pith. machine review for the scientific record. sign in

arxiv: 2604.19838 · v2 · submitted 2026-04-21 · 💻 cs.AI

Recognition: 2 theorem links

· Lean Theorem

Resolving space-sharing conflicts in road user interactions through uncertainty reduction: An active inference-based computational model

Arkady Zgonnikov, Jens Kober, Johan Engstr\"om, Julian F. Schumann, Ran Wei, Shu-Yuan Liu

Authors on Pith no claims yet

Pith reviewed 2026-05-12 04:06 UTC · model grok-4.3

classification 💻 cs.AI
keywords active inferenceroad user interactionsuncertainty reductionnormative expectationsexplicit communicationspace-sharing conflictstraffic safetyautonomous vehicles
0
0 comments X

The pith

An active inference model shows normative and explicit cues raise successful resolution of road conflicts but increase collisions if agents violate expectations.

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

The paper extends an active inference model of driver behavior to simulate two agents interacting at an intersection. It models three ways to reduce uncertainty: implicit cues from direct movements, following traffic norms like stop signs, and explicit signals. Simulations show that norms and explicit communication raise the odds of safe space-sharing, yet this advantage disappears when one agent breaks the rules or sends false signals, leading to more collisions. Understanding these dynamics supports safer traffic systems and autonomous vehicle design by revealing when reliance on expectations succeeds or fails.

Core claim

Extending active inference to two-agent road interactions captures three complementary uncertainty-reduction mechanisms: implicit communication through behavioral coupling, reliance on normative expectations such as priority rules, and explicit communication. In simplified intersection scenarios, normative and explicit cues increase successful conflict resolution, but violations of expectations or misleading signals by one agent raise collision likelihood.

What carries the argument

Active inference, where agents minimize uncertainty by updating beliefs and selecting actions, extended to multi-agent settings through three uncertainty reduction mechanisms for social interactions.

Load-bearing premise

That the simulated agents act according to normative expectations and that the three mechanisms accurately represent how real humans reduce uncertainty in intersections.

What would settle it

Measure collision rates in human driving simulator studies at intersections where one agent violates a stop sign or gives misleading signals, and check whether rates rise compared to conditions where all agents follow expectations.

Figures

Figures reproduced from arXiv: 2604.19838 by Arkady Zgonnikov, Jens Kober, Johan Engstr\"om, Julian F. Schumann, Ran Wei, Shu-Yuan Liu.

Figure 1
Figure 1. Figure 1: The modeled scenario: two agents on crossing paths representing a space-sharing conflict with ∆D denoting agent B’s lead over agent A (i.e., ∆D < 0 if A is closer to the intersection). Below it, the basic action-perception (hermeneutic) cycle resulting from modeling two active inference agents observing each other (based on Friston and Frith32) probabilistic belief q(s) over s). While it is possible to ass… view at source ↗
Figure 2
Figure 2. Figure 2: The likelihoods of different simulation outcomes under different model setups, for different initial differences in distance to the intersection ∆D(0). “With norms” refers to models considering both first-come priority rules and stop signs. The vertical lines indicate the 95% confidence intervals of the shown probabilities according to the Wilson score51 . agent to always go first. As the difference in ini… view at source ↗
Figure 3
Figure 3. Figure 3: Three illustrative examples of model behavior in the baseline scenario: the agents interact based on implicit communication only. For each example simulation, the panels illustrate (from top to bottom) the top-down view at a specified instant t, the difference in distance ∆D(t) = DB(t) − DA(t) (∆D > 0 m means that agent B is closer to the intersection), the agents’ velocities v, their accumulated evidence … view at source ↗
Figure 4
Figure 4. Figure 4: Three example interactions of the model endowed with normative expectations (priority rule and stop signs). a) The two agents start at an equal distance to the intersection, both agents consider the less-aggressive braking agent A as the leading agent which allows A (after stopping) to pass the intersection first. b) The two agents start at an equal distance to the intersection, but due to lagging percepti… view at source ↗
Figure 5
Figure 5. Figure 5: An example scenario for the interaction in which agents could use explicit communication. In addition to variables presented in previous examples, the explicit communication signals are shown (prompting γA and yielding γY have slightly offset x-axis values for better visibility); shaded regions denote the range in the other agent’s beliefs about each signal. Both agents start at equal distance to the inter… view at source ↗
Figure 6
Figure 6. Figure 6: Three example interactions of the model equipped with explicit communication and normative expectations. a) At an intersection where agent A is leading (∆D(0) = −3 m), it manages to pass first, solely relying on the normative expectation, with communication not necessary. b) The two agents start at an equal distance. As both agents assign themselves priority, they do not consider signaling yielding. In the… view at source ↗
Figure 7
Figure 7. Figure 7: presents an example of such interaction. There, at the first re-plan (t = 4.4 s, first dashed line), convinced by the yielding signal of the other vehicle, the leading agent A considers light acceleration to be sufficient to cross the intersection safely. The collision part of the pragmatic values falls by over an order of magnitude, and A’s predictions of ∆D show high confidence that the gap will become w… view at source ↗
Figure 8
Figure 8. Figure 8: a) The likelihoods of the different simulation outcomes in the provoked conflict scenario, with error bars indicating 95% confidence intervals. b) The time tE>1 at which the agent A first triggered a full re-plan of its policy (lower panel). Error bars show the 5th and 95th percentiles. signal, agent A is able to choose a policy which avoids a collision. Discussion In this paper, we extended an existing ac… view at source ↗
read the original abstract

