Quantum Decision Theory

Herve Zwirn

REVIEW 4 major objections 5 minor 8 references

Quantum Decision Theory

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read The paper argues that preferences can be genuinely indeterminate before a decision, and that choosing acts like a quantum measurement that collapses the state onto one preference.

desk verdict The alleged Section 3.2 inconsistency is a misread; the real problem is that the model is untested and can fit anything. read the letter →

arxiv 2412.02165 v1 pith:PQHUC3QL submitted 2024-11-22 physics.soc-ph quant-ph

Herve Zwirn This is my paper · ORCID

classification physics.soc-phquant-ph MSC 91B0681P16

keywords quantumdecisiontheorypreferenceindeterminacyframingeffectprisoner'sdilemmameasurementcollapsenon-commutingobservablesinterferencebehaviouralanomalies

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

An agent's preferences, the paper argues, can be genuinely indeterminate (in the quantum sense) before a choice, and the choice itself acts as a measurement that projects the agent's state onto one preference. The motivation is to explain framing effects and other behavioural anomalies without assuming that agents have fixed preferences that a question or a presentation secretly changes. The central mathematical claim is that commuting decision observables reproduce classical Bayesian predictions, while non-commuting observables generate interference terms that alter choice probabilities. The paper applies this to the decomposed prisoner's dilemma, where equivalent presentations of the same game should yield different cooperation rates, and it proposes a two-population experiment patterned on the double-slit experiment to test the effect. If correct, the model implies that eliciting a preference, or even presenting a problem, is not an inert operation on the agent.

What carries the argument

The load-bearing object is the agent's state vector $|\psi\rangle$ in a Hilbert space whose dimension is at least the number of alternatives, together with the association of each decision situation with an observable $A$ whose eigenvalues label the choices. The rule that does the work is the projection postulate: measuring $A$ yields eigenvalue $i$ with probability $|\langle i|\psi\rangle|^2$ and replaces $|\psi\rangle$ by the eigenstate $|i\rangle$. For two observables $A$ and $B$ that do not commute, writing the state in the $B$-basis and then in the $A$-basis produces an amplitude $\sum_j \nu_j\mu_{ij}$ whose squared modulus contains cross terms; after a prior $B$-measurement the same quantity is $\sum_j|\nu_j\mu_{ij}|^2$, so the difference between the two is exactly the interference that drives the paper's predictions.

What would settle it

Conduct the two-population prisoner's dilemma experiment described in the paper: one group answers a yes/no question about a trait such as altruism before playing, and the other group plays immediately; if the cooperation rates are equal, the predicted interference is absent and the central mechanism fails.

Watch

Extended reading notes

Core claim

At the paper's centre is the identification of a decision with a quantum measurement. The agent's state $|\psi\rangle$ is a superposition of basis states corresponding to the alternatives, and the probability of choosing alternative $i$ is the squared amplitude $|\lambda_i|^2$; immediately after the choice, the state collapses to the eigenstate $|i\rangle$. For two decision situations, the paper distinguishes commuting observables, where a joint probability distribution over pairs of choices exists and the classical conditional-probability formula holds, from non-commuting observables, where the probability of a choice contains interference cross terms of the form $\nu_j^*\mu_{ij}^*\nu_k\mu_{ik}$ that vanish if a prior measurement has been made. This gives a formal mechanism for order effects and context effects. The paper's explanation of framing is that presentation acts as another measurement, projecting the agent's state onto one of several mental-representation eigenstates; because two presentations project onto different states, the subsequent choice probabilities $p_{GA}(C)$ and $p_{GB}(C)$ differ even though the games are payoff-equivalent. It then predicts that a preliminary question about altruism can change prisoner's dilemma cooperation rates only when the corresponding observables do not commute.

Load-bearing premise

The paper treats a mere presentation or question as something that collapses a person's mental state while also writing the resulting choice probabilities as if no collapse had occurred.

