Pith. sign in

REVIEW 3 major objections 4 minor 58 references

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

T0 review · 3 major / 4 minor · reviewed 2026-08-09 · deepseek-v4-flash

Pith's one-line read This paper argues that social power is a self-reinforcing cycle of attracting attention and expanding information-processing capacity.

desk verdict Worth a careful read for its integrated framing of power as attention plus computation, but the formal centerpiece—the claim that belief divergence vanishes under repeated interaction—is asserted, not derived, and that limits what the paper can currently support. read the letter →

arxiv 2501.19368 v3 pith:HOA2PSG2 submitted 2025-01-31 physics.soc-ph

classification physics.soc-ph
keywords powerscriptsactiveinferencenarrativessemanticattractionsharedintentionalitycollectiveattention
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish a computational theory of social power using active inference, the view that agents act to minimize the gap between their predictions and their observations. Its central claim is that power has two ingredients: the ability to attract the attention of people or groups whose actions can bring about the outcomes you want, and the capacity to process information effectively enough to choose and execute those outcomes. Those two ingredients feed each other: winning attention gives access to more policies and computing power, and greater computing power makes an agent more worth attending to. Narratives and social scripts are the levers in this loop, acting as semantic attractors that absorb uncertainty and direct collective attention toward preferred futures. The payoff is that power stops being a vague sociological quality and becomes a quantity that could in principle be measured, compared, and redistributed through the same formal machinery that describes perception and action.

What carries the argument

The central machinery is the active-inference formalism applied to social interaction: the expected-free-energy objective with a deontic value term, the empowerment measure defined as maximum mutual information between actions and future states, and the belief-alignment limit $\lim_{t\to\infty} D_{KL}[Q_t^i \| Q_t^j] \approx 0$. The load-bearing component is precision modulation: the power to adjust how much confidence agents place in particular beliefs, which lets a leader or influencer pull other agents' generative models toward their own without coercion. Scripts and narratives are analyzed as semantic attractors—stable patterns of meaning that pool attention, reduce the dimensionality of the state space, and make certain futures seem natural or inevitable—and are therefore the instruments through which precision is exercised.

What would settle it

Run a controlled dyadic interaction experiment or a simulation of two active-inference agents in which each agent's beliefs about a shared hidden state are measured across repeated exchanges; if the KL divergence between their belief distributions does not reliably decrease, or settles at a nonzero value, the claim that free-energy minimization synchronizes beliefs under appropriate interaction conditions is falsified.

Watch

Extended reading notes

Core claim

The paper's discovery is that social power can be understood as a dual mechanism: an agent's power is a function of (1) how well it attracts the attention of others who can facilitate desired policies, and (2) how well it can process information to pursue those policies. These reinforce each other, producing a cycle in which the powerful expand the set of policies and state spaces available to them while also buffering against risk. The formal core of the argument is that social interaction drives the belief distributions of agents toward each other—the Kullback-Leibler divergence between their generative models shrinks under repeated exchange—and that power is the capacity to make that convergence happen on one's own terms. This is achieved by shaping semantic attractors and modulating the precision of beliefs, so that others' generative models align with the powerful agent's model even in the absence of overt coercion. Because one agent's internal state is another agent's external state, maximizing empowerment in a social world amounts to making other agents more predictable and useful, which is most easily done by aligning their beliefs with one's own.

Load-bearing premise

The load-bearing premise is that repeated social interaction genuinely drives agents' belief distributions toward each other—that under 'appropriate interaction conditions' the KL divergence between two agents' generative models converges to zero—because the entire account of power as belief alignment, narrative control, and precision modulation hinges on that convergence.