Understanding how road users resolve space-sharing conflicts is important both for traffic safety and the safe deployment of autonomous vehicles. While existing models have captured specific aspects of such interactions (e.g., explicit communication), a theoretically-grounded computational framework has been lacking. In this paper, we extend a previously developed active inference-based driver behavior model to simulate interactive behavior of two agents. Our model captures three complementary mechanisms for uncertainty reduction in interaction: (i) implicit communication via direct behavioral coupling, (ii) reliance on normative expectations (stop signs, priority rules, etc.), and (iii) explicit communication. In a simplified intersection scenario, we show that normative and explicit communication cues can increase the likelihood of a successful conflict resolution. However, this relies on agents acting as expected. In situations where another agent (intentionally or unintentionally) violates normative expectations or communicates misleading information, reliance on these cues may induce collisions. These findings illustrate how active inference can provide a novel framework for modeling road user interactions which is also applicable in other fields.

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

1 major / 2 minor

Summary. The paper extends a prior active-inference driver model to two agents interacting at a simplified intersection. It implements three uncertainty-reduction mechanisms—implicit behavioral coupling, normative expectations (e.g., stop signs, priority rules), and explicit communication—and reports simulation outcomes showing that normative and explicit cues raise the probability of successful space-sharing conflict resolution, while violations or misleading signals can produce collisions. The work positions active inference as a general framework for such interactions.

Significance. The integration of active inference with the three specified uncertainty-reduction channels supplies a coherent computational account that could, if empirically grounded, inform AV interaction design and traffic-safety modeling. The manuscript correctly notes the dependence on agents acting according to normative expectations and thereby avoids over-claiming generality; however, the absence of any calibration, trajectory matching, or out-of-sample validation against human data keeps the quantitative effect sizes tied to the chosen generative-model priors and likelihoods.

major comments (1)
  1. [Simulation results / abstract] The central simulation result—that normative and explicit cues increase successful-resolution likelihood—is demonstrated only under the assumption that both agents encode and act on the normative expectations exactly as encoded in the generative model (see abstract and the simulation section). No empirical calibration or comparison to human trajectory/decision data is reported, so the reported probability gains cannot yet be treated as predictions about real road users.
minor comments (2)
  1. [Model section] Clarify the precise functional forms and any free parameters used for the three uncertainty-reduction mechanisms; the current description leaves the implementation details opaque.
  2. [Discussion] Add a short paragraph distinguishing the scope of the simulation findings from any intended real-world implications, to prevent readers from over-generalizing the collision-risk result.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their constructive review. We address the single major comment below, clarifying the scope and assumptions of our computational study while offering a targeted revision to strengthen the existing qualifications.

read point-by-point responses

  1. Referee: [Simulation results / abstract] The central simulation result—that normative and explicit cues increase successful-resolution likelihood—is demonstrated only under the assumption that both agents encode and act on the normative expectations exactly as encoded in the generative model (see abstract and the simulation section). No empirical calibration or comparison to human trajectory/decision data is reported, so the reported probability gains cannot yet be treated as predictions about real road users.

