Pith. sign in

REVIEW 4 major objections 6 minor 21 references

Realism about the external world: an adversarial collaboration

T0 review · 4 major / 6 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read An adversarial collaboration argues that Algorithmic Idealism resolves the Boltzmann brain paradox by making induction, not microstate counting, the arbiter of what you should expect next.

desk verdict A clear, honest adversarial collaboration that sharpens Algorithmic Idealism, but the Boltzmann-brain resolution still rests on a deferred inequality and an unformalized typicality claim. read the letter →

arxiv 2607.16379 v1 pith:S65GYKFS submitted 2026-07-17 physics.hist-ph

Kelvin J. McQueen , Markus P. Mueller This is my paper · ORCID
classification physics.hist-ph MSC 03A0503D3268Q30
keywords AlgorithmicIdealismBoltzmannbrainadversarialcollaborationSolomonoffinductionrealismvsself-locatinguncertaintyprobabilityfoundationsofphysics
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper stages a structured disagreement between a realist and an idealist over Algorithmic Idealism, a formal model in which reality is made of first-person self states — unembedded patterns — and the next state is chosen by universal induction. The idealist side claims this dissolves the Boltzmann brain paradox: no matter how many random brains a cosmological model produces, induction says your next experience is 'business as usual', not disintegration. The realist side raises two objections — that the transition probabilities lack a clear subject, and that a Boltzmann brain's self state could encode its entropic environment, making a violent future predictable. The paper defends the framework against both and concludes that Boltzmann brain counting cannot constrain cosmological models, while conceding that the probabilities' interpretation and the conscious/unconscious split need more work. If the argument holds, cosmology loses a common theoretical argument and the realism/idealism dispute gains a concrete, if private, predictive arena.

What carries the argument

The engine of the argument is Algorithmic Idealism's state-transition postulate: if your current self state is x, the objective private chance of next being xy is the conditional algorithmic probability P(y|x) defined by Solomonoff induction on a universal monotone Turing machine. Under this measure, compressible continuations of one's data are likely and incompressible ones unlikely. The bit model, in which self states are binary strings and transitions append bits, provides the formal setting; the emergent-world theorem says that agents will long-run behave as if embedded in a simple computable probabilistic world. The BB resolution turns on asserting that typical BB realizations have near

What would settle it

Take a specific computable cosmological model that is BB-dominated and in which BBs are generated by a process that correlates them with their environment; compute whether the conditional algorithmic probability of a 'disintegration' next state is actually small. If even one such model yields P(BB-like future | x) comparable to P(OO-like future | x), the resolution fails. More directly, formalize a uniform measure over BB realizations of a given self state and check whether the subset that is uncorrelated with the environment has measure one.

Watch

Extended reading notes

Core claim

The paper's central claim is that Algorithmic Idealism predicts that what happens to you next is what universal induction would predict from your current self state, not what counting microstates plus a principle of indifference would predict. Applied to the Boltzmann brain problem, the claim is that OO-like future experiences have overwhelmingly higher conditional algorithmic probability than BB-like disintegration experiences, regardless of how many Boltzmann brains exist in the universe. From this the authors conclude that the mere fact that a cosmological model is Boltzmann-brain-dominated cannot be used to rule it out. In response to the objection that a Boltzmann brain's self state mig

Load-bearing premise

The Boltzmann-brain resolution hinges on the unproved assertion that the vast majority of Boltzmann brain realizations have their entire self state — conscious and unconscious parts — algorithmically uncorrelated with their environment.

Editorial extensions

If this is right

  • Cosmologists should not reject a model merely because it predicts vastly more Boltzmann brains than ordinary observers; the counting argument is declared irrelevant.
  • Private-experiment questions — duplication, simulation, split-brain, survival — get well-defined answers in principle: the more compressible the future given your state, the more likely it is.
  • The theory predicts that self states will long-run behave as if embedded in a simple computable probabilistic world, recovering the appearance of an external physical world.
  • First-person and third-person probability assignments can diverge in exotic situations, leading to 'probabilistic zombies' and nonstandard predictions about what you will see happening to others.

