pith. sign in

arxiv: 2604.03387 · v1 · submitted 2026-04-03 · 💻 cs.AI

Hume's Representational Conditions for Causal Judgment: What Bayesian Formalization Abstracted Away

Yiling Wu This is my paper

Pith reviewed 2026-05-13 19:43 UTC · model grok-4.3

classification 💻 cs.AI
keywords Humecausal judgmentBayesian epistemologypredictive processingrepresentational conditionsvivacitylarge language modelsformalization
0
0 comments X

The pith

Hume's causal judgment depends on experiential grounding, structured retrieval, and vivacity transfer, conditions that Bayesian epistemology and predictive processing have abstracted away.

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

The paper extracts from Hume's texts three representational conditions that his account of causal judgment presupposes: ideas must trace back to impressions, associations must operate through organized networks rather than mere pairwise links, and inferences must produce a felt sense of conviction rather than bare probability updates. These conditions are presented as integral to Hume's causal psychology. The argument then follows the formalization path from Hume through Bayesian epistemology and predictive processing, showing that later frameworks retain the core updating mechanism while dropping the representational requirements. Large language models are used to illustrate the result, as they perform statistical updating on text without satisfying any of the three conditions, making visible what earlier frameworks had treated as background.

Core claim

Hume's account of causal judgment presupposes three representational conditions—experiential grounding (ideas must trace to impressions), structured retrieval (association must operate through organized networks exceeding pairwise connection), and vivacity transfer (inference must produce felt conviction, not merely updated probability)—that are integral to his causal psychology but abstracted away in the formalization trajectory from Hume to Bayesian epistemology and predictive processing.

What carries the argument

The three representational conditions (experiential grounding, structured retrieval, and vivacity transfer) that carry the argument by specifying what must be preserved for a judgment to count as genuinely causal in Hume's sense.

If this is right

  • Bayesian epistemology preserves only the probabilistic updating structure of Hume's insight while discarding the representational conditions.
  • Predictive processing frameworks continue the same abstraction by focusing on prediction error minimization without requiring experiential grounding or vivacity transfer.
  • Large language models perform statistical updating on linguistic data yet fail all three conditions, thereby exposing the requirements that were previously implicit in Hume's framework.
  • Any account of causal judgment that drops the conditions risks reducing inference to bare correlation tracking.

Where Pith is reading between the lines

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

  • Contemporary AI systems built on statistical learning may therefore generate outputs that look like causal reasoning while lacking the psychological features Hume took to be essential.
  • Attempts to build more human-like causal reasoners could test whether adding mechanisms for experiential grounding or vivacity transfer changes model behavior on tasks that distinguish Humean from Bayesian inference.
  • The paper's contrast between Hume and later formalisms suggests that representational constraints may be needed in any model that aims to replicate not just the accuracy but the phenomenology of causal judgment.

Load-bearing premise

That Hume's texts treat the three conditions as load-bearing presuppositions for causal judgment rather than optional background features.

What would settle it

A close reading of Hume's texts showing that causal judgment can proceed without one or more of the three conditions, or an LLM-based causal judgment that produces the same felt conviction and structured associations Hume requires.

Figures

Figures reproduced from arXiv: 2604.03387 by Yiling Wu.

Figure 1
Figure 1. Figure 1: Hume’s dual-layer model of mental representation, illustrating the distinction between [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Hume’s account of causal inference, showing the pathway from perception through [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Hume’s model of how the idea of necessary connection arises. The impression of [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
read the original abstract

Hume's account of causal judgment presupposes three representational conditions: experiential grounding (ideas must trace to impressions), structured retrieval (association must operate through organized networks exceeding pairwise connection), and vivacity transfer (inference must produce felt conviction, not merely updated probability). This paper extracts these conditions from Hume's texts and argues that they are integral to his causal psychology. It then traces their fate through the formalization trajectory from Hume to Bayesian epistemology and predictive processing, showing that later frameworks preserve the updating structure of Hume's insight while abstracting away these further representational conditions. Large language models serve as an illustrative contemporary case: they exhibit a form of statistical updating without satisfying the three conditions, thereby making visible requirements that were previously background assumptions in Hume's framework.

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 / 3 minor

Summary. The paper extracts three representational conditions from Hume's texts—experiential grounding (ideas trace to impressions), structured retrieval (associations operate via organized networks beyond pairwise links), and vivacity transfer (inference yields felt conviction)—as integral to his causal psychology. It traces their abstraction in the formalization from Hume through Bayesian epistemology and predictive processing, using LLMs as an illustrative case of statistical updating that lacks these conditions.

