Pith. sign in

REVIEW 3 major objections 4 minor 55 references

This paper introduces a computational-level theory of mind, ToM-U, which derives belief-like states from ordered information access history and source credibility via candidate world models evaluated against observed behavior.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-02 11:42 UTC pith:UBC7RMSD

load-bearing objection ToM-U is a genuinely new conceptual schema for epistemic state inference with clear limits: the load-bearing fit score is deliberately unspecified, so the 'formal derivation' claim overreaches, but as a computational-level schema it is worth a serious referee. the 3 major comments →

arxiv 2606.12721 v2 pith:UBC7RMSD submitted 2026-06-10 cs.AI

The Theory of Mind Utility: Formal Specification of a Mentalizing Mechanism

Nikolos Gurney , Stacy Marsella This is my paper
classification cs.AI MSC 68T2703B4268T30
keywords theory of mindmentalizingbelief inferenceepistemic stateworld modelcomputational levelgenerate-and-filterbounded rationality
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

ToM-U aims to formalize the epistemic state inference problem at the heart of theory of mind: how one agent infers what another believes. Rather than presupposing belief states as inputs (as Bayesian approaches do), the paper claims beliefs can be derived from who told the target what, in what order, and how credibly. The formal system represents this as a directed typed graph called a Local Epistemic World Model, evaluates discrete candidate models against observed behavior, and accumulates confidence until a threshold or a bounded recursion ceiling is reached. If correct, this gives the first formal account that derives belief states rather than assuming them, and it yields directional, falsifiable predictions about when mentalizing fails. The reader should care because this positions epistemic state inference as a distinct, upstream cognitive problem that other formal accounts leave unmodeled.

Core claim

The paper's central claim is that mentalizing reduces to a formalizable epistemic state inference problem: given a target agent's ordered information access history, source credibility, and observability constraints, the focal agent constructs and evaluates Local Epistemic World Models (LEWMs) — directed typed graphs whose edges carry belief-like states — and selects the candidate that best accounts for observed behavior. Five definitions specify the LEWM structure, agent node properties, a bounded proliferation mechanism for recursive mentalizing, three inference procedures (backward inference, self-projection, mutual reconciliation), and a residue function that records the structured trace

What carries the argument

The LEWM is the central object: a directed typed graph W = ⟨A, O, E, BLS, obs, cred⟩ where A is agents, O is state nodes, E is directed typed edges, BLS maps edges to belief-like states with a type and scalar value, obs maps edges to observability, and cred maps edges to accumulated inferential reliability. Agent nodes carry an ordered information access history H_i, a sophistication parameter S_i, and a snapshot timestamp. The proliferation mechanism builds a bounded branching tree of projected agents, treating nested selves as constructed others, and stops at a sophistication ceiling or when marginal coherence gain falls below a cost threshold. Inference proceeds by generate-and-filter: ca

Load-bearing premise

The load-bearing premise is that human belief formation is actually determined by ordered exposure history and source credibility — not merely content, recency, or salience — such that a model built on ordered information access can generatively produce the epistemic states we attribute to others.

What would settle it

A controlled behavioral experiment in which participants attribute beliefs to a target whose information sources are held constant but whose order is reversed (e.g., credible testimony contradicts a later label, or a salient signal appears first instead of second) would settle the claim if it shows attribution insensitive to ordering. Equally, a demonstration that repeated failed inferences about the same epistemic relationship do not reduce the reliance on that relationship in later judgments would falsify the residue function.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

If this is right

  • If ToM-U is right, belief states are outputs of a formal mechanism rather than inputs, so downstream models of goal inference and action understanding could consume epistemic state estimates from a shared, domain-agnostic utility.
  • The theory generates directional predictions about mentalizing failure modes: misrepresenting a target's exposure history produces systematically miscalibrated LEWMs, not randomly wrong ones; low observability yields certainty errors rather than content errors; sophistication miscalibration produces over- or under-recursive trees.
  • Falsifiable architectural commitments are identified: belief attribution insensitive to ordering would undermine the representational foundation; evidence that mentalizing is a continuous distribution rather than discrete candidate commitment would falsify generate-and-filter; evidence that failures leave no trace would falsify the residue function.
  • The distinction between absent and false beliefs offers a structural explanation for false-belief task performance differences, separating failure to represent from misrepresentation.
  • The theory connects to relevance theory and grounding in communication, suggesting that ordered H_i and mutual reconciliation formalize cognitive environment and joint understanding processes.