Editorial extensions

If this is right

  • If power is the control of attention plus computation, then interventions on public attention—media, education, narrative design—are direct interventions on power, not precursors to it.
  • The self-reinforcing cycle predicts that power asymmetries will tighten over time unless some friction, such as divisive narratives, context constraints, or resistance, interrupts the loop.
  • Because belief alignment can arise from confidence and precision alone, the framework predicts that a sufficiently confident communicator can shift group behavior without formal authority or sanctions.
  • The formalism suggests measurable proxies for power: mutual information between an agent's actions and future states, KL divergence between belief distributions, and the precision of communicated predictions.
  • The same logic implies that territorial, material, and cultural power can be reframed as informational and computational power, with resource inequality reflecting unequal access to computation and attention.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: the convergence result could be tested empirically with conversation data; if belief alignment does not reliably increase in natural dyadic interactions, the framework's core mechanism is context-dependent rather than universal.
  • Editorial inference: one could build a quantitative 'possibilistic power' index from attention share and information-processing capacity, and check whether it reproduces familiar concentration curves such as wealth or citation distributions; the paper does not take this step.
  • Editorial inference: a natural simulation experiment is to place two active-inference agents with mismatched generative models in repeated interaction and observe whether the KL divergence reaches zero or a nonzero steady state, since the 'appropriate interaction conditions' for convergence are left unspecified.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper proposes a conceptual framework for understanding social power through the active inference framework (AIF), arguing that power is a function of an agent's or group's ability to attract others' attention and to process information effectively. It introduces the notion of 'possibilistic power' and analyzes narrative scripts, attention, and belief alignment as mechanisms of power. The paper includes a formal layer in Section 3.3 with equations for deontic value, KL divergence between agents' belief distributions, and a precision-modulated policy selection rule, but the authors state that detailed modeling and empirical testing are left for future work.

Significance. If accepted as a theoretical essay, the paper offers a valuable interdisciplinary synthesis, linking active inference, social theory, and political philosophy in a way that may inspire testable models of social influence. The emphasis on attention as a resource and on 'semantic attractors' as devices for shaping collective belief is a plausible and potentially generative reframing. The paper is honest about the limits of its formalization, and it explicitly flags the intractability of many ingredients for future modeling. However, the central formal claims in Section 3.3 are asserted rather than proved, and no empirical or simulation evidence is provided; as a result, the paper's contribution is mostly conceptual, with the formal apparatus serving as an illustrative metaphor rather than a workable quantitative framework.

major comments (3)
  1. [Section 3.3, KL divergence limit] The claim lim_{t→∞} D_KL[Q^t_i || Q^t_j] ≈ 0 is load-bearing for the paper's central thesis that social power operates through aligning other agents' generative models, but it is asserted without specifying the interaction conditions under which it holds. The text says 'under appropriate interaction conditions' but never defines a class of generative models, a coupling structure, a precision schedule, or observability assumptions. Standard active inference minimizes each agent's own expected free energy and does not guarantee across-agent convergence; joint-action models with private observations and distinct priors can reach equilibria with nonzero between-agent divergence. The paper's own footnote 17 concedes that synchronization need not occur in adversarial or mixed-motive settings, but this caveat is not integrated into the formal statement. The authors should either provide a theorem with explicit conditions, or explicitly present this as a conjecture that is not yet established. As written, the subsequent claims about the 'smaller update' criterion and 'precision-driven leadership' do not follow.
  2. [Section 3.3, π* equation] The equation π* = arg min_π [G(π) + α · precision(Q(s|o,π))] contains an undefined modulating factor α and an ambiguous precision term. The text states that α 'reflects the power to affect the perceived reliability of beliefs,' but there is no account of how α is determined, learned, or bounded, nor of what distribution the argument of precision() denotes (precision of the observation likelihood? of the posterior? of a policy-dependent prediction?). Without these definitions, the equation is not a formal result but a placeholder. The authors should either define α and precision() concretely or remove the equation and describe the mechanism verbally as a hypothesis.
  3. [Section 3.3, 'smaller update' criterion] The inference that the agent making the significantly smaller belief update 'may do so as a result of being in a position of greater social power' is not derived from the KL convergence limit. Even if both limits converge to zero, the relative magnitudes of D_KL[Q^t_i || Q^0_i] and D_KL[Q^t_j || Q^0_j] depend on the agents' priors, the sensitivity of their generative models to new observations, and the parametrization of the belief distributions. Attributing a smaller update to 'social power' is an extra interpretive step that requires a model or at least a clearly stated assumption. This statement should be explicitly labeled as a conjecture, or supported with a minimal simulation.
minor comments (4)
  1. [Section 0.1] The name 'Bordieu' appears as a typo on page 3; it should be 'Bourdieu'.
  2. [Section 1.1, free energy decomposition] The equation for variational free energy is displayed with a bracket notation that is not standard and could confuse readers; the decomposition into complexity and accuracy is correct, but the labels under the first equation appear garbled ('Divergence' and 'surprise' are placed under terms in a way that is not self-explanatory).
  3. [Section 2.1] The citation '[] (Constant et al., 2018)' is incomplete; the bracket is empty and the year does not match the 'Constant et al., 2019' reference in the bibliography. Please correct the citation.
  4. [Section 3.3] The paper states that 'a deeper investigation into how the two are precisely related is beyond the scope of this paper' regarding deontic value and empowerment; this is fine as a limitation, but the text should also flag that the formal connection between D(π,o) and the KL synchronization mechanism is currently left open, so that the reader does not over-read the equations as a unified formal model.