Editorial extensions

If this is right

Commuting decision situations give back classical Bayesian probabilities, so the quantum model is a strict generalisation rather than a replacement.
If a decision observable does not commute with a preliminary question, answering that question changes the probabilities of later choices, so the model predicts non-classical order effects.
The decomposed prisoner's dilemma should show more cooperation when the game is presented in the giving form than in the standard payoff-table form, because the two presentations project agents onto different mental-representation states.
In the proposed experiment, comparing cooperation rates between a population that first answers an altruism question and one that does not can reveal whether the two observables commute.
Once a choice is made, repeating the same decision situation immediately yields the same choice with certainty, matching the stability of ordinary revealed preferences.

Reading between the lines

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

A consequence the author leaves implicit is that preference elicitation is never neutral: any questionnaire that measures a trait changes the probabilities of subsequent choices, which would be a design constraint on behavioural experiments.
The formalism suggests an empirical criterion for non-commutation: if reversing the order of two decision questions changes the joint distribution of choices, the corresponding observables do not commute.
Because the interference amplitude is controlled by the basis-change matrix, the size of order effects is bounded, and measuring that bound could distinguish a genuinely quantum-like model from a classical mixture model.
A natural extension, not developed in the paper, would reinterpret preference reversals over time as sequential measurements of non-commuting decision observables rather than as changes in underlying utility.

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.

Desk Editor's Note

read the letter

Quick take: the alleged Born-rule error in Section 3.2 isn't there. The paper subtracts the interference cross-term after projecting onto the framing observable, which is exactly what the projection postulate requires. The stress-test note has it right. I read the displayed formulas as p_GA(C) = |alpha_1|^2|gamma_11|^2 + |alpha_2|^2|gamma_21|^2, which is p_G(C) minus the interference term—consistent with collapse, not contradictory.

What's actually new: a double-slit-style two-population prisoner's dilemma experiment that would test type indeterminacy, and a framing model built on the author's earlier type-indeterminacy work. The exposition of the formalism is clean: it recovers classical Bayesian results when decision observables commute, and it gives a coherent account of how framing could shift choice without assuming preferences are simply changed by the question.

Soft spots, in proportion. The central claim is empirically untested. The proposed experiment is schematic—no results, no protocol for controlling the basis-change matrix or the initial amplitudes. Those parameters are free, so any observed P_II(coop) != P_I(coop) difference could be accommodated. That makes the framing explanation post hoc rather than predictive. The paper also does not engage the existing quantum decision literature (Busemeyer, Pothos, Yukalov, etc.), which already contains quantum models of order effects and framing. The novelty is bounded because the formal apparatus comes from reference [4], the author's own prior work. Non-commutation of the relevant observables is asserted, not derived from behavioral principles.

None of that is fatal to the program. The idea that preferences are indeterminate until elicited is a legitimate modeling hypothesis, and the double-slit experiment is a plausible way to distinguish it from a classical 'question changes preference' account—though even that needs a design that fixes the free parameters.

For whom: someone working on quantum cognition or behavioral decision theory might find this a useful teaching example or a starting point for an experiment. It is not a self-contained new framework. I'd send it to review, with the expectation that major revision is needed—an actual experiment, or at least a parameter-free signature—and better positioning against existing QDT models. As is, it's a coherent but untested proposal.

Referee Report

4 major / 5 minor

Summary. The paper proposes a quantum-mechanical formalism for decision theory, in which an agent's preferences are represented as a superposition of potential preferences before a decision, and the decision itself is modeled as a projective measurement of the agent's state. The author shows that when the observables associated with two decision situations commute, the formalism reduces to classical Bayesian probability, whereas non-commuting observables introduce interference terms that can produce order and context effects. The framework is applied to two examples: a proposed two-population prisoner's dilemma experiment (analogous to the double-slit experiment) and the framing effect in a decomposed prisoner's dilemma. The central claim is that preference indeterminacy is essential, not merely epistemic, and that decision is a measurement-like process.

