pith. sign in

arxiv: 1907.08097 · v1 · pith:B32JKI3Vnew · submitted 2019-07-18 · ⚛️ physics.bio-ph · cond-mat.soft· q-bio.NC

Intrinsically motivated collective motion

Henry J. Charlesworth , Matthew S. Turner This is my paper

Pith reviewed 2026-05-24 19:13 UTC · model grok-4.3

classification ⚛️ physics.bio-ph cond-mat.softq-bio.NC
keywords collective motionfuture state maximisationintrinsic motivationvisual perceptionemergent behavioranimal groupsactive matterneural network
0
0 comments X

The pith

Collective motion behaviors such as cohesion and alignment emerge when agents maximize the variety of future visual environments they expect to access.

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

The paper establishes that features of collective motion seen in animals arise spontaneously from Future State Maximisation, a principle in which agents move to increase the number of distinct visual scenes they anticipate encountering later. No explicit rules for group alignment, cohesion or collision avoidance are built into the model. Agents use a retinal-style visual input to select actions that expand their expected future options. If this principle holds, it supplies a direct route from individual control to group behaviors that could confer fitness in uncertain settings and could be implemented in artificial sensing and moving systems.

Core claim

Under Future State Maximisation, agents that perceive a visual representation of their surroundings and move to maximise the number of different visual environments they expect to access in the future produce collective dynamics that include cohesion, co-alignment and collision suppression. None of these features are explicitly encoded. A multi-layered neural network trained on the resulting trajectories learns a heuristic that approximates the same control rule, indicating that comparable reasoning is within reach of animal cognition and offering a candidate encoding for future realisations of artificial intelligent matter.

What carries the argument

Future State Maximisation (FSM), the rule by which agents select moves that increase the expected number of distinct future visual environments based on simple retinal input.

If this is right

  • Cohesion, co-alignment and collision suppression appear without any explicit encoding of those properties.
  • A neural network can be trained on FSM trajectories to produce a computationally lighter heuristic that still generates the same group behaviors.
  • The same control rule supplies a candidate mechanism for the appearance of collective motion in social animals.
  • Models based on FSM are candidates for encoding into artificial systems that sense light, process information and move.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • FSM could be implemented on physical robots to test whether the same qualitative group features appear outside simulation.
  • The visual-diversity objective might be compared with other intrinsic-motivation objectives to identify which ones reliably produce alignment and cohesion.
  • Varying the angular resolution or field of view of the retinal input would reveal how sensitive the emergent behaviors are to perceptual detail.
  • The principle could be examined in heterogeneous groups where only a subset of agents follow FSM to determine whether the collective features persist.

Load-bearing premise

Maximizing the number of different visual environments agents expect to access in the future is a control principle that confers evolutionary fitness and produces the observed collective behaviors.

What would settle it

A simulation in which agents governed by FSM do not form cohesive groups, align their directions, or reduce collisions relative to random motion would contradict the emergence result.

Figures

Figures reproduced from arXiv: 1907.08097 by Henry J. Charlesworth, Matthew S. Turner.

Figure 1
Figure 1. Figure 1: Sketch showing an agent’s movement options, a representation of the visual state of the world around it and its future decision-tree. (a) The five actions available to each agent at every time step, given that its previous move was in the direction of the dashed line, continue in the same direction at a nominal/slow/fast speed, or turn left/right, respectively. (b) A representative agent (red) sees the oth… view at source ↗
Figure 2
Figure 2. Figure 2: (a) Structure of collective swarms that emerge under FSM dynamics, as described in [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Snapshots of a swarm made up of N = 500 agents with τ = 5, shown at different times in a frame co-moving with the swarm’s centre of mass. Panel (a) shows the initial state of the swarm and then (b), (c) and (d) show snapshots of its subsequent evolution (in chronological order). In this example we use a continuous measure of visual degeneracy (see SI for details). The full simulation is shown in SI Movie 4… view at source ↗
Figure 4
Figure 4. Figure 4: Convergence of heuristic A (order targeting, blue) and B (topological Vicsek (5), red) to a value of the order parameter that is self-consistent with the value realized by FSM in each case. An initial (iteration 0) order parameter for the heuristic (φA and φB, respectively) is chosen. This parameterises the model of all (other) agents to be used when constructing their trajectories into the future in order… view at source ↗
Figure 5
Figure 5. Figure 5: Training a neural network as a heuristic approximating FSM. (a) Sketch of the network architecture. The network takes as its input the agent’s current speed and the state of each sensor in both the current and previous time steps, represented as light and dark blue squares on each sensor (left). This is then passed through four hidden layers of neurons of sizes 200, 100, 50 and 25 which have RelU activatio… view at source ↗
read the original abstract