Circularity Check

2 steps flagged · score 4.0 of 10

Modest circularity: the smaller-update-as-greater-power criterion is read off the paper's own definition of power, and the script premise is inherited from a same-author prior paper; the convergence limit is an unsupported assertion rather than a derivation.

  1. self definitional [Section 3.3 (Formal Framework Integration), around the DKL convergence equation and the power-as-smaller-update gloss]
    "The agent who makes the significantly smaller update may do so as a result of being in a position of greater social power. ... As we conceptualized it, power is an agent’s capacity to shape and control the (semantic) “attractors” within a social or informational landscape."

    The claim that the smaller updater is the more powerful agent is not a consequence of expected free energy minimization; it is a restatement of the definition of power adopted in the same passage, namely the capacity to shape attractors and modulate precision. The preceding limit lim DKL→0 is assumed, not derived, so the subsequent reading of that limit as evidence that powerful agents induce convergence, and of small update size as indicating power, is read off the paper’s own definition rather than independently established.

  2. self citation load bearing [Section 2.1 (Scripts as frameworks for social behavior)]
    "Social scripts are blueprints that enable agents to navigate social situations and select situationally appropriate behaviors [Albarracin et al., 2021]. ... They thus reduce uncertainty and create predictable social patterns that benefit those already in power [Albarracin et al., 2021]."

    The paper’s central mechanism — that narratives and scripts are the structures through which power is mediated — is imported from Albarracin et al. (2021), whose first author is also the first author of the present paper. The manuscript does not re-derive or independently test the ‘scripts as variational models’ premise; instead it uses that same-author result as the load-bearing starting point for its account of how attention-shaping narratives confer power. This is a self-citation chain rather than a derivation from first principles, though it is not the only basis for the paper’s broader conceptual claim.

full rationale

Most of the paper is a conceptual reframing rather than a derivation: no data are fitted, so the usual fitted-input-called-prediction circularity does not arise. The main formal move in Section 3.3 is an asserted limit, DKL→0, which is not proved; that is a correctness gap or unsupported assumption, not itself a circular step. What is circular is the glossing of that asserted convergence as evidence for the power mechanism: identifying the smaller updater with the more powerful agent is a restatement of the paper’s own definition of power as shaping attractors and modulating precision. A second, milder circular feature is the heavy reliance on Albarracin et al. (2021) for the script premise that carries much of the argument. These features warrant a modest score: the central philosophical interpretation remains an independent position, but its formal bridge is partly self-definitional and partly inherited from same-author prior work.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The central claim rests on the active inference framework, the script-as-generative-model assumption, and two social-dynamics assumptions (belief convergence and precision modulation) that are asserted rather than proved. One free parameter, α, appears in the proposed power equation but is unfitted. No new physical entities are introduced.

free parameters (1)
  • α (precision modulation factor)
    Introduced in Section 3.3 in the equation π* = argmin G(π) + α·precision(Q(s|o,π)) to weight the power to affect perceived reliability of beliefs. No value or estimation procedure is provided.
assumptions (4)
  • domain assumption Agents and groups minimize variational free energy and expected free energy; active inference is a valid account of cognition and social behavior.
    Section 1 presents active inference as the theoretical ground for the entire argument, assuming its applicability to human social dynamics.
  • domain assumption Narratives and scripts act as shared generative models that shape agents' policies and beliefs.
    Section 2, drawing on Albarracin et al. 2021 and Constant et al. 2019, assumes scripts are internalized schemas and externalized social orders with deontic cues.
  • ad hoc to paper Under 'appropriate interaction conditions,' repeated social interaction drives the KL divergence between agents' belief distributions to zero.
    Section 3.3 asserts lim DKL -> 0 without specifying the conditions, and the power-as-belief-alignment mechanism rests on this convergence.
  • ad hoc to paper Control over precision (perceived reliability of beliefs) is a feasible and effective way to align other agents' beliefs.
    Section 3.3 introduces precision modulation as the mechanism of leader-follower power, but no evidence or derivation is given that agents can actually modulate the precision of others' beliefs.

