by taking lessons from the brain. In particular, we suggest conjoining the free energy principle (FEP) with the dynamic temporo-spatial (TSD) view of neuro-mental processes. Our proposed integration of FEP and TSD – in the implementation of artificial agents – offers a novel, expressive, and explainable way for artificial agents to adapt to different environmental contexts. The targeted applications are broad: from adaptive intelligence augmenting agents (IA’s) that assist psychiatric self-regulation to environmental disaster prediction and personal assistants. This reflects the central role of the mind and moral decision-making in most of what we do as humans.
Full-text available
Recent advances in neuroscience have characterised brain function using mathematical formalisms and first principles that may be usefully applied elsewhere. In this paper, we explain how active inference—a well-known description of sentient behaviour from neuroscience—can be exploited in robotics. In short, active inference leverages the processes thought to underwrite human behaviour to build effective autonomous systems. These systems show state-of-the-art performance in several robotics settings; we highlight these and explain how this framework may be used to advance robotics.
Full-text available
The aim of this paper is twofold: it critically analyses and rejects accounts blending active inference as theory of mind and enactivism; and it advances an enactivist-dynamic account of social cognition that is compatible with active inference. While some inference models of social cognition seemingly take an enactive perspective on social cognition, they explain it as the attribution of mental states to other people, via representational machinery, in line with Theory of Mind (ToM). Holding both enactivism and ToM, we argue, entails contradiction and confusion due to two ToM assumptions rejected by enactivism: (1) that social cognition reduces to mental representation and (2) cognition must be hardwired with a social cognition contentful “toolkit” or “starter pack” for fueling the model-like theorising supposed in (1). The paper offers a positive alternative, one that avoids contradictions or confusions. After clarifying the profile of social cognition under enactivism, i.e. without assumptions (1) and (2), the last section advances an enactivist-dynamic model of cognition as dynamic, real time, fluid, dynamic, contextual social action, where we use the formalisms of dynamical systems theory to explain the origins of sociocognitive novelty in developmental change and active inference as a tool to explain social understanding as generalised synchronisation.
Full-text available
In theoretical biology, we are often interested in random dynamical systems—like the brain—that appear to model their environments. This can be formalized by appealing to the existence of a (possibly non-equilibrium) steady state, whose density preserves a conditional independence between a biological entity and its surroundings. From this perspective, the conditioning set, or Markov blanket, induces a form of vicarious synchrony between creature and world—as if one were modelling the other. However, this results in an apparent paradox. If all conditional dependencies between a system and its surroundings depend upon the blanket, how do we account for the mnemonic capacity of living systems? It might appear that any shared dependence upon past blanket states violates the independence condition, as the variables on either side of the blanket now share information not available from the current blanket state. This paper aims to resolve this paradox, and to demonstrate that conditional independence does not preclude memory. Our argument rests upon drawing a distinction between the dependencies implied by a steady state density, and the density dynamics of the system conditioned upon its configuration at a previous time. The interesting question then becomes: What determines the length of time required for a stochastic system to ‘forget’ its initial conditions? We explore this question for an example system, whose steady state density possesses a Markov blanket, through simple numerical analyses. We conclude with a discussion of the relevance for memory in cognitive systems like us.
Full-text available
The free energy principle (FEP) has been presented as a unified brain theory, as a general principle for the self-organization of biological systems, and most recently as a principle for a theory of every thing. Additionally, active inference has been proposed as the process theory entailed by FEP that is able to model the full range of biological and cognitive events. In this paper, we challenge these two claims. We argue that FEP is not the general principle it is claimed to be, and that active inference is not the all-encompassing process theory it is purported to be either. The core aspects of our argumentation are that (i) FEP is just a way to generalize Bayesian inference to all domains by the use of a Markov blanket formalism, a generalization we call the Markov blanket trick; and that (ii) active inference presupposes successful perception and action instead of explaining them.
Full-text available
Recent characterisations of self-organising systems depend upon the presence of a ‘Markov blanket’: a statistical boundary that mediates the interactions between the inside and outside of a system. We leverage this idea to provide an analysis of partitions in neuronal systems. This is applicable to brain architectures at multiple scales, enabling partitions into single neurons, brain regions, and brain-wide networks. This treatment is based upon the canonical micro-circuitry used in empirical studies of effective connectivity, so as to speak directly to practical applications. The notion of effective connectivity depends upon the dynamic coupling between functional units, whose form recapitulates that of a Markov blanket at each level of analysis. The nuance afforded by partitioning neural systems in this way highlights certain limitations of ‘modular’ perspectives of brain function that only consider a single level of description.
The Free Energy Principle (FEP) states that under suitable conditions of weak coupling, random dynamical systems with sufficient degrees of freedom will behave so as to minimize an upper bound, formalized as a variational free energy, on surprisal (a.k.a., self-information). This upper bound can be read as a Bayesian prediction error. Equivalently, its negative is a lower bound on Bayesian model evidence (a.k.a., marginal likelihood). In short, certain random dynamical systems evince a kind of self-evidencing. Here, we reformulate the FEP in the formal setting of spacetime-background free, scale-free quantum information theory. We show how generic quantum systems can be regarded as observers, which with the standard freedom of choice assumption become agents capable of assigning semantics to observational outcomes. We show how such agents minimize Bayesian prediction error in environments characterized by uncertainty, insufficient learning, and quantum contextuality. We show that in its quantum-theoretic formulation, the FEP is asymptotically equivalent to the Principle of Unitarity. Based on these results, we suggest that biological systems employ quantum coherence as a computational resource and – implicitly – as a communication resource. We summarize a number of problems for future research, particularly involving the resources required for classical communication and for detecting and responding to quantum context switches.
In their impressive paper, Bruineberg et al. (2021) make a significant contribution to the Free Energy Principle literature by distinguishing between 'Pearl blankets' and 'Friston blankets', identifying the former as an epistemic tool, and the latter in terms of its novel metaphysical use. We note the oft-forgotten theoretical context of these statistical tools and the need for empirical testing next to computational modeling. A peculiar aspect of the FEP is its use in support of radically opposed ontologies of the mind. In our view, the objective ontological aspiration itself should be rejected; we propose a more thoroughly pragmatic instrumentalist view.
At the inception of human brain mapping, two principles of functional anatomy underwrote most conceptions – and analyses – of distributed brain responses: namely functional segregation and integration. There are currently two main approaches to characterising functional integration. The first is a mechanistic modelling of connectomics in terms of directed effective connectivity that mediates neuronal message passing and dynamics on neuronal circuits. The second phenomenological approach usually characterises undirected functional connectivity (i.e., measurable correlations), in terms of intrinsic brain networks, self-organised criticality, dynamical instability, etc. This paper describes a treatment of effective connectivity that speaks to the emergence of intrinsic brain networks and critical dynamics. It is predicated on the notion of Markov blankets that play a fundamental role in the self-organisation of far from equilibrium systems. Using the apparatus of the renormalisation group, we show that much of the phenomenology found in network neuroscience is an emergent property of a particular partition of neuronal states, over progressively coarser scales. As such, it offers a way of linking dynamics on directed graphs to the phenomenology of intrinsic brain networks.