    Authors: We agree that the reported increases in successful-resolution probability are obtained under the assumption that both agents share and act upon the normative expectations encoded in the generative model. This assumption is already stated in the abstract (final sentence) and discussed at length in Section 4. The manuscript presents a theoretical simulation study whose goal is to demonstrate how active inference supplies a unified account of three uncertainty-reduction channels; it does not claim to deliver calibrated predictions or effect sizes for real human road users. No trajectory matching or out-of-sample validation against human data was performed, precisely because the contribution is the normative framework itself rather than an empirical model. We will add one clarifying sentence to the discussion (and update the abstract if the editor prefers) to reiterate that all quantitative outcomes are illustrative and prior-dependent, thereby preventing any misreading as empirical forecasts. revision: partial

Circularity Check

0 steps flagged

No significant circularity; simulations are independent demonstrations of an extended model

full rationale

The paper extends a prior active-inference driver model to a two-agent intersection scenario and reports simulation outcomes under explicit normative and communication assumptions. No load-bearing step reduces a claimed result to a fitted parameter, self-citation chain, or definitional equivalence; the quantitative effects are generated by running the generative model forward rather than by construction from data subsets or prior outputs. Self-citation of the base model is present but does not carry the central claim, which remains falsifiable against external human trajectory data.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on the active inference framework from prior literature plus assumptions that agents use the three listed mechanisms and that normative expectations are shared. No new free parameters or invented entities are introduced in the abstract.

axioms (2)
  • domain assumption Agents minimize uncertainty about other agents' intentions and actions through active inference
    Core premise of the extended model stated in the abstract.
  • domain assumption Normative expectations (stop signs, priority rules) are reliably shared among agents
    Invoked when claiming that normative cues increase successful resolution.

pith-pipeline@v0.9.0 · 5497 in / 1253 out tokens · 56386 ms · 2026-05-12T04:06:53.356866+00:00 · methodology

Review history (2 revisions) →

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

63 extracted references · 63 canonical work pages · 1 internal anchor

  1. [1]

    Fatal intersection crashes in Norway: Patterns in contributing factors and data collection challenges.Accident; analysis and prevention2012; 45: 782–91

    Ljung Aust M, Fagerlind H and Sagberg F. Fatal intersection crashes in Norway: Patterns in contributing factors and data collection challenges.Accident; analysis and prevention2012; 45: 782–91. DOI:10.1016/j.aap.2011.11.001

    work page doi:10.1016/j.aap.2011.11.001 2011

  2. [2]

    Driver behavior in car-to-pedestrian incidents: An application of the driving reliability and error analysis method (dream).Accident Analysis & Prevention2013; 50: 554–565

    Habibovic A, Tivesten E, Uchida N et al. Driver behavior in car-to-pedestrian incidents: An application of the driving reliability and error analysis method (dream).Accident Analysis & Prevention2013; 50: 554–565

    work page

  3. [3]

    Vehicle automation– other road user communication and coordination: Theory and mechanisms.IEEE Access2020; 8: 19860–19872

    Domeyer JE, Lee JD and Toyoda H. Vehicle automation– other road user communication and coordination: Theory and mechanisms.IEEE Access2020; 8: 19860–19872

    work page

  4. [4]

    Editorial to the special issue on road user interactions in the age of vehicle automation, 2026

    Lee YM and Domeyer J. Editorial to the special issue on road user interactions in the age of vehicle automation, 2026

    work page 2026

  5. [5]

    Defining interactions: a conceptual framework for understanding interactive behaviour in human and automated road traffic

    Markkula G, Madigan R, Nathanael D et al. Defining interactions: a conceptual framework for understanding interactive behaviour in human and automated road traffic. Theoretical Issues in Ergonomics Science2020; DOI:10.1080/ 1463922X.2020.1736686

    work page arXiv 2020

  6. [6]