how reviews work

0 comments
Cite this review

Pith. "Pith review of The Physics and Metaphysics of Social Powers: Bridging Cognitive Processing and Social Dynamics, a New Perspective on Power through Active Inference." pith.science (2026). https://pith.science/paper/HOA2PSG2

@misc{pith2026250119368,
  author       = {Pith},
  title        = {Pith review of: The Physics and Metaphysics of Social Powers: Bridging Cognitive Processing and Social Dynamics, a New Perspective on Power through Active Inference},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HOA2PSG2}},
  note         = {Machine review of arXiv:2501.19368}
}
read the original abstract

The concept of power can be explored at several scales: from physical action and process effectuation, all the way to complex social dynamics. A spectrum-wide analysis of power requires attention to the fundamental principles that constrain these processes. In the social realm, the acquisition and maintenance of power is intertwined with both social interactions and cognitive processing capacity: socially-facilitated empowerment grants agents more information-processing capacities and opportunities, either by relying on others to bring about desired policies or ultimately outcomes, and/or by enjoying more information-processing possibilities as a result of relying on others for the reproduction of (material) tasks. The effects of social empowerment thus imply an increased ability to harness computation toward desired ends, thereby augmenting the evolution of a specific state space. Empowered individuals attract the attention of others, who contribute to increasing the scale of their access to various policies effectuating these state spaces. The presented argument posits that social power, in the context of active inference, is a function of several variables. As a result of its power-amplifying effects, this extended computational ability also buffers against possible vulnerabilities. We propose that individuals wield power not only by associating with others possessing desirable policies, but also by enhancing their ability to intake and compute information effectively. This dual mechanism is argued to create a cyclical, reinforcing pattern wherein the empowered are able to incrementally expand the scope of policies and state spaces available to them while minimizing risk-exposure.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

58 extracted references · 55 canonical work pages

  1. [1]

    A variational approach to scripts

    Mahault Albarracin, Axel Constant, Karl J Friston, and Maxwell James D Ramstead. A variational approach to scripts. Frontiers in Psychology, 12: 0 585493, 2021

  2. [2]

    Epistemic communities under active inference

    Mahault Albarracin, Daphne Demekas, Maxwell JD Ramstead, and Conor Heins. Epistemic communities under active inference. Entropy, 24 0 (4): 0 476, 2022

  3. [3]

    The math is not the territory: navigating the free energy principle

    Mel Andrews. The math is not the territory: navigating the free energy principle. Biology & Philosophy, 36 0 (3): 0 30, 2021

  4. [4]

    Emergence of scaling in random networks

    Albert-L \'a szl \'o Barab \'a si and R \'e ka Albert. Emergence of scaling in random networks. Science, 286 0 (5439): 0 509--512, 1999

  5. [5]

    Vulture Capitalism: Corporate Crimes, Backdoor Bailouts, and the Death of Freedom

    Grace Blakeley. Vulture Capitalism: Corporate Crimes, Backdoor Bailouts, and the Death of Freedom. Simon and Schuster, 2024

  6. [6]

    Narrative as active inference: an integrative account of the functions of narratives

    Nabil Bouizegarene, MJ Ramstead, Axel Constant, Karl J Friston, and Laurence J Kirmayer. Narrative as active inference: an integrative account of the functions of narratives. Preprint, 10 0 (10.31234), 2020

  7. [7]

    Structures, habitus, practices

    Pierre Bourdieu. Structures, habitus, practices. In Rethinking the subject, pages 31--45. Routledge, 2018

  8. [8]

    Active inference and the primacy of the'i can'

    Jelle Bruineberg. Active inference and the primacy of the'i can'. In Philosophy and predictive processing, pages 1--18. MIND Group, 2017

Show all 58 references
  1. [9]