Reading between the lines

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

  • One could test the framework's practical content by deriving concrete probability numbers for a simple duplication scenario and comparing them with intuitive or operational credences; the paper only gives qualitative inequalities.
  • The 'vast majority' claim about BB realizations is a concentration-of-measure statement; a formal probability bound over realizations of a self state would either shore up or sink the BB resolution.
  • The adversarial-collaboration format itself looks portable to other foundational disputes (e.g., interpretations of quantum mechanics), though the paper only gestures at this possibility.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper reports an adversarial collaboration between a realist and a defender of Algorithmic Idealism. It presents a condensed version of Algorithmic Idealism, including the bit model with self states as binary strings and transition probabilities given by conditional algorithmic probability, and claims that the framework predicts the emergence of a simple external world. The central application is the Boltzmann brain problem: the paper argues that an agent should prioritize induction over counting, so that an OO-like future is overwhelmingly more probable than a BB-like future regardless of how many Boltzmann brains the physics predicts. A critic raises two objections: (O1) the probabilities P(y|x) are not well-defined because the theory lacks a clear account of what the probabilities are probabilities of; and (O2) a BB's self state may encode correlations with its maximum-entropy environment, which would make a BB-like future predictable and hence not improbable. The idealist responds that most BB realizations are almost uncorrelated with their environments and that the analysis gives valid intuition but not the exact technical arguments, which are deferred to earlier work. The authors conclude that their collaboration was successful and that both objections identify important work for the future.

Significance. If the central claim were established, the paper would be significant: it would offer a principled first-person resolution of the Boltzmann brain paradox and would block the use of BB counting to constrain cosmological models. The adversarial collaboration format is a genuine strength: the two positions are stated carefully, the objections are not strawmen, and the authors explicitly concede the places where the theory needs further development. The distinction between 'deceiving' and 'surprisal' self states, and the admission that O1 and O2 are valid open problems, are honest and useful. However, the paper's headline contribution—the resolution of the BB problem—is not actually derived here. The decisive inequality is deferred to reference [7], and the response to O2 relies on an unformalized 'vast majority' typicality claim. As it stands, the manuscript is a valuable programmatic and clarificatory document, but not a demonstration that Algorithmic Idealism solves the BB problem.

major comments (4)
  1. [Section III.B] The displayed inequality ∑_{y∈Y_OO} P(y|x) ≫ ∑_{y∈Y_BB} P(y|x) is the load-bearing result of the paper, but it is not proved here. The text says only that 'going through the mathematical details provided in [7] leads to the conclusion', and Section V.B explicitly states that the analysis 'gives valid intuition, but not the exact technical arguments' and that the details are 'discussed in more depth in [7]'. A reader of this paper cannot verify the central claim without consulting an external source. Please either state the relevant theorem with its assumptions and a proof sketch, or clearly reframe the paper's conclusion as conditional on [7] rather than as a self-contained resolution of the Boltzmann brain problem.
  2. [Section V.B] The response to O2 hinges on the assertion that 'the vast majority of BB realizations of x will have close to maximal entropy under the constraint of realizing x' and later that 'the vast majority of BBs will have all of x, and hence also u, uncorrelated with the BB's environment.' No measure is defined over the space of BB realizations or over the infinite space of self states, so 'vast majority' has no determinate content. Moreover, the paper concedes that deceiving self states exist and are not a priori unlikely. An agent with introspective access only to the conscious part c cannot tell whether their x is one of the typical states or one of the deceiving states; the private probability P(y|x) is supposed to answer exactly this question from x alone. Appealing to a majority over physical realizations reintroduces the third-person counting that Section III.B declares irrelevant. A form
  3. [Section IV.A and Section V.C (O1)] The paper concedes that 'the interpretation of the probabilities P(y|x) has to be further elaborated' and that, unlike quantum theory, there is no intersubjective repeated-experiment grounding for these probabilities. This is not a peripheral worry: the BB resolution is a quantitative comparison of conditional probabilities. The analogy to quantum mechanics is suggestive but incomplete, because quantum probabilities come with an extensive operational apparatus that is absent here. Until the notion of probability used in the BB inequality is specified more precisely, the central conclusion cannot be fully evaluated.
  4. [Section III.B and Section V.B] There is a circularity concern that the text does not dispel. The sets Y_OO and Y_BB are characterized phenomenologically ('business as usual' versus 'weird... disintegrating'), and the paper then asserts that OO-like continuations are compressible while BB-like continuations are not. Algorithmic probability is defined to favor compressible extensions, so the inequality is to a significant degree built into the classification. Section V.B's 'information-theoretic definition' for distinguishing ordinary-planet-like from BB-like realizations itself uses algorithmic correlation, which is precisely what is at issue. To avoid the objection that the conclusion is predetermined, the authors should provide independent, theory-neutral characterizations of OO-like and BB-like futures and then prove the compressibility ordering, rather than defining the categories in terms of compressibility.
minor comments (6)
  1. [Section IV heading] Typo: 'Bolzmann' should be 'Boltzmann'.
  2. [Section II.B, Definition 2] P_U(b|x) is defined for a single bit b, but Section III.B applies it to y representing 'the next, say, hundred bits'. Please clarify how conditional algorithmic probability is extended to multi-bit continuations.
  3. [Section II.B] Typo: 'desribed' should be 'described'. Also, the claim that predictions are invariant under the choice of universal machine U is stated informally; later passages correctly note that the invariance is asymptotic up to multiplicative constants, and this should be made explicit at the first occurrence.
  4. [Section V.C] The objections are labeled O1 and O2 for the first time in Section V.C, but they are not labeled in Section IV. Adding the labels at the point where the objections are stated would make the cross-references easier to follow.
  5. [Section II.C, Eq. (1)] The statement 'With P-probability of at least 2^{-K(µ_W)}' is not fully formal as written; please specify the underlying measure and the quantifier order over n and the probabilistic event.
  6. [References] The paper relies heavily on the companion manuscript [1], which is described as unpublished. Since the success condition of the collaboration is defined there, a brief self-contained summary of that condition would help readers evaluate the authors' claim that the collaboration was successful.