    Proceedings of first workshop on traffic conflicts.Oslo, TTI, Oslo, Norway and LTH Lund, Sweden1977; 78

    Amundsen F and Hyden C. Proceedings of first workshop on traffic conflicts.Oslo, TTI, Oslo, Norway and LTH Lund, Sweden1977; 78

    work page

  7. [7]

    Drivers’ commu- nicative interactions: on-road observations and modelling for integration in future automation systems.Ergonomics2014; 57(12): 1795–1805

    Portouli E, Nathanael D and Marmaras N. Drivers’ commu- nicative interactions: on-road observations and modelling for integration in future automation systems.Ergonomics2014; 57(12): 1795–1805

    work page

  8. [8]

    Grounding in communication

    Clark HH and Brennan SE. Grounding in communication. Perspectives on socially shared cognition1991

    work page

  9. [9]

    Sadigh D, Landolfi N, Sastry SS et al. Planning for cars that coordinate with people: leveraging effects on human actions for planning and active information gathering over human internal state.Autonomous Robots2018; 42(7): 1405–1426. DOI: 10.1007/s10514-018-9746-1

    work page doi:10.1007/s10514-018-9746-1

  10. [10]

    Proxemics and kinesics in automated vehicle–pedestrian communication: Representing ethnographic observations.Transportation research record2019; 2673(10): 70–81

    Domeyer J, Dinparastdjadid A, Lee JD et al. Proxemics and kinesics in automated vehicle–pedestrian communication: Representing ethnographic observations.Transportation research record2019; 2673(10): 70–81

    work page

  11. [11]

    External human-machine interfaces on automated vehicles: Effects on pedestrian crossing decisions.Human factors2019; 61(8): 1353–1370

    De Clercq K, Dietrich A, N ´u˜nez Velasco JP et al. External human-machine interfaces on automated vehicles: Effects on pedestrian crossing decisions.Human factors2019; 61(8): 1353–1370

    work page

  12. [12]

    Road users rarely use explicit communication when interacting in today’s traffic: implications for automated vehicles.Cognition, Technology & Work2021; 23(2): 367–380

    Lee YM, Madigan R, Giles O et al. Road users rarely use explicit communication when interacting in today’s traffic: implications for automated vehicles.Cognition, Technology & Work2021; 23(2): 367–380

    work page

  13. [13]

    Interdependence in vehicle-pedestrian encounters and its implications for vehicle automation.IEEE transactions on intelligent transportation systems2020; 23(5): 4122–4134

    Domeyer JE, Lee JD, Toyoda H et al. Interdependence in vehicle-pedestrian encounters and its implications for vehicle automation.IEEE transactions on intelligent transportation systems2020; 23(5): 4122–4134

    work page

  14. [14]

    Interdependence of driver and pedestrian behavior in naturalistic roadway negotiations.Traffic injury prevention2022; 23(sup1): S62– S67

    Noonan TZ, Gershon P, Domeyer J et al. Interdependence of driver and pedestrian behavior in naturalistic roadway negotiations.Traffic injury prevention2022; 23(sup1): S62– S67

    work page

  15. [15]

    Driver-pedestrian perceptual models demonstrate coupling: Implications for vehicle automation.IEEE Transactions on Human-Machine Systems2022; 52(4): 557–566

    Domeyer JE, Lee JD, Toyoda H et al. Driver-pedestrian perceptual models demonstrate coupling: Implications for vehicle automation.IEEE Transactions on Human-Machine Systems2022; 52(4): 557–566

    work page

  16. [16]

    Towards pedestrian- av interaction: method for elucidating pedestrian preferences

    Camara F, Cosar S, Bellotto N et al. Towards pedestrian- av interaction: method for elucidating pedestrian preferences. InIEEE International Workshop on Intelligent Robots and Systems (IROS)

    work page

  17. [17]

    Pedestrian Models for Autonomous Driving Part II: High-Level Models of Human Behavior.IEEE Transactions on Intelligent Transportation Systems2021; 22(9): 5453–5472