Where Pith is reading between the lines

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

  • A testable extension would be to compare ToM-U's predictions against Bayesian ToM in experimental tasks where the target's information access history is manipulated independently of the surface signal; ToM-U predicts order-sensitive belief attribution, whereas models that rely only on current perceptual access would not.
  • ToM-U's residue mechanism could be interpreted as a formal account of trust dynamics: repeated failures on the same epistemic relationship should reduce inferential reliance even when new evidence partially rehabilitates it, a prediction that could be probed in repeated social inference games.
  • One could extend the framework by instantiating the deliberately underspecified fit score and candidate generation functions at the algorithmic level (e.g., as a Bayesian sampler), then testing whether the resulting model reproduces the documented egocentric anchoring and adjustment patterns in perspective taking.
  • The paper leaves mid-construction abandonment without residue; specifying this boundary and its dynamics would allow the model to address cases where new information interrupts mentalizing and shapes later inferences.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. The paper presents ToM-U, a computational-level (Marr) theory of belief-like state inference in theory of mind. ToM-U constructs Local Epistemic World Models (LEWMs), directed typed graphs over agents and state nodes, where each agent node stores an ordered information access history H_i, a sophistication parameter, and a timestamp. Five definitions formalize (1) LEWM structure, (2) agent nodes, (3) bounded proliferation of projected agents, (4) three generate-and-filter inference procedures (backward inference, self-projection, mutual reconciliation), and (5) a residue function for failed mentalizing attempts. The paper claims that this is the first formal mechanism that derives, rather than presupposes, belief states from ordered exposure and source credibility, and that it yields falsifiable predictions about mentalizing failures. A popcorn-bag example illustrates how the machinery would attribute 'popcorn' when trusted testimony precedes a conflicting label.

Significance. ToM-U is a genuinely useful conceptual scaffold. It makes a persuasive case that ordered information access and source credibility are representational primitives that many formal accounts (BToM, DEL, AGM) leave implicit, and its distinction between absent and false beliefs is a substantive theoretical contribution. The nested-selves-as-others proliferation axiom is an elegant way to bound recursive mentalizing without a homunculus, and the residue construct gives failure a formal (if currently qualitative) role. The paper is also commendably explicit about its falsifiable commitments in §5.2. However, the current manuscript overstates what has been specified: the fit-score function, candidate-generation procedure, and residue decay are not defined, so ToM-U is not yet a function from observations to belief-state estimates; it is an architecture schema. With a completed scoring/generation layer — or with claims carefully reframed as schema-level — the framework could be a valuable contribution.