Circularity Check

3 steps flagged · score 6.0 of 10

BB resolution is substantially encoded in the definition of algorithmic probability, an unproven 'vast majority' typicality premise, and a load-bearing self-citation; the paper is transparent about the gaps, but the central conclusion is not independently derived in this text.

  1. self definitional [Section III.B, using the definition of P(y|x) from Definition 2 and Section II.B]
    "P(y|x) is larger if and only if xy is a more compressible extension of x, i.e. one that is a more natural guess for induction. ... Going through the mathematical details provided in [7] leads to the conclusion that, indeed, sum_{y in Y_OO} P(y|x) >> sum_{y in Y_BB} P(y|x)."

    Y_OO is introduced as the set of 'business as usual' continuations, while Y_BB is described as 'weird and different' high-temperature radiation. Algorithmic probability has just been defined as favoring more compressible extensions. Thus the advertised result that business-as-usual futures dominate BB-like futures is obtained by labelling the compressible continuations as OO-like and the incompressible ones as BB-like, and then reading off the definition of the probability measure. The conclusion is therefore largely an unpacking of the chosen definitions rather than an independently derived first-principles prediction.

  2. self citation load bearing [Section V.B; also III.B]
    "Note that the analysis here gives valid intuition, but not the exact technical arguments: it is not literally the algorithmic mutual information between x and y that is relevant here, but the question of whether there exists a computable probability measure of short program length for which both x and xy are typical outcomes. The technical details are discussed in more depth in [7]."

    The decisive inequality of the Boltzmann-brain resolution is the load-bearing result of the paper. In III.B it is asserted by saying 'Going through the mathematical details provided in [7]', and in V.B the paper concedes that the present argument is only intuition and that the exact technical argument lives in [7]. Reference [7] is the advocate's own earlier work. The central claim is thus supported in this text mainly by a self-citation whose content is not reproduced or verified here.

1 more flagged steps
  1. other [Section V.B, response to objection O2]
    "However, the analysis above has shown that this premise is not satisfied: the vast majority of BBs will have all of x, and hence also u, uncorrelated with the BB's environment."

    This sentence is the response to the critic's counterexample of a deceiving self state whose unconscious part u is correlated with the BB's environment. The 'vast majority' claim is load-bearing: it is what makes P(BB-like future | x) small. But no measure over BB realizations or over the infinite space of self states is introduced, and the paper itself earlier described the analogous move as a restriction 'defined just so that it solves the BB problem'. The conclusion is thus supported by an unformalized typicality assertion chosen to make the desired inequality true, rather than by a derivation from the postulates.