Collective motion is found in various animal systems, active suspensions and robotic or virtual agents. This is often understood using high level models that directly encode selected empirical features, such as co-alignment and cohesion. Can these features be shown to emerge from an underlying, low-level principle? We find that they emerge naturally under Future State Maximisation (FSM). Here agents perceive a visual representation of the world around them, such as might be recorded on a simple retina, and then move to maximise the number of different visual environments that they expect to be able to access in the future. Such a control principle may confer evolutionary fitness in an uncertain world by enabling agents to deal with a wide variety of future scenarios. The collective dynamics that spontaneously emerge under FSM resemble animal systems in several qualitative aspects, including cohesion, co-alignment and collision suppression, none of which are explicitly encoded in the model. A multi-layered neural network trained on simulated trajectories is shown to represent a heuristic mimicking FSM. Similar levels of reasoning would seem to be accessible under animal cognition, demonstrating a possible route to the emergence of collective motion in social animals directly from the control principle underlying FSM. Such models may also be good candidates for encoding into possible future realisations of artificial "intelligent" matter, able to sense light, process information and move.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript claims that collective motion features such as cohesion, co-alignment and collision suppression emerge spontaneously in simulations of agents that select actions to maximise the number of distinct future visual states they expect to access (Future State Maximisation, FSM). None of these features are explicitly encoded in the model. The work further shows that a multi-layer neural network trained on FSM trajectories can approximate the resulting policy, offering a possible cognitively plausible route to such behaviour in animals.

Significance. If the emergence result holds under quantitative scrutiny, the paper supplies a single, low-level, parameter-free control principle that could explain the spontaneous appearance of collective motion across biological and artificial systems. The visual-perception framing and the neural-network approximation add biological plausibility and suggest applications to engineered active matter.

major comments (2)
  1. [Model definition / Methods] The central claim that cohesion, co-alignment and collision suppression arise without explicit encoding rests on the simulation results. The manuscript must supply the precise mathematical definition of the FSM objective (how the expected number of distinct future visual states is computed and optimised) together with the agent dynamics and visual representation; without these the independence from the observed features cannot be verified.
  2. [Results / Simulation analysis] The resemblance to animal systems is asserted on qualitative grounds only. Quantitative order parameters (e.g., global polarisation, pair-correlation functions, collision frequency) and direct comparisons with established models (Vicsek, Couzin et al.) are required to substantiate the claim that the emergent dynamics are comparable to empirical data.
minor comments (2)
  1. [Introduction / Discussion] The evolutionary-fitness interpretation is presented as a possible rationale rather than a required step; it should be clearly separated from the mechanistic emergence argument.
  2. [Figures] Figure captions and axis labels should explicitly state the visual-field resolution, number of agents, and simulation time steps used in each panel.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive review and for recommending major revision. The suggested clarifications and quantitative analyses will strengthen the manuscript. We address each major comment below and will incorporate the necessary revisions.

read point-by-point responses

  1. Referee: [Model definition / Methods] The central claim that cohesion, co-alignment and collision suppression arise without explicit encoding rests on the simulation results. The manuscript must supply the precise mathematical definition of the FSM objective (how the expected number of distinct future visual states is computed and optimised) together with the agent dynamics and visual representation; without these the independence from the observed features cannot be verified.