major comments (3)
  1. [§3.6.1, Eq. (10)-(11); §4.4; §5.5] The central derivational claim is not supported. Backward inference BI in Eq. (10) is described as a generate-and-filter procedure whose output is the candidate with highest f_j, but Step 1 leaves candidate generation and its ordering entirely unspecified, and f_j is 'deliberately underspecified' apart from the [0,1] bound. The early-stopping rule in Step 4 (terminate when C_n > τ_high) makes the output order-dependent, so even fixing f_j would not determine a unique result without a generation order. The worked example in §4.4 asserts that the popcorn candidate's fit score is higher; this is a narrative assumption, not a consequence of Definitions 1–4. The paper admits in §5.5 that no functional forms are provided. I do not dispute that one can define such an f_j; but the abstract's claim of a formal specification of 'what mentalizing computes' is too strong. Please either fix f_j and g
  2. [§3.6.3, Eq. (13)-(14)] Mutual reconciliation, the third inference procedure, contains an undefined step: 'Update W^{n+1}, advance snapshot timestamps, rebuild or prune proliferated tree as needed' (Step 4). In a formal computational-level specification, when and how the tree is rebuilt/pruned is part of the function, not an implementation detail; different pruning rules will yield different terminal outputs and different Cjoint. Additionally, the joint confidence Cjoint = min_i min(...) in Eq. (14) is presented as the only option, while the paper acknowledges an alternative weighted-average formulation. The choice is reasonable, but it is a theoretical assumption that itself needs justification or empirical grounding, not a derived result. As written, the multi-agent component of the theory is a sketch, not a specification.
  3. [§3.7, Eq. (15); §5.5] The residue function is not a function. Definition 5 states the signature R:E×R→[0,1] and five boundary conditions, but the 'specific functional form of R, including the mathematical structure of disconfirmation weighting, time decay, and rehabilitation, is deferred to the algorithmic level.' This deferral is problematic at the computational level: the rate and selectivity of residue accumulation determine how cred(e) changes and hence which future candidate LEWMs are viable. Those effects are part of what ToM-U computes, not merely how. The 'Monotonicity in disconfirmation' and 'Monotonicity in time' conditions are too weak to determine even ordinal predictions without specifying the relative weights of different rejection types. To make contribution 4 substantive, please provide a minimal recurrence for cred updates (or a precise family of update rules) and show its consequences for at
minor comments (4)
  1. [§4.3, Eq. (7)] The statement that b(1,3)=⌊3ρ⌋ 'yields a single branch for any ρ<1' is incorrect: for ρ∈[2/3,1), floor(3ρ)=2. Also, the assertion ΔC(2)<κ is not computed anywhere; it is a stipulation. Suggest specifying ρ<2/3 and either computing ΔC or labeling it an assumption.
  2. [§3.2 / Eq. (1)] Fit score f_j is defined as a scalar in [0,1] but never written as a function of the candidate BLS configuration, H_i, obs, and β_i. Adding notation such as f_j = F(BLS_j, H_i, obs, β_i) would prevent ambiguity in Eq. (11) and make the dependency structure explicit.
  3. [§2 / §5.5] The paper's level-of-analysis defense for leaving f_j and candidate generation unspecified would be stronger if it cited prior computational-level frameworks in which the objective function itself is left open. As written, the analogy to rational analysis is only partial, because a rational analysis normally specifies the utility/prior, not just the architecture.
  4. [§5.2] The claim that 'BLS types are not free parameters' is followed by a description of type attribution as a structured inference problem, but no formal criterion for type assignment is given. Consider adding a constraint (e.g., type determined by edge origin and H_i composition) or soften the claim.

Circularity Check

0 steps flagged

No significant circularity: the core LEWM/backward-inference formalism is self-contained, with only a minor non-load-bearing self-citation and an explicitly acknowledged under-specification of f_j.

full rationale

ToM-U's claimed derivation is the five definitions producing belief-like-state estimates from H_i, obs, cred, and β_i. Backward inference (Def. 4, Eq. 10) maps β_i × W_k to a candidate BLS with confidence C_n = 1 − ∏(1 − f_j). This mapping is not a single specified function because f_j and candidate generation are deliberately underspecified (§3.6.1) and §5.5 explicitly states: "it does not provide guidance for the functional forms of fit scoring, candidate generation, or residue decay." That is an incompleteness/underdetermination of the formal specification, not a circular reduction: the paper never fits f_j to data and then calls the result a prediction. In the worked example (§4.4) the higher fit score for the popcorn candidate is assigned informally, so the popcorn output is an illustrative choice rather than a theorem forced by the definitions. This weakens the strength of the claimed prediction but does not make the derivation circular. The only self-citation used to support a premise is [20] (an in-press paper by one of the authors) for the parent-process engagement assumption in §3.2. That assumption is upstream of the LEWM/backward-inference core and does not supply the content of the five definitions or the structural predictions; it is therefore a minor, non-load-bearing self-citation. No equation is defined in terms of its own output, no fitted parameter is renamed a prediction, and no uniqueness theorem is imported from the authors' prior work. External commitments such as source monitoring [24], bounded rationality [43], and relevance theory [44] are used as independent support rather than as the derivation. Score 2 reflects the minor self-citation plus the acknowledged under-specification; the central formal architecture has independent content.

Axiom & Free-Parameter Ledger

5 free parameters · 6 axioms · 3 invented entities

The central claim rests on several domain-specific axioms about how belief formation and mentalizing work, plus five free parameters (thresholds and decay) that are declared fixed but not calibrated. The invented entities are theoretical constructs without independent evidence.