    Camara F, Bellotto N, Cosar S et al. Pedestrian Models for Autonomous Driving Part II: High-Level Models of Human Behavior.IEEE Transactions on Intelligent Transportation Systems2021; 22(9): 5453–5472. DOI:10.1109/TITS.2020. 3006767

    work page doi:10.1109/tits.2020 2020

  18. [18]

    When should the chicken cross the road.Game theory for autonomous vehicle-human interactions2018

    Fox C, Camara F, Markkula G et al. When should the chicken cross the road.Game theory for autonomous vehicle-human interactions2018

    work page

  19. [19]

    Hierarchical game-theoretic planning for autonomous vehicles

    Fisac JF, Bronstein E, Stefansson E et al. Hierarchical game-theoretic planning for autonomous vehicles. In2019 International conference on robotics and automation (ICRA). IEEE, pp. 9590–9596

    work page

  20. [20]

    Game theoretical approach to model decision making for merging maneuvers at freeway on-ramps

    Kang K and Rakha HA. Game theoretical approach to model decision making for merging maneuvers at freeway on-ramps. Transportation Research Record2017; 2623(1): 19–28

    work page

  21. [21]

    Kirkpatrick,et al., Overcoming catastrophic forgetting in neural networks.Proceedings of the National Academy of Sciences114(13), 3521–3526 (2017), doi:10.1073/pnas

    Schwarting W, Pierson A, Alonso-Mora J et al. Social behavior for autonomous vehicles.Proceedings of the National Academy of Sciences2019; : 201820676DOI:10.1073/pnas. 1820676116

    work page doi:10.1073/pnas

  22. [22]

    Modeling driver behavior in interactions with other road users

    Thalya P, Kovaceva J, Knauss A et al. Modeling driver behavior in interactions with other road users. InProceedings of 8th Transport Research Arena. Prepared usingsagej.cls 16 Collective Intelligence XX(X)

    work page

  23. [23]

    A computational driver model to predict driver control at unsignalised intersections

    Boda CN, Lehtonen E and Dozza M. A computational driver model to predict driver control at unsignalised intersections. IEEE Access2020; 8: 104619–104631

    work page

  24. [24]

    At the zebra crossing: Modelling complex decision processes with variable- drift diffusion models

    Giles O, Markkula G, Pekkanen J et al. At the zebra crossing: Modelling complex decision processes with variable- drift diffusion models. InProceedings of the Annual Meeting of the Cognitive Science Society, volume 41

    work page

  25. [25]

    Evidence Accumulation Account of Human Operators’ Decisions in Intermittent Control During Inverted Pendulum Balancing

    Markkula G and Zgonnikov A. Evidence Accumulation Account of Human Operators’ Decisions in Intermittent Control During Inverted Pendulum Balancing. InIEEE International Conference on Systems, Man, and Cybernetics (SMC). Miyazaki, Japan. ISBN 978-1-5386-6650-0, pp. 716–

    work page

  26. [26]

    DOI:10.1109/SMC.2018.00130

    work page doi:10.1109/smc.2018.00130 2018

  27. [27]

    Variable-Drift Diffusion Models of Pedestrian Road-Crossing Decisions

    Pekkanen J, Giles OT, Lee YM et al. Variable-Drift Diffusion Models of Pedestrian Road-Crossing Decisions. Computational Brain & Behavior2022; 5(1): 60–80. DOI: 10.1007/s42113-021-00116-z

    work page doi:10.1007/s42113-021-00116-z

  28. [28]

    Should I Stay or Should I Go? Cognitive Modeling of Left- Turn Gap Acceptance Decisions in Human Drivers

    Zgonnikov A, Abbink D and Markkula G. Should I Stay or Should I Go? Cognitive Modeling of Left- Turn Gap Acceptance Decisions in Human Drivers. Human Factors2024; : 00187208221144561DOI: 10.1177/00187208221144561. URLhttps://doi. org/10.1177/00187208221144561. Publisher: SAGE Publications Inc

    work page doi:10.1177/00187208221144561

  29. [29]

    Explaining human interactions on the road by large-scale integration of computational psychological theory.PNAS Nexus2023; 2(6): pgad163