    Ai survival stories: a taxonomic analysis of ai existential risk

    H Cappelen, S Goldstein, and J Hawthorne. Ai survival stories: a taxonomic analysis of ai existential risk. preprint for Philosophy of AI Journal, 2024

  2. [10]

    Regimes of expectations: an active inference model of social conformity and human decision making

    Axel Constant, Maxwell JD Ramstead, Samuel PL Veissi \`e re, and Karl Friston. Regimes of expectations: an active inference model of social conformity and human decision making. Frontiers in psychology, 10: 0 679, 2019

  3. [11]

    Inevitably falling into place: Intuition as a function of spatial prediction

    S de Jager. Inevitably falling into place: Intuition as a function of spatial prediction. In Philosophy after AI: Meaning and Understanding, Society for the Study of Artificial Intelligence and the Simulation of Behavior, 2021

  4. [12]

    Fiction, speculation, prediction

    S de Jager. Fiction, speculation, prediction. preprint, 2022

  5. [13]

    Semantic noise and conceptual stagnation in natural language processing

    S de Jager. Semantic noise and conceptual stagnation in natural language processing. Angelaki, 28 0 (3): 0 111--132, 2023

  6. [14]

    Postscript on the societies of control

    Gilles Deleuze. Postscript on the societies of control. October, 59: 0 3--7, 1992

  7. [15]

    The self as a center of narrative gravity

    Daniel C Dennett. The self as a center of narrative gravity. In F Kessel, P Cole, and D Johnson, editors, Self and Consciousness: Multiple Perspectives. Erlbaum, Hillsdale, NJ, 1992

  8. [16]

    Capitalist realism: Is there no alternative? Zero Books, 2009

    Mark Fisher. Capitalist realism: Is there no alternative? Zero Books, 2009

  9. [17]

    The subject and power

    Michel Foucault. The subject and power. Critical Inquiry, 8 0 (4): 0 777--795, 1982

  10. [18]

    The history of sexuality: An introduction, volume I

    Michel Foucault. The history of sexuality: An introduction, volume I. Vintage, New York, 1990. Originally published 1978

  11. [19]

    Active inferants: An active inference framework for ant colony behavior

    Daniel Ari Friedman, Alec Tschantz, Maxwell JD Ramstead, Karl Friston, and Axel Constant. Active inferants: An active inference framework for ant colony behavior. Frontiers in behavioral neuroscience, 15: 0 647732, 2021

  12. [20]

    A free energy principle for the brain

    Karl Friston, James Kilner, and Lee Harrison. A free energy principle for the brain. Journal of Physiology-Paris, 100 0 (1-3): 0 70--87, 2006

  13. [21]

    Active inference and epistemic value

    Karl Friston, Francesco Rigoli, Dimitri Ognibene, Christoph Mathys, Thomas Fitzgerald, and Giovanni Pezzulo. Active inference and epistemic value. Cognitive neuroscience, 6 0 (4): 0 187--214, 2015

  14. [22]

    Active inference: A process theory

    Karl Friston, Thomas FitzGerald, Francesco Rigoli, Philipp Schwartenbeck, and Giovanni Pezzulo. Active inference: A process theory. Neural Computation, 29 0 (1): 0 1--49, 2016. doi:10.1162/NECO_a_00912

  15. [23]

    Action and behavior: A free-energy formulation

    Karl J Friston, Jean Daunizeau, James Kilner, and Stefan J Kiebel. Action and behavior: A free-energy formulation. Biological Cybernetics, 102: 0 227--260, 2010

  16. [24]

    Generative models, linguistic communication and active inference

    Karl J Friston, Thomas Parr, Yan Yufik, Noor Sajid, Catherine J Price, and Emma Holmes. Generative models, linguistic communication and active inference. Neuroscience & Biobehavioral Reviews, 118: 0 42--64, 2020

  17. [25]

    Designing ecosystems of intelligence from first principles

    Karl J Friston, Maxwell JD Ramstead, Alex B Kiefer, Alexander Tschantz, Christopher L Buckley, Mahault Albarracin, Riddhi J Pitliya, Conor Heins, Brennan Klein, Beren Millidge, et al. Designing ecosystems of intelligence from first principles. Collective Intelligence, 3 0 (1):...