free parameters (5)
  • tau_high
    Confidence threshold for early stopping, introduced in §3.2; treated as fixed, no calibrated value.
  • tau_low
    Lower bound for low-confidence output at sophistication ceiling, §3.2; fixed.
  • rho
    Branching decay parameter in b(k,S1)=floor(S1·rho^k), Definition 3 Eq (7); fixed but no numeric value.
  • kappa
    Proliferation cost threshold in stopping rule DeltaC(k)<kappa, Definition 3 Eq (9); positive real, fixed.
  • S_max = 6 (recommended)
    Empirical ceiling for sophistication normalization in §3.4 Eq (5), following Kinderman et al. [26]; provisional.
axioms (6)
  • domain assumption Belief states are determined by ordered exposure history and source credibility
    Stated in Section 1 as a foundational position; no empirical evidence supplied.
  • domain assumption Mentalizing proceeds by generating and filtering discrete candidate world models
    Section 1 and Definition 4; the framework is built on this generate-and-filter commitment.
  • domain assumption Recursive mentalizing is a bounded branching tree with nested-selves-as-others
    Definition 3; prevents infinite regress by construction.
  • domain assumption Failed mentalizing leaves a persistent residue that reduces edge credibility
    Definition 5; residue function is stipulated, not derived from other commitments.
  • domain assumption Computational-level analysis can defer functional forms to algorithmic level
    Section 2 and §3.6.1 justify leaving fit scoring, candidate generation, and residue shape unspecified.
  • domain assumption A parent process decides whether ToM-U is invoked and whether output is accepted
    Section 3.2 preliminary definitions; relies on the author's own in-press reference [20].
invented entities (3)
  • Local Epistemic World Model (LEWM) no independent evidence
    purpose: Directed typed graph used to represent and evaluate another's epistemic state
    The central new representational substrate; no direct measurement or external falsifiable handle provided.
  • Residue function R(e,t) no independent evidence
    purpose: Persistent trace of rejected LEWMs that reduces credibility of implicated edges
    Only boundary conditions specified; no empirical evidence that such a trace exists.
  • Projected agent nodes (nested-selves-as-others) no independent evidence
    purpose: Represent recursive ToM at each depth as distinct constructed others, avoiding homunculus regress
    Theoretical device in Definition 3; no independent evidence.

pith-pipeline@v1.3.0-alltime-deepseek · 19294 in / 15689 out tokens · 166877 ms · 2026-08-02T11:42:49.375699+00:00 · methodology

0 comments
Cite this review

Pith. "Pith review of The Theory of Mind Utility: Formal Specification of a Mentalizing Mechanism." pith.science (2026). https://pith.science/paper/UBC7RMSD

@misc{pith2026260612721,
  author       = {Pith},
  title        = {Pith review of: The Theory of Mind Utility: Formal Specification of a Mentalizing Mechanism},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UBC7RMSD}},
  note         = {Machine review of arXiv:2606.12721}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Inferring others' beliefs requires more than reading surface signals; it requires tracking who told them what, in what order, and how credibly. The Theory of Mind Utility (ToM-U) formalizes this epistemic state inference problem at the computational level of analysis, specifying what mentalizing computes and why without commitment to algorithmic or neural implementation. ToM-U achieves this by constructing Local Epistemic World Models (LEWMs) -- directed typed graphs that represent agents, state nodes, and the epistemic relationships among them -- and evaluating discrete candidate LEWMs against observed behavior until one achieves sufficient confidence. Five formal definitions specify the LEWM structure, agent node properties including ordered information access history, a bounded proliferation mechanism for recursive mentalizing, three inference procedures, and a residue function that captures the structured trace left by failed mentalizing attempts. ToM-U differs from Bayesian Theory of Mind and adjacent formal accounts, which presuppose rather than derive belief states, and from simulation theory and theory-theory, which lack a formal apparatus for epistemic state inference. The architecture generates directional, falsifiable predictions about mentalizing failure that follow from structural properties of the model rather than auxiliary assumptions, and positions ToM-U as a domain-agnostic mechanism upstream of goal inference and other downstream social cognitive processes.

Review history (2 revisions) →

discussion (0)

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

Reference graph

Works this paper leans on

55 extracted references · 2 linked inside Pith

  1. [1]

    On the logic of theory change: Partial meet contraction and revision functions

    Carlos E Alchourr´ on, Peter G¨ ardenfors, and David Makinson. “On the logic of theory change: Partial meet contraction and revision functions”. In:The journal of symbolic logic50.2 (1985), pp. 510–530

    1985

  2. [2]

    Meeting of minds: the medial frontal cortex and social cognition

    David M. Amodio and Chris D. Frith. “Meeting of minds: the medial frontal cortex and social cognition”. In:Nature reviews neuroscience 7.4 (2006), pp. 268–277

    2006

  3. [3]

    Do humans have two sys- tems to track beliefs and belief-like states?