    Authors: We agree that explicit mathematical definitions are required for independent verification. While the manuscript describes the FSM objective, agent dynamics and visual representation in the Methods, we will add the full set of equations (including the precise computation of expected distinct future visual states and the optimisation procedure) to ensure the independence claim can be fully assessed. revision: yes

  2. Referee: [Results / Simulation analysis] The resemblance to animal systems is asserted on qualitative grounds only. Quantitative order parameters (e.g., global polarisation, pair-correlation functions, collision frequency) and direct comparisons with established models (Vicsek, Couzin et al.) are required to substantiate the claim that the emergent dynamics are comparable to empirical data.

    Authors: We acknowledge that the current presentation relies on qualitative resemblance. In the revised manuscript we will introduce quantitative order parameters (global polarisation, pair-correlation functions, collision frequency) and perform direct numerical comparisons against the Vicsek model and the Couzin et al. model to provide a more rigorous substantiation of the emergent dynamics. revision: yes

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper defines the FSM objective independently as maximizing the number of distinct future visual states accessible to agents, then demonstrates via simulation that collective behaviors (cohesion, alignment, collision avoidance) emerge without being explicitly encoded. No equations, derivations, or claims reduce these outcomes to self-definitions, fitted inputs renamed as predictions, or load-bearing self-citations. The evolutionary-fitness interpretation is offered only as a possible rationale, not as a required step in the emergence argument. The central claim rests on observable simulation results from the stated control principle.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

Based solely on the abstract; full details on any parameters or additional assumptions in the model are not available.

axioms (2)
  • domain assumption Agents perceive a visual representation of the world around them such as might be recorded on a simple retina
    This is the basis for the FSM control principle as stated in the abstract.
  • domain assumption Maximizing the number of different visual environments accessible in the future is a viable and fitness-conferring control principle
    Central to the model and its evolutionary interpretation.

pith-pipeline@v0.9.0 · 5763 in / 1166 out tokens · 24064 ms · 2026-05-24T19:13:18.198392+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

35 extracted references · 35 canonical work pages · 2 internal anchors

  1. [1]

    intelligent matter

    where it is used to aid exploration in environments where rewards are sparse. The key principle is that such behaviour should offer a generic and universal benefit to the agent, not because it is useful for solving any one particular problem, but because it is beneficial for a wide range of scenarios that the agent may encounter in the future. A similar idea...

    work page 2019

  2. [2]

    Valadares, Yu-Guo Tao, Nicole S

    Leonardo F . Valadares, Yu-Guo Tao, Nicole S. Zacharia, Fernando Galembeck, Raymond Kapral, and Geoffrey A. Ozin. Catalytic nanomotors: Self-propelled sphere dimers. Small, 6: 565–572, 2010

    work page 2010

  3. [3]

    Pine, and Paul M

    Jeremie Palacci, Stefano Sacanna, Asher Preska Steinberg, David J. Pine, and Paul M. Chaikin. Living crystals of light-activated colloidal surfers. Science, 339(6122):936–940, 2013

    work page 2013

  4. [4]

    Active matter

    Sriram Ramaswamy. Active matter. Journal of Statistical Mechanics: Theory and Experiment, 2017(5):054002, may 2017. . URL https://doi.org/10.1088%2F1742-5468%2Faa6bc5

    work page 2017

  5. [5]

    Cates and Julien Tailleur

    Michael E. Cates and Julien Tailleur. When are active brownian particles and run-and-tumble particles equivalent? consequences for motility-induced phase separation.EPL (Europhysics Letters), 101(2):20010, 2013

    work page 2013

  6. [6]

    Relevance of metric-free interactions in flocking phe- nomena

    Hugues Chaté and Francesco Ginelli. Relevance of metric-free interactions in flocking phe- nomena. Phys. Rev. Lett., 105:168103, 2010

    work page 2010

  7. [7]