    Markkula G, Lin YS, Srinivasan AR et al. Explaining human interactions on the road by large-scale integration of computational psychological theory.PNAS Nexus2023; 2(6): pgad163. DOI:10.1093/pnasnexus/pgad163. URLhttps: //doi.org/10.1093/pnasnexus/pgad163

    work page doi:10.1093/pnasnexus/pgad163

  30. [30]

    Zone-specific interaction modeling of pedestrians and cars in shared spaces.Transportation research procedia2020; 47: 251–258

    Johora FT and M ¨uller JP. Zone-specific interaction modeling of pedestrians and cars in shared spaces.Transportation research procedia2020; 47: 251–258

    work page

  31. [31]

    A model of dyadic merging interactions explains human drivers’ behavior from control inputs to decisions.PNAS Nexus2024; 3(10): pgae420

    Siebinga O, Zgonnikov A and Abbink DA. A model of dyadic merging interactions explains human drivers’ behavior from control inputs to decisions.PNAS Nexus2024; 3(10): pgae420. DOI:10.1093/pnasnexus/pgae420. URLhttps: //doi.org/10.1093/pnasnexus/pgae420

    work page doi:10.1093/pnasnexus/pgae420

  32. [32]

    Modelling communication-enabled traffic interactions.Royal Society open science2023; 10(5)

    Siebinga O, Zgonnikov A and Abbink DA. Modelling communication-enabled traffic interactions.Royal Society open science2023; 10(5)

    work page

  33. [33]

    Active inference, communication and hermeneutics.cortex2015; 68: 129–143

    Friston KJ and Frith CD. Active inference, communication and hermeneutics.cortex2015; 68: 129–143

    work page

  34. [34]

    A duet for one.Consciousness and cognition2015; 36: 390–405

    Friston K and Frith C. A duet for one.Consciousness and cognition2015; 36: 390–405

    work page

  35. [35]

    A Predic- tive Processing Framework for Joint Action and Commu- nication, 2026

    Pezzulo G, Knoblich G, Maisto D et al. A Predic- tive Processing Framework for Joint Action and Commu- nication, 2026. URLhttps://osf.io/preprints/ psyarxiv/q4jnr_v4/

    work page 2026

  36. [36]

    Oxford University Press, 2016

    Bicchieri C.Norms in the wild: How to diagnose, measure, and change social norms. Oxford University Press, 2016

    work page 2016

  37. [37]

    Being good (at driving): Characterizing behavioral expectations on automated and human driven vehicles.arXiv preprint arXiv:250208121 2025

    Fraade-Blanar L, Favar `o F, Engstr ¨om J et al. Being good (at driving): Characterizing behavioral expectations on automated and human driven vehicles.arXiv preprint arXiv:250208121 2025

    work page 2025

  38. [38]

    A variational approach to niche construction.Journal of the Royal Society Interface2018; 15(141)

    Constant A, Ramstead MJ, Veissiere SP et al. A variational approach to niche construction.Journal of the Royal Society Interface2018; 15(141)

    work page

  39. [39]

    Free-energy minimization in joint agent-environment systems: A niche construction perspective.Journal of theoretical biology2018; 455: 161–178

    Bruineberg J, Rietveld E, Parr T et al. Free-energy minimization in joint agent-environment systems: A niche construction perspective.Journal of theoretical biology2018; 455: 161–178

    work page

  40. [40]

    Regimes of expectations: An active inference model of social conformity and human decision making.Frontiers in psychology2019; 10: 679

    Constant A, Ramstead MJ, Veissi `ere SP et al. Regimes of expectations: An active inference model of social conformity and human decision making.Frontiers in psychology2019; 10: 679

    work page

  41. [41]

    Neural Computation , volume=

    Friston K, FitzGerald T, Rigoli F et al. Active Inference: A Process Theory.Neural Computation2017; 29(1): 1–49. DOI: 10.1162/NECO a 00912. URLhttps://doi.org/10. 1162/NECO_a_00912

    work page doi:10.1162/neco

  42. [42]