    Ian A. Apperly and Stephen A. Butterfill. “Do humans have two sys- tems to track beliefs and belief-like states?” In:Psychological review 116.4 (2009), pp. 953–969. 34

    2009

  4. [4]

    Action un- derstanding as inverse planning

    Chris L. Baker, Rebecca Saxe, and Joshua B. Tenenbaum. “Action un- derstanding as inverse planning”. In:Cognition113.3 (2009), pp. 329– 349

    2009

  5. [5]

    Bayesian theory of mind: Modeling joint belief-desire attribution

    Chris L. Baker, Rebecca Saxe, and Joshua B. Tenenbaum. “Bayesian theory of mind: Modeling joint belief-desire attribution”. In:Proceed- ings of the annual meeting of the cognitive science society. Vol. 33. 33. 2011

    2011

  6. [6]

    Synchrony in the onset of mental-state reason- ing: Evidence from five cultures

    Tara Callaghan et al. “Synchrony in the onset of mental-state reason- ing: Evidence from five cultures”. In:Psychological Science16.5 (2005), pp. 378–384

    2005

  7. [7]

    Experience-weighted attraction learn- ing in normal form games

    Colin Camerer and Teck-Hua Ho. “Experience-weighted attraction learn- ing in normal form games”. In:Econometrica67.4 (1999), pp. 827–874

    1999

  8. [8]

    A cognitive hi- erarchy model of games

    Colin Camerer, Teck-Hua Ho, and Juin-Kuan Chong. “A cognitive hi- erarchy model of games”. In:The quarterly journal of economics119.3 (2004), pp. 861–898

    2004

  9. [9]

    The curse of knowledge in economic settings: An experimental analysis

    Colin Camerer, George Loewenstein, and Martin Weber. “The curse of knowledge in economic settings: An experimental analysis”. In:Journal of political Economy97.5 (1989), pp. 1232–1254

    1989

  10. [10]

    Grounding in communica- tion

    Herbert H. Clark and Susan E. Brennan. “Grounding in communica- tion”. In:Perspectives on Socially Shared Cognition. Ed. by Lauren B. Resnick, John M. Levine, and Stephanie D. Teasley. Washington, DC: American Psychological Association, 1991, pp. 127–149

    1991

  11. [11]

    Dennett.The intentional stance

    Daniel C. Dennett.The intentional stance. MIT press, 1989

    1989

  12. [12]

    Perspective taking as egocentric anchoring and adjustment

    Nicholas Epley et al. “Perspective taking as egocentric anchoring and adjustment.” In:Journal of personality and social psychology87.3 (2004), pp. 327–339

    2004

  13. [13]

    Jurist.Affect regulation, mentalization and the development of the self

    Peter Fonagy, Gyorgy Gergely, and Elliot L. Jurist.Affect regulation, mentalization and the development of the self. Routledge, 2018

    2018

  14. [14]

    Active inference: a process theory

    Karl Friston et al. “Active inference: a process theory”. In:Neural computation29.1 (2017), pp. 1–49

    2017

  15. [15]

    The spotlight effect in social judgment: an egocentric bias in estimates of the salience of one’s own actions and appearance

    Thomas Gilovich, Victoria Husted Medvec, and Kenneth Savitsky. “The spotlight effect in social judgment: an egocentric bias in estimates of the salience of one’s own actions and appearance”. In:Journal of per- sonality and social psychology78.2 (2000), pp. 211–222. 35

    2000

  16. [16]

    The illusion of transparency: biased assessments of others’ ability to read one’s emotional states

    Thomas Gilovich, Kenneth Savitsky, and Victoria Husted Medvec. “The illusion of transparency: biased assessments of others’ ability to read one’s emotional states”. In:Journal of personality and social psychology 75.2 (1998), pp. 332–346

    1998

  17. [17]

    Goldman.Simulating minds: The philosophy, psychology, and neuroscience of mindreading

    Alvin I. Goldman.Simulating minds: The philosophy, psychology, and neuroscience of mindreading. Oxford University Press, 2006

    2006

  18. [18]

    Why the child’s theory of mind really is a theory

    Alison Gopnik and Henry M. Wellman. “Why the child’s theory of mind really is a theory”. In:Mind & Language7.1–2 (1992), pp. 145–171

    1992

  19. [19]

    Folk psychology as simulation

    Robert M. Gordon. “Folk psychology as simulation”. In:Mind & lan- guage1.2 (1986), pp. 158–171

    1986

  20. [20]