full rationale

The paper is unusually candid: it labels its own analysis as giving 'valid intuition, but not the exact technical arguments', identifies two open points O1 and O2, and frames the collaboration as leaving the verdict to the reader. That transparency is not itself circularity. The non-circular core is the substantive proposal that if one accepts Postulate 2, one obtains a different answer to the Boltzmann-brain puzzle than from third-person counting. However, the specific quantitative claim that OO-like futures dominate BB-like futures is not self-contained. It is obtained by combining a probability measure that by construction favors compressible continuations with a classification of 'business as usual' futures as compressible and 'weird high-temperature' futures as incompressible, and the exact proof is deferred to the authors' own prior work. Moreover, the response to the strongest objection—correlated unconscious content in a BB—relies on an unmeasured 'vast majority of BBs are uncorrelated with their environment' premise, which is effectively an input chosen to preserve the conclusion. These features make the central resolution substantially predetermined by definition and self-citation, so a score of 6 is appropriate. I do not assign a higher score because the paper explicitly acknowledges the gaps, does not present a uniqueness theorem as external authority, and the underlying postulates are stated independently of the BB problem.

Assumptions & free parameters 2 free parameters · 5 assumptions · 1 invented entities

The central argument rests on idealist postulates, an arbitrary choice of universal machine, and an unproven entropic assumption about Boltzmann brains. These are not fitted numbers, but they are also not independently established outside the framework.

free parameters (2)
  • Universal monotone Turing machine U
    Transition probabilities P_U(b|x) are defined relative to an arbitrary universal monotone Turing machine. The paper claims U-invariance only asymptotically and up to multiplicative constants, so finite-string predictions such as the BB inequality depend on this choice.
  • Encoding map φ between self states and binary strings
    The dictionary between self states and binary strings is chosen by hand; the paper argues predictions are invariant under structure-preserving maps, but the specific sets Y_OO and Y_BB are defined in a particular encoding.
assumptions (5)
  • domain assumption There exists a countably-infinite set S of self states, and everything about a first person at a moment is determined by its self state.
    Postulate 1 in Section II.A; the core idealist claim, not derived from anything else.
  • domain assumption The next self state is determined by a universal method of induction, formalized as algorithmic probability.
    Postulate 2 in Section II.A; this is the mechanism that is supposed to ground physics and resolve observer paradoxes.
  • ad hoc to paper In the bit model, self states are binary strings, every transition appends exactly one bit, and agents fundamentally never forget anything.
    Section II.B acknowledges this is a simplifying assumption that is "manifestly only approximately correct" in some scenarios.
  • standard math Solomonoff induction converges to computable measures and universal machines are multiplicatively equivalent.
    Background algorithmic information theory from [13,14], used to justify U-invariance and the emergence theorem in Eq. (1).
  • ad hoc to paper The vast majority of Boltzmann brain realizations have their self states almost uncorrelated with their environment.
    Section V.B; this is load-bearing for the response to objection O2, but no measure or proof is given.
invented entities (1)
  • Self states (observer states)
    purpose: Primitive entities of Algorithmic Idealism; abstract patterns that replace an external physical world as the fundamental substrate.
    No independent falsifiable handle is provided. The paper's evidence is internal consistency and claimed private predictions, which cannot be intersubjectively tested.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Realism about the external world: an adversarial collaboration." pith.science (2026). https://pith.science/paper/S65GYKFS

@misc{pith2026260716379,
  author       = {Pith},
  title        = {Pith review of: Realism about the external world: an adversarial collaboration},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/S65GYKFS}},
  note         = {Machine review of arXiv:2607.16379}
}
read the original abstract

One of us is a realist and believes that reality fundamentally consists in an external physical world, governed by laws of physics. Observers are either emergent (physicalism) or exist in addition to physical reality. The other defends a version of idealism and believes that reality fundamentally consists in first-person states, unembedded into worlds, on which laws of nature act directly. Shared "physical worlds" are merely emergent. The disagreement ultimately centers on the overall coherence of a formal model, known as Algorithmic Idealism, and its ability to resolve observer paradoxes, such as the Boltzmann brain paradox. Our aim is to confront this issue in the form of an adversarial collaboration, where we avoid misunderstanding each other, so that readers can see the true source of the disagreement, and decide for themselves. Our work represents an attempt to establish a form of exchange that may help overcome the fragmentation of scholarly communities in philosophy, physics, and elsewhere. The collaboration was conducted using a discipline-neutral template for theoretical adversarial collaboration developed in a companion paper [1], which we hope will be useful in other theoretical disciplines.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

21 extracted references · 1 linked inside Pith

  1. [7]

    M. P. M ¨uller, Law without law: from observer states to physics via algorithmic information theory, Quantum4, 301 (2020)

    2020

  2. [1]