    M. C. Marchetti, J. F . Joanny, S. Ramaswamy, T. B. Liverpool, J. Prost, Madan Rao, and R. Aditi Simha. Hydrodynamics of soft active matter. Rev. Mod. Phys., 85:1143–1189, 2013

    work page 2013

  8. [8]

    Long-range order in a two-dimensional dynamicalXY model: How birds fly together

    John Toner and Yuhai Tu. Long-range order in a two-dimensional dynamicalXY model: How birds fly together. Phys. Rev. Lett., 75:4326–4329, 1995

    work page 1995

  9. [9]

    Boltzmann and hydrodynamic description for self-propelled particles

    Eric Bertin, Michel Droz, and Guillaume Grégoire. Boltzmann and hydrodynamic description for self-propelled particles. Phys. Rev. E, 74:022101, Aug 2006. . URL https://link.aps.org/ doi/10.1103/PhysRevE.74.022101

    work page doi:10.1103/physreve.74.022101 2006

  10. [10]

    Scale-free correlations in starling flocks

    A Cavagna et al. Scale-free correlations in starling flocks. Proceedings of the National Academy of Sciences, 107(26):11865–11870, 2010

    work page 2010

  11. [11]

    Couzin, and Martin Wikelski

    Andrea Flack, Máté Nagy, Wolfgang Fiedler, Iain D. Couzin, and Martin Wikelski. From local collective behavior to global migratory patterns in white storks. Science, 360(6391):911–914, 2018

    work page 2018

  12. [12]

    Zitterbart, Barbara Wienecke, James P

    Daniel P . Zitterbart, Barbara Wienecke, James P . Butler, and Ben Fabry. Coordinated move- ments prevent jamming in an emperor penguin huddle. PLoS ONE, 6, 2011

    work page 2011

  13. [13]

    Social force model for pedestrian dynamics

    Dirk Helbing and Péter Molnár. Social force model for pedestrian dynamics. Phys. Rev. E, 51:4282–4286, 1995

    work page 1995

  14. [14]

    How simple rules determine pedestrian behavior and crowd disasters

    Mehdi Moussaïd, Dirk Helbing, and Guy Theraulaz. How simple rules determine pedestrian behavior and crowd disasters. Proceedings of the National Academy of Sciences , 108(17): 6884–6888, 2011. ISSN 0027-8424. . URL https://www.pnas.org/content/108/17/6884

    work page 2011

  15. [15]

    Ballerini, N

    M. Ballerini, N. Cabibbo, R. Candelier, A. Cavagna, E. Cisbani, I. Giardina, V. Lecomte, A. Or- landi, G. Parisi, A. Procaccini, M. Viale, and V. Zdravkovic. Interaction ruling animal collective behavior depends on topological rather than metric distance: Evidence from a field study. Proceedings of the National Academy of Sciences, 105(4):1232–1237, 2008

    work page 2008

  16. [16]

    Hildenbrandt, C

    H. Hildenbrandt, C. Carere, and C.K. Hemelrijk. Self-organized aerial displays of thousands of starlings: a model. Behavioral Ecology, 21:1349–1359, 2010

    work page 2010

  17. [17]

    Dan J. G. Pearce, Adam M. Miller, George Rowlands, and Matthew S. Turner. Role of projec- tion in the control of bird flocks. Proceedings of the National Academy of Sciences, 111(29): 10422–10426, 2014

    work page 2014

  18. [18]

    Gallup, Joseph J

    Andrew C. Gallup, Joseph J. Hale, David J. T. Sumpter, Simon Garnier, Alex Kacelnik, John R. Krebs, and Iain D. Couzin. Visual attention and the acquisition of information in human crowds. Proceedings of the National Academy of Sciences , 109(19):7245–7250, 2012. ISSN 0027-

    work page 2012

  19. [19]

    A. S. Kluybin, D Polani, and C. L. Nehaniv. Empowerment: A universal agent-centric measure of control. In The 2005 IEEE Congress on Evolutionary Computation vol. 1, pages 128–135, 2005

    work page 2005

  20. [20]