    A Causal Model of Theory of Mind in Conflict for Artificial Intelligence

    Nikolos Gurney. “A Causal Model of Theory of Mind in Conflict for Artificial Intelligence”. In:HCI International 2026 – Late Breaking Work. Lecture Notes in Computer Science. Accepted, in press. Cham: Springer, 2026

    2026

  21. [21]

    World models

    David Ha and J¨ urgen Schmidhuber. “World models”. In:arXiv preprint arXiv:1803.10122(2018)

    Pith/arXiv arXiv 2018

  22. [22]

    Interpretation as abduction

    Jerry R. Hobbs et al. “Interpretation as abduction”. In:Artificial in- telligence63.1-2 (1993), pp. 69–142

    1993

  23. [23]

    The naive utility calculus as a unified, quantitative framework for action understanding

    Julian Jara-Ettinger, Laura E. Schulz, and Joshua B. Tenenbaum. “The naive utility calculus as a unified, quantitative framework for action understanding”. In:Cognitive Psychology123 (2020), p. 101334

    2020

  24. [24]

    Source monitoring

    Marcia K. Johnson, Shahin Hashtroudi, and D. Stephen Lindsay. “Source monitoring”. In:Psychological bulletin114.1 (1993), pp. 3–28

    1993

  25. [25]

    Johnson-Laird.Mental Models: Towards a Cognitive Science of Language, Inference, and Consciousness

    Philip N. Johnson-Laird.Mental Models: Towards a Cognitive Science of Language, Inference, and Consciousness. Cambridge, MA: Harvard University Press, 1983

    1983

  26. [26]

    Theory-of- mind deficits and causal attributions

    Peter Kinderman, Robin Dunbar, and Richard P. Bentall. “Theory-of- mind deficits and causal attributions”. In:British Journal of Psychol- ogy89.2 (1998), pp. 191–204

    1998

  27. [27]

    Toward a model of text compre- hension and production

    Walter Kintsch and Teun A. Van Dijk. “Toward a model of text compre- hension and production.” In:Psychological review85.5 (1978), pp. 363– 394. 36

    1978

  28. [28]

    Evaluating large language models in theory of mind tasks

    Michal Kosinski. “Evaluating large language models in theory of mind tasks”. In:Proceedings of the National Academy of Sciences121.45 (2024), e2405460121

    2024

  29. [29]

    The case for motivated reasoning

    Ziva Kunda. “The case for motivated reasoning”. In:Psychological Bul- letin108.3 (1990), pp. 480–498

    1990

  30. [30]

    Emotion and decision making

    Jennifer S. Lerner et al. “Emotion and decision making”. In:Annual review of psychology66.1 (2015), pp. 799–823

    2015

  31. [31]

    Reconstruction of auto- mobile destruction: An example of the interaction between language and memory

    Elizabeth F. Loftus and John C. Palmer. “Reconstruction of auto- mobile destruction: An example of the interaction between language and memory”. In:Journal of verbal learning and verbal behavior13.5 (1974), pp. 585–589

    1974

  32. [32]

    Cambridge, MA: MIT Press, 1982

    David Marr.Vision: A Computational Investigation into the Human Representation and Processing of Visual Information. Cambridge, MA: MIT Press, 1982

    1982

  33. [33]

    Toward a mechanistic psy- chology of dialogue

    Martin J. Pickering and Simon Garrod. “Toward a mechanistic psy- chology of dialogue”. In:Behavioral and Brain Sciences27.2 (2004), pp. 169–190

    2004

  34. [34]

    Early understanding of perception as a source of knowledge

    Bradford H. Pillow. “Early understanding of perception as a source of knowledge”. In:Journal of experimental child psychology47.1 (1989), pp. 116–129

    1989

  35. [35]

    Young children understand that looking leads to knowing (so long as they are looking into a single barrel)

    Chris Pratt and Peter Bryant. “Young children understand that looking leads to knowing (so long as they are looking into a single barrel)”. In: Child development61.4 (1990), pp. 973–982

    1990

  36. [36]

    Does the chimpanzee have a the- ory of mind?