    K. J. McQueen and M. P. M ¨uller, Theoretical adversarial collaboration: A template (2026), manuscript

    2026

  3. [2]

    Mellers, R

    B. Mellers, R. Hertwig, and D. Kahneman, Do frequency representations eliminate conjunction effects? an exercise in adver- sarial collaboration, Psychological Science12, 269 (2001)

    2001

  4. [3]

    A. W. Corcoran, J. Hohwy, and K. J. Friston, Accelerating scientific progress through bayesian adversarial collaboration, Neuron111, 3505 (2023)

    2023

  5. [4]

    Chalmers, Idealism and the mind-body problem, inThe Routledge Handbook of Panpsychism, edited by W

    D. Chalmers, Idealism and the mind-body problem, inThe Routledge Handbook of Panpsychism, edited by W. Seager (Rout- ledge, 2019) pp. 353–373

    2019

  6. [5]

    Berkeley,A treatise concerning the principles of human knowledge(JB Lippincott & Company, 1881)

    G. Berkeley,A treatise concerning the principles of human knowledge(JB Lippincott & Company, 1881)

  7. [6]

    Hylton,Russell, Idealism, and the Emergence of Analytic Philosophy(Oxford University Press, 1992)

    P. Hylton,Russell, Idealism, and the Emergence of Analytic Philosophy(Oxford University Press, 1992)

    1992

  8. [8]

    C. L. Jones and M. P. M ¨uller, On the significance of Wigner’s Friend in contexts beyond quantum foundations, Quantum10, 2147 (2026)

    2026

Show all 21 references
  1. [9]

    Chalmers,The Conscious Mind(Oxford University Press, 1996)

    D. Chalmers,The Conscious Mind(Oxford University Press, 1996)

    1996

  2. [10]

    P M ¨uller, Algorithmic Idealism: What Should You Believe to Experience Next?, Foundations of Physics56, 11 (2026)

    M. P M ¨uller, Algorithmic Idealism: What Should You Believe to Experience Next?, Foundations of Physics56, 11 (2026)

    2026

  3. [11]

    Sider, All the world’s a stage, Australasian Journal of Philosophy74, 433 (1994)

    T. Sider, All the world’s a stage, Australasian Journal of Philosophy74, 433 (1994)

    1994

  4. [12]

    D. C. Dennett, Real patterns, The journal of Philosophy88, 27 (1991)

    1991

  5. [13]

    Li and P

    M. Li and P. Vit´anyi,An Introduction to Kolmogorov Complexity and Its Applications (Third Edition)(Springer, 2008)

    2008

  6. [14]

    Hutter,Universal Artificial Intelligence: Sequential Decisions Based on Algorithmic Probability(Springer, 2005)

    M. Hutter,Universal Artificial Intelligence: Sequential Decisions Based on Algorithmic Probability(Springer, 2005)

    2005

  7. [15]

    S. M. Carroll, Why boltzmann brains are bad, inCurrent Controversies in Philosophy of Science(Routledge, 2020) p. 14

    2020

  8. [16]

    Adlam, Against self-location, arXiv (2024), 2409.05259 [physics.hist-ph]

    E. Adlam, Against self-location, arXiv (2024), 2409.05259 [physics.hist-ph]

    2024 arXiv

  9. [17]

    Elga, Defeating dr

    A. Elga, Defeating dr. evil with self-locating belief, Philosophy and Phenomenological Research69, 383 (2004)

    2004

  10. [18]

    Parfit, Divided minds and the nature of persons, inScience Fiction and Philosophy: From Time Travel to Superintelligence, edited by S

    D. Parfit, Divided minds and the nature of persons, inScience Fiction and Philosophy: From Time Travel to Superintelligence, edited by S. Schneider (Wiley-Blackwell, 2009) pp. 91–98

    2009

  11. [19]

    D. Z. Albert,Quantum Mechanics and Experience(Cambridge University Press, 1992)

    1992

  12. [20]

    Sider, Personal identity, inRiddles of Existence: A Guided Tour of Metaphysics (Second Edition), edited by E

    T. Sider, Personal identity, inRiddles of Existence: A Guided Tour of Metaphysics (Second Edition), edited by E. Conee and T. Sider (Oxford University Press, Oxford, 2014) pp. 7–21

    2014

  13. [21]

    Wolfram,A New Kind of Science(Wolfram Media Inc., 2002)

    S. Wolfram,A New Kind of Science(Wolfram Media Inc., 2002)

    2002

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.