    MIT Press, 2022

    Parr T, Pezzulo G and Friston KJ.Active Inference: The Free Energy Principle in Mind, Brain, and Behavior. MIT Press, 2022. ISBN 978-0-262-04535-3. Google-Books-ID: UrZNEAAAQBAJ

    work page 2022

  43. [43]

    Resolving uncertainty on the fly: modeling adaptive driving behavior as active inference.Frontiers in Neurorobotics 2024; 18

    Engstr ¨om J, Wei R, McDonald AD et al. Resolving uncertainty on the fly: modeling adaptive driving behavior as active inference.Frontiers in Neurorobotics 2024; 18. DOI:10.3389/fnbot.2024.1341750. URL https://www.frontiersin.org/articles/ 10.3389/fnbot.2024.1341750. Publisher: Frontiers

    work page doi:10.3389/fnbot.2024.1341750 2024

  44. [44]

    Active inference as a unified model of collision avoidance behavior in human drivers, 2025

    Schumann JF, Engstr ¨om J, Johnson L et al. Active inference as a unified model of collision avoidance behavior in human drivers, 2025. DOI:10.48550/arXiv.2506.02215. URLhttp: //arxiv.org/abs/2506.02215. ArXiv:2506.02215 [cs]

    work page internal anchor Pith review doi:10.48550/arxiv.2506.02215 2025

  45. [45]

    A World Unto Itself: Human Communication as Active Inference.Frontiers in Psychology2020; 11

    Vasil J, Badcock PB, Constant A et al. A World Unto Itself: Human Communication as Active Inference.Frontiers in Psychology2020; 11. DOI:10.3389/fpsyg.2020.00417. URLhttps://www.frontiersin.org/journals/ psychology/articles/10.3389/fpsyg.2020. 00417/full. Publisher: Frontiers

    work page doi:10.3389/fpsyg.2020.00417 2020

  46. [46]

    Traffic safety and norms of compliance with rules: An exploratory study.Economics Bulletin2021; 41(4): 2464–2483

    Laurent H, Sangnier M and Treibich C. Traffic safety and norms of compliance with rules: An exploratory study.Economics Bulletin2021; 41(4): 2464–2483. URL http://www.scopus.com/inward/record.url? scp=85125263034&partnerID=8YFLogxK

    work page

  47. [47]

    Deceleration parameters as implicit communication signals for pedestrians’ crossing decisions and estimations of automated vehicle behaviour

    Tian K, Tzigieras A, Wei C et al. Deceleration parameters as implicit communication signals for pedestrians’ crossing decisions and estimations of automated vehicle behaviour. Accident Analysis & Prevention2023; 190: 107173

    work page

  48. [48]

    Pedestrian interaction with vehicles: roles of explicit and implicit communication

    Dey D and Terken J. Pedestrian interaction with vehicles: roles of explicit and implicit communication. InProceedings of the 9th international conference on automotive user interfaces and interactive vehicular applications. pp. 109–113

    work page

  49. [49]

    Miller L, Leitner J, Kraus J et al. Implicit intention communication as a design opportunity for automated vehicles: Understanding drivers’ interpretation of vehicle trajectory at narrow passages.Accident Analysis & Prevention2022; 173: 106691

    work page

  50. [50]

    Nudging human drivers via implicit communication by automated vehicles: Empirical evidence and computational cognitive modeling

    Zgonnikov A, Beckers N, George A et al. Nudging human drivers via implicit communication by automated vehicles: Empirical evidence and computational cognitive modeling. International Journal of Human-Computer Studies2024; 185: 103224

    work page

  51. [51]

    Routledge, 2017

    Goffman E.Interaction ritual: Essays in face-to-face behavior. Routledge, 2017

    work page 2017

  52. [52]

    Probable inference, the law of succession, and statistical inference.Journal of the American Statistical Association1927; 22(158): 209–212

    Wilson EB. Probable inference, the law of succession, and statistical inference.Journal of the American Statistical Association1927; 22(158): 209–212

    work page

  53. [53]

    Pedestrian-driver communication and decision strategies at marked crossings

    Sucha M, Dostal D and Risser R. Pedestrian-driver communication and decision strategies at marked crossings. Accident Analysis & Prevention2017; 102: 41–50. Prepared usingsagej.cls Schumannet al. 17

    work page

  54. [54]