    David Premack and Guy Woodruff. “Does the chimpanzee have a the- ory of mind?” In:Behavioral and brain sciences1.4 (1978), pp. 515– 526

    1978

  37. [37]

    Minimal mental models

    David V. Pynadath and Stacy C. Marsella. “Minimal mental models”. In:AAAI. Vol. 22. 2. 2007, pp. 1038–1044

    2007

  38. [38]

    PsychSim: Modeling the- ory of mind with decision-theoretic agents

    David V. Pynadath and Stacy C. Marsella. “PsychSim: Modeling the- ory of mind with decision-theoretic agents”. In:IJCAI. Vol. 5. 2005, pp. 1181–1186

    2005

  39. [39]

    Bayesian brains without proba- bilities

    Adam N. Sanborn and Nick Chater. “Bayesian brains without proba- bilities”. In:Trends in cognitive sciences20.12 (2016), pp. 883–893. 37

    2016

  40. [40]

    People thinking about thinking people: the role of the temporo-parietal junction in “theory of mind

    Rebecca Saxe and Nancy Kanwisher. “People thinking about thinking people: the role of the temporo-parietal junction in “theory of mind””. In:Social neuroscience. Psychology Press, 2013, pp. 171–182

    2013

  41. [41]

    The seven sins of memory: Insights from psychol- ogy and cognitive neuroscience

    Daniel L. Schacter. “The seven sins of memory: Insights from psychol- ogy and cognitive neuroscience”. In:American psychologist54.3 (1999), pp. 182–203

    1999

  42. [42]

    Culture and the sequence of steps in the- ory of mind development

    Ameneh Shahaeian et al. “Culture and the sequence of steps in the- ory of mind development”. In:Developmental psychology47.5 (2011), pp. 1239–1247

    2011

  43. [43]

    A behavioral model of rational choice

    Herbert A. Simon. “A behavioral model of rational choice”. In:Quar- terly Journal of Economics69.1 (1955), pp. 99–118

    1955

  44. [44]

    Dan Sperber and Deirdre Wilson.Relevance: Communication and cog- nition. Vol. 142. Harvard University Press Cambridge, MA, 1986

    1986

  45. [45]

    Decision by sam- pling

    Neil Stewart, Nick Chater, and Gordon D.A. Brown. “Decision by sam- pling”. In:Cognitive psychology53.1 (2006), pp. 1–26

    2006

  46. [46]

    Explanatory coherence

    Paul Thagard. “Explanatory coherence”. In:Behavioral and brain sci- ences12.3 (1989), pp. 435–467

    1989

  47. [47]

    Large language models fail on trivial alterations to theory-of-mind tasks

    Tomer Ullman. “Large language models fail on trivial alterations to theory-of-mind tasks”. In:arXiv preprint arXiv:2302.08399(2023)

    Pith/arXiv arXiv 2023

  48. [48]

    Van Dijk and Walter Kintsch.Strategies of discourse compre- hension

    Teun A. Van Dijk and Walter Kintsch.Strategies of discourse compre- hension. New York: Academic Press, 1983.isbn: 9780127120508

    1983

  49. [49]

    Springer, 2008

    Hans Van Ditmarsch, Wiebe van Der Hoek, and Barteld Kooi.Dynamic epistemic logic. Springer, 2008

    2008

  50. [50]

    One and done? Optimal decisions from very few samples

    Edward Vul et al. “One and done? Optimal decisions from very few samples”. In:Cognitive science38.4 (2014), pp. 599–637

    2014

  51. [51]

    Meta-analysis of theory-of-mind development: The truth about false belief

    Henry M. Wellman, David Cross, and Julanne Watson. “Meta-analysis of theory-of-mind development: The truth about false belief”. In:Child development72.3 (2001), pp. 655–684

    2001

  52. [52]

    From simple desires to ordinary beliefs: The early development of everyday psychology

    Henry M. Wellman and Jacqueline D. Woolley. “From simple desires to ordinary beliefs: The early development of everyday psychology”. In:Cognition35.3 (1990), pp. 245–275

    1990

  53. [53]

    Wellman et al.The child’s theory of mind

    Henry M. Wellman et al.The child’s theory of mind. The MIT Press, 1990. 38

    1990

  54. [54]

    Beliefs about beliefs: Representation and constraining function of wrong beliefs in young children’s under- standing of deception

    Heinz Wimmer and Josef Perner. “Beliefs about beliefs: Representation and constraining function of wrong beliefs in young children’s under- standing of deception”. In:Cognition13.1 (1983), pp. 103–113

    1983

  55. [55]

    Event perception: a mind-brain perspective

    Jeffrey M. Zacks et al. “Event perception: a mind-brain perspective”. In:Psychological bulletin133.2 (2007), pp. 273–293. 39

    2007