    P Capdepuy, Daniel Polani, and C. L. Nehaniv. Maximization of potential information flow as a universal utility for collective behaviour. InIEEE Symposium on Artificial Life, pages 207–213. IEEE, 2007

    work page 2007

  21. [21]

    P Capdepuy, Daniel Polani, and C. L. Nehaniv. Perception-action loops of multiple agents: informational aspects and the impact of coordination.Theory in Biosciences, 131(3):149–159, 2012

    work page 2012

  22. [22]

    Changing the environment based on empowerment as intrinsic motivation

    Christopher Salge, Cornelius Glackin, and Daniel Polani. Changing the environment based on empowerment as intrinsic motivation. Entropy, 16, 2014

    work page 2014

  23. [23]

    Intrinsically Motivated Learning in Natural and Arti- ficial Systems

    Gianluca Baldassarre and Marco Mirolli. Intrinsically Motivated Learning in Natural and Arti- ficial Systems. Springer, 2013

    work page 2013

  24. [24]

    Robert W. White. Motivation reconsidered: The concept of competence. Psychological Re- view, 66(5), 1959

    work page 1959

  25. [25]

    Ryan and Edward L

    Richard M. Ryan and Edward L. Deci. Intrinsic and extrinsic motivations: Classic definitions and new directions. Contemporary Educational Psychology, 25:54–67, 2000

    work page 2000

  26. [26]

    A. G. Barto, S Singh, and N Chentanez. Intrinsically motivated learning of hierarchical collec- tions of skills. In The 3rd International Conference on Developmental Learning, 2004

    work page 2004

  27. [27]

    Shakir Mohamed and Daniel J. Rezende. Variational information maximisation for intrinsically motivated reinforcement learning. In Advances in Neural Information Processing Systems , pages 2116–2124, 2015

    work page 2015

  28. [28]

    Diversity is All You Need: Learning Skills without a Reward Function

    Benjamin Eysenbach, Abhishek Gupta, Julian Ibarz, and Sergey Levine. Diversity is all you need: Learning diverse skills without a reward function. arXiv preprint arXiv:1802.06070 , 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  29. [29]

    Wissner-Gross and C

    Alexander D. Wissner-Gross and C. E. Freer. Causal entropic forces.Physical Review Letters, 110:168702, 2013

    work page 2013

  30. [30]

    Hannes Hornischer, Stephan Herminghaus, and Marco G. Mazza. Intelligence of agents pro- duces a structural phase transition in collective behaviour. arXiv preprint arXiv:1706.01458, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  31. [31]

    Mann and Roman Garnett

    Richard P . Mann and Roman Garnett. The entropic basis of collective behaviour. Journal of The Royal Society Interface, 12(106), 2015. ISSN 1742-5689

    work page 2015

  32. [32]

    Active motion of a janus particle by self-thermophoresis in a defocused laser beam

    Hong-Ren Jiang, Natsuhiko Y oshinaga, and Masaki Sano. Active motion of a janus particle by self-thermophoresis in a defocused laser beam. Phys. Rev. Lett., 105:268302, 2010

    work page 2010

  33. [33]

    Goldstein

    Vasily Kantsler, Jörn Dunkel, Marco Polin, and Raymond E. Goldstein. Ciliary contact inter- actions dominate surface scattering of swimming eukaryotes. Proceedings of the National Academy of Sciences, 110(4):1187–1192, 2013

    work page 2013

  34. [34]

    Tim Sanchez, Daniel T. N. Chen, Stephen J. DeCamp, Michael Heymann, and Zvonimir Dogic. Spontaneous motion in hierarchically assembled active matter. Nature, 491, 2012

    work page 2012

  35. [35]

    Are biological systems poised at criticality? Journal of Statistical Physics, 144:268–302, 2011

    Thierry Mora and William Bialek. Are biological systems poised at criticality? Journal of Statistical Physics, 144:268–302, 2011. 6 | Charlesworth et al

    work page 2011