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

    Maisto D, Donnarumma F and Pezzulo G. Interactive inference: A multi-agent model of cooperative joint actions. IEEE Transactions on Systems, Man, and Cybernetics: Systems 2023; 54(2): 704–715

    work page 2023

  55. [55]

    Road users rarely use explicit communication when interacting in today’s traffic: implications for automated vehicles.Cognition, Technology & Work2021; 23(2): 367–380

    Lee YM, Madigan R, Giles O et al. Road users rarely use explicit communication when interacting in today’s traffic: implications for automated vehicles.Cognition, Technology & Work2021; 23(2): 367–380. DOI:10.1007/ s10111-020-00635-y

    work page

  56. [56]

    Formal scenario-based testing of autonomous vehicles: From simulation to the real world

    Fremont DJ, Kim E, Pant YV et al. Formal scenario-based testing of autonomous vehicles: From simulation to the real world. In2020 IEEE 23rd International Conference on Intelligent Transportation Systems (ITSC). IEEE, pp. 1–8

    work page

  57. [57]

    The Waymo Open Sim Agents Challenge.Advances in Neural Information Processing Systems2023; 36: 59151–59171

    Montali N, Lambert J, Mougin P et al. The Waymo Open Sim Agents Challenge.Advances in Neural Information Processing Systems2023; 36: 59151–59171. URLhttps://proceedings. neurips.cc/paper_files/paper/2023/hash/ b96ce67b2f2d45e4ab315e13a6b5b9c5-Abstract-Datasets_ and_Benchmarks.html

    work page 2023

  58. [58]

    Testing autonomous vehicles in virtual environments: A review of simulation tools and techniques.Control Syst Optim Lett2025; 3: 151–157

    Uzzaman A, Islam M and Hossain MS. Testing autonomous vehicles in virtual environments: A review of simulation tools and techniques.Control Syst Optim Lett2025; 3: 151–157

    work page

  59. [59]

    A model of the sidewalk salsa

    Siebinga O. A model of the sidewalk salsa. In2025 IEEE International Conference on Systems, Man, and Cybernetics (SMC). IEEE, pp. 7298–7305

    work page

  60. [60]

    Drivers’ knowledge, attitude and practices towards traffic rules and regulations.Journal of Road Safety2024; 35(3): 24–31

    Abbas S, Fatima S, Sharif A et al. Drivers’ knowledge, attitude and practices towards traffic rules and regulations.Journal of Road Safety2024; 35(3): 24–31

    work page

  61. [61]

    Information Particle Filter Tree: An Online Algorithm for POMDPs with Belief-Based Rewards on Continuous Domains

    Fischer J and Tas OS. Information Particle Filter Tree: An Online Algorithm for POMDPs with Belief-Based Rewards on Continuous Domains. InProceedings of the 37th International Conference on Machine Learning. PMLR, pp. 3177–3187. URLhttps://proceedings.mlr. press/v119/fischer20a.html. ISSN: 2640-3498

    work page

  62. [62]

    A Tutorial on the Cross-Entropy Method.Annals of Operations Research 2005; 134(1): 19–67

    de Boer PT, Kroese DP, Mannor S et al. A Tutorial on the Cross-Entropy Method.Annals of Operations Research 2005; 134(1): 19–67. DOI:10.1007/s10479-005-5724-z. URL https://doi.org/10.1007/s10479-005-5724-z

    work page doi:10.1007/s10479-005-5724-z 2005

  63. [63]

    first come, first go

    Dinparastdjadid A, Supeene I and Engstrom J. Measuring Surprise in the Wild, 2023. DOI:10.48550/arXiv.2305. 07733. URLhttp://arxiv.org/abs/2305.07733. ArXiv:2305.07733 [cs]. Prepared usingsagej.cls 18 Collective Intelligence XX(X) Appendix Model implementation In the original model 43, the kinematic state of an agentv∈Vis defined asx v ={x v, yv, θv, δv, ...

    work page doi:10.48550/arxiv.2305 2023