  18. [26]

    Why heuristics work

    Gerd Gigerenzer. Why heuristics work. Perspectives on Psychological Science, 3 0 (1): 0 20--29, 2008

  19. [27]

    Understanding the constraints in maximum entropy methods for modeling and inference

    Amos Golan and Duncan K Foley. Understanding the constraints in maximum entropy methods for modeling and inference. IEEE Transactions on Pattern Analysis and Machine Intelligence, 45 0 (3): 0 3994--3998, 2022

  20. [28]

    Enactive-dynamic social cognition and active inference

    In \^e s Hip \'o lito and Thomas van Es. Enactive-dynamic social cognition and active inference. Frontiers in Psychology, 13: 0 855074, 2022

  21. [29]

    Free-energy equilibria: Toward a theory of interactions between boundedly-rational agents

    David Hyland, Tom \'a s Gaven c iak, Lancelot Da Costa, Conor Heins, Vojtech Kovarik, Julian Gutierrez, Michael J Wooldridge, and Jan Kulveit. Free-energy equilibria: Toward a theory of interactions between boundedly-rational agents. In ICML 2024 Workshop on Models of Human Fe...

  22. [30]

    Division of labour and the evolution of multicellularity

    Iaroslav Ispolatov, Martin Ackermann, and Michael Doebeli. Division of labour and the evolution of multicellularity. Proceedings of the Royal Society B: Biological Sciences, 279 0 (1734): 0 1768--1776, 2012

  23. [31]

    Information theory and statistical mechanics

    Edwin T Jaynes. Information theory and statistical mechanics. Physical Review, 106 0 (4): 0 620--630, 1957

  24. [32]

    Small steps for mankind: Modeling the emergence of cumulative culture from joint active inference communication

    Natalie Kastel, Casper Hesp, K Richard Ridderinkhof, and Karl J Friston. Small steps for mankind: Modeling the emergence of cumulative culture from joint active inference communication. Frontiers in neurorobotics, 16: 0 944986, 2023

  25. [33]

    Mental gravity: Modelling the embodied self on the physical environment

    Lachlan Kent. Mental gravity: Modelling the embodied self on the physical environment. Journal of Environmental Psychology, page 102245, 2024

  26. [34]

    Obsessive compulsive disorder: a pathology of self-confidence? Trends in Cognitive Sciences, 23 0 (5): 0 369--372, 2019

    Julian Kiverstein et al. Obsessive compulsive disorder: a pathology of self-confidence? Trends in Cognitive Sciences, 23 0 (5): 0 369--372, 2019

  27. [35]

    Empowerment: A universal agent-centric measure of control

    Alexander S Klyubin, Daniel Polani, and Chrystopher L Nehaniv. Empowerment: A universal agent-centric measure of control. In Proceedings of the 2005 IEEE Congress on Evolutionary Computation, pages 128--135, 2005

  28. [36]

    Division of labour in groups of robots

    Thomas Halva Labella. Division of labour in groups of robots. Universite Libre de Bruxelles, 2007

  29. [37]

    Xenofeminist Manifesto

    Laboria Cuboniks . Xenofeminist Manifesto. 2015

  30. [38]

    Metaphors we live by

    George Lakoff and Mark Johnson. Metaphors we live by. University of Chicago Press, 2003. Originally published 1980

  31. [39]

    A unified bellman optimality principle combining reward maximization and empowerment

    Felix Leibfried, Sergio Pascual-Diaz, and Jordi Grau-Moya. A unified bellman optimality principle combining reward maximization and empowerment. Advances in Neural Information Processing Systems, 32, 2019

  32. [40]

    Interactive inference: A multi-agent model of cooperative joint actions

    Domenico Maisto, Francesco Donnarumma, and Giovanni Pezzulo. Interactive inference: A multi-agent model of cooperative joint actions. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 54 0 (2): 0 704--715, 2024

  33. [41]

    Distributed generation for economic benefit maximization through coalition formation--based game theory concept

    Mousa Marzband, Reza Rouhi Ardeshiri, Milad Moafi, and Hassan Uppal. Distributed generation for economic benefit maximization through coalition formation--based game theory concept. International Transactions on Electrical Energy Systems, 27 0 (6): 0 e2313, 2017

  34. [42]