Significance. If the interpretive extraction holds, the work usefully identifies representational requirements that formal models have backgrounded, offering a bridge between Humean psychology and contemporary AI. The textual grounding of the conditions and the contrast with LLMs provide a clear, falsifiable illustration of what the abstraction omits, strengthening the paper's contribution to philosophy of cognitive science.

major comments (1)
  1. [§4] §4 (trajectory section): the argument that Bayesian models abstract away vivacity transfer would be strengthened by a direct comparison to a specific formulation (e.g., a cited Bayesian network or predictive-processing equation) showing where felt conviction is replaced by probability updating alone.
minor comments (3)
  1. [Abstract and §1] The abstract and introduction use 'structured retrieval' without an explicit contrast to Hume's associationism in the opening paragraphs; a brief definitional sentence would improve readability.
  2. [§5] LLM illustration in §5: the claim that models 'exhibit statistical updating without satisfying the three conditions' would benefit from one concrete example (e.g., a failure mode in causal inference) rather than remaining at the level of general contrast.
  3. [Throughout] A short table or bullet list summarizing the three conditions and their status in each framework (Hume, Bayesian, predictive processing, LLM) would aid comparison.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the supportive summary and recommendation of minor revision. The suggestion to strengthen §4 with a concrete formal comparison is helpful and will be incorporated.

read point-by-point responses

  1. Referee: [§4] §4 (trajectory section): the argument that Bayesian models abstract away vivacity transfer would be strengthened by a direct comparison to a specific formulation (e.g., a cited Bayesian network or predictive-processing equation) showing where felt conviction is replaced by probability updating alone.

    Authors: We agree that an explicit side-by-side comparison would make the abstraction of vivacity transfer more precise. In the revised version we will add a short paragraph in §4 that contrasts Hume’s vivacity transfer with (i) the standard Bayesian update P(H|E) = P(E|H)P(H)/P(E) in a simple causal network and (ii) the minimization of variational free energy in predictive-processing accounts. The text will note that both replace the phenomenological “felt conviction” with a purely quantitative measure of posterior probability or expected information gain, without any counterpart to Hume’s transfer of vivacity from impression to idea. revision: yes

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper performs an interpretive extraction of three conditions from Hume's texts and traces their absence in subsequent formal frameworks via textual comparison and illustrative contrast with LLMs. No equations, fitted parameters, self-citations, or ansatzes appear in the derivation chain; the central claim rests on direct quotation and philosophical analysis of the source material rather than any reduction to its own inputs by construction. The argument is therefore self-contained against external textual benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on interpretive assumptions about Hume's texts and the representational properties of Bayesian models and LLMs; no free parameters or invented entities are introduced.

axioms (1)
  • domain assumption Hume's texts presuppose experiential grounding, structured retrieval, and vivacity transfer as integral to causal judgment
    Stated as extracted from Hume's texts in the abstract.

pith-pipeline@v0.9.0 · 5414 in / 1173 out tokens · 46195 ms · 2026-05-13T19:43:19.447633+00:00 · methodology

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

7 extracted references · 7 canonical work pages

  1. [1]

    Climbing towards NLU: On Meaning, Form, and Understanding in the Age of Data

    Bender, Emily M., and Alexander Koller. “Climbing towards NLU: On Meaning, Form, and Understanding in the Age of Data.”Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics(2020): 5185-98. Blackburn, Simon. “Hume and Thick Connexions.”Philosophy and Phenomenological Research50 (1990): 237-50. Broughton, Janet. “Explaining...

    work page 2020

  2. [2]

    Bayesian Cognitive Science, Unification, and Explana- tion

    Colombo, Matteo, and Stephan Hartmann. “Bayesian Cognitive Science, Unification, and Explana- tion.”British Journal for the Philosophy of Science68 (2017): 451-84. Earman, John.Bayes or Bust? A Critical Examination of Bayesian Confirmation Theory. Cam- bridge, MA: MIT Press,

    work page 2017

  3. [3]

    Bayes, Hume, and Miracles

    Earman, John. “Bayes, Hume, and Miracles.”Faith and Philosophy10 (1993): 293-310. Everson, Stephen. “The Difference between Feeling and Thinking.”Mind97 (1988): 401-13. Flage, Daniel E. “Hume’s Relative Ideas.”Hume Studies7 (1981): 55-73. Fodor, Jerry A.Hume Variations. Oxford: Clarendon Press,

    work page 1993

  4. [4]

    The Free-Energy Principle: A Unified Brain Theory?

    Friston, Karl. “The Free-Energy Principle: A Unified Brain Theory?”Nature Reviews Neuroscience 11 (2010): 127-36. Garrett, Don.Cognition and Commitment in Hume’s Philosophy. New York: Oxford University Press,

    work page 2010

  5. [5]

    General Rules in Hume’sTreatise

    16 Hearn, Thomas K. “General Rules in Hume’sTreatise.”Journal of the History of Philosophy8 (1970): 405-22. Hohwy, Jakob.The Predictive Mind. Oxford: Oxford University Press,

    work page 1970

  6. [6]

    Causal reasoning and large language models: Opening a new frontier for causality

    Kıcıman, Emre, Robert Ness, Amit Sharma, and Chenhao Tan. “Causal Reasoning and Large Language Models: Opening a New Frontier for Causality.” arXiv preprint arXiv:2305.00050 (2023). Kirchhoff, Michael D., and Tom Froese. “Where There Is Life There Is Mind: In Support of a Strong Life-Mind Continuity Thesis.”Entropy19 (2017):

    work page arXiv 2023

  7. [7]

    General Rules and the Justification of Probable Belief in Hume’sTreatise

    Lyons, Jack C. “General Rules and the Justification of Probable Belief in Hume’sTreatise.”Hume Studies27 (2001): 247-77. Mahowald, Kyle, Anna A. Ivanova, Idan A. Blank, Nancy Kanwisher, Joshua B. Tenenbaum, and Evelina Fedorenko. “Dissociating Language and Thought in Large Language Models: A Cognitive Perspective.”Trends in Cognitive Sciences28 (2024): 51...

    work page 2001