    Mortal computation: A foundation for biomimetic intelligence

    Alexander Ororbia and Karl Friston. Mortal computation: A foundation for biomimetic intelligence. arXiv preprint arXiv:2311.09589, 2023

  35. [43]

    Neuronal message passing using mean-field, bethe, and marginal approximations

    Thomas Parr, Dimitrije Markovic, Stefan J Kiebel, and Karl J Friston. Neuronal message passing using mean-field, bethe, and marginal approximations. Scientific Reports, 9 0 (1): 0 1889, 2019

  36. [44]

    Active Inference: The Free Energy Principle in Mind, Brain, and Behavior

    Thomas Parr, Giovanni Pezzulo, and Karl J Friston. Active Inference: The Free Energy Principle in Mind, Brain, and Behavior. MIT Press, Cambridge, MA, 2022

  37. [45]

    How preferences enslave attention: calling into question the endogenous/exogenous dichotomy from an active inference perspective

    Darius Parvizi-Wayne. How preferences enslave attention: calling into question the endogenous/exogenous dichotomy from an active inference perspective. Phenomenology and the Cognitive Sciences, pages 1--47, 2024

  38. [46]

    Complex behavior from intrinsic motivation to occupy future action-state path space

    Jorge Ram \'i rez-Ruiz, Dmytro Grytskyy, Chiara Mastrogiuseppe, Yasir Habib, and Ruben Moreno-Bote. Complex behavior from intrinsic motivation to occupy future action-state path space. Nature Communications, 15 0 (1), 2024

  39. [47]

    On bayesian mechanics: a physics of and by beliefs

    Maxwell JD Ramstead, Dalton AR Sakthivadivel, Conor Heins, Magnus Koudahl, Beren Millidge, Lancelot Da Costa, Brennan Klein, and Karl J Friston. On bayesian mechanics: a physics of and by beliefs. Interface Focus, 13 0 (20220029), 2023. doi:10.1098/rsfs.2022.0029

  40. [48]

    Seeing like a state: How certain schemes to improve the human condition have failed

    James C Scott. Seeing like a state: How certain schemes to improve the human condition have failed. Yale University Press, 2020

  41. [49]

    Toward a universal law of generalization for psychological science

    Roger N Shepard. Toward a universal law of generalization for psychological science. Science, 237 0 (4820): 0 1317--1323, 1987

  42. [50]

    Theory of collective mind

    Garriy Shteynberg et al. Theory of collective mind. Trends in Cognitive Sciences, 27 0 (11): 0 1019--1031, 2023

  43. [51]

    Rationality as process and as product of thought

    Herbert A Simon. Rationality as process and as product of thought. The American Economic Review, 68 0 (2): 0 1--16, 1978

  44. [52]

    Intrinsic motivation in dynamical control systems

    Stas Tiomkin, Ilya Nemenman, Daniel Polani, and Naftali Tishby. Intrinsic motivation in dynamical control systems. PRX Life, 2 0 (3): 0 033009, 2024

  45. [53]

    Communication as socially extended active inference: An ecological approach to communicative behavior

    Remi Tison and Pierre Poirier. Communication as socially extended active inference: An ecological approach to communicative behavior. Ecological Psychology, 33 0 (3-4): 0 197--235, 2021

  46. [54]

    A world unto itself: human communication as active inference

    Jared Vasil et al. A world unto itself: human communication as active inference. Frontiers in Psychology, 11: 0 417, 2020

  47. [55]

    Goal-directed empowerment: combining intrinsic motivation and task-oriented behavior

    Nicola Catenacci Volpi and Daniel Polani. Goal-directed empowerment: combining intrinsic motivation and task-oriented behavior. IEEE Transactions on Cognitive and Developmental Systems, 15 0 (2): 0 361--372, 2020

  48. [56]

    Swarm control for distributed construction: A computational complexity perspective

    Todd Wareham et al. Swarm control for distributed construction: A computational complexity perspective. ACM Transactions on Human-Robot Interaction, 12 0 (1): 0 1--45, 2023

  49. [57]

    Autonomous systems

    David P Watson and David H Scheidt. Autonomous systems. Johns Hopkins APL Technical Digest, 26 0 (4): 0 368--376, 2005

  50. [58]

    The theory of social and economic organization

    Max Weber. The theory of social and economic organization. Simon and Schuster, 2009

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.