Significance. If the central claim were empirically supported, the formalism would offer a novel and mathematically rigorous way to model context-dependent preferences, order effects, and the robustness of preferences. The paper's consistency check for commuting observables (Section 2.4) is a useful sanity check, and the derivation in Section 3.2 correctly removes the interference term after projective measurement, contrary to the reader's initial concern. However, the paper presents no experimental data, the model's free parameters make its main predictions highly tunable, and the framing application is explicitly based on arbitrary mental-representation coefficients. The manuscript is a preliminary modeling proposal rather than an empirically grounded theory.

major comments (4)

[Section 3.1] The proposed experiment is not carried out and no data are reported. The paper explicitly calls it a 'fictitious experiment' and the conclusion states that 'a proper experimental protocol' is still needed. Without actual data or at least a fully specified protocol with concrete predicted effect sizes, the central claim that preferences are indeterminate and decision is a measurement remains untested. This is load-bearing because the abstract and conclusion present this claim as the article's main result.
[Sections 3.1 and 2.4] The model's free parameters (initial amplitudes λ1, λ2 and basis-change coefficients μij) are unconstrained. The inequality P_II(coop) ≠ P_I(coop) can be produced or suppressed by choosing the μij appropriately, and any observed difference can be fitted with suitable coefficients. The paper does not derive a parameter-free quantum signature—for example, an inequality among choice probabilities that classical models cannot satisfy—and it does not rule out the classical explanation that answering a question changes preferences. Thus the proposed experiment cannot discriminate the quantum model from a classical model with altered preference distributions.
[Section 3.2] The explanation of the framing effect relies on freely chosen basis states and coefficients, and the paper itself admits in footnote 7 that the mental representations are 'arbitrary and makes no claim to psychological accuracy.' Consequently, the observed difference in cooperation probabilities between presentations A and B is not explained in a predictive sense; it is merely encoded in the products α*γ and β*δ. The model would offer a genuine explanation only if these coefficients were derived from independent principle or fixed by separate measurements; as written, the framework can accommodate any observed framing difference.
[Section 2.4] The paper asserts, without justification, that the questionnaire observable and the prisoner's dilemma observable (and similarly the presentation observables and the decision observable in Section 3.2) do not commute. All new quantum predictions follow from non-commutation, yet no criterion is provided for deciding when two decision situations should be represented by non-commuting observables. Without such a criterion, the model cannot make predictions for new contexts; non-commutation is chosen post hoc to fit the examples.

minor comments (5)

[Section 2.4] In the degenerate-eigenvalue discussion, the operators A and B are not fully written out; they should be defined explicitly as A = Σ_i a_i |i><i| and B = Σ_i b_i |i><i| before the probabilities p_AB(i|j) are derived.
[Section 3.2] The notation p_GA(C) is ambiguous because it could be read as the probability of C in a collapsed state, whereas it actually denotes the mixture over collapsed outcomes after the projective measurement. Writing p(C | presentation A) would be clearer.
[Section 3.1] The argument that population proportions equal quantum probabilities assumes that all agents are in the same initial state and that the law of large numbers applies; this assumption should be stated explicitly when equating P_I(coop) and P_II(coop) with the corresponding quantum probabilities.
[References] Reference [7] lists the journal as 'Journal of Personality and Psychology'; the correct title is 'Journal of Personality and Social Psychology'. Reference [8] for Selten lacks page numbers; please complete it.
[Throughout] There are several language issues: the French word 'avec' appears in Section 2.1, 'lhe game' in Section 3.2 should be 'the game', and capitalization of 'Her' is inconsistent in Section 2.1. A careful proofread is needed.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the quantum derivations follow from explicit postulates; the main limitation is that the model is untested and underdetermined, plus one non-load-bearing self-citation.

full rationale

The paper's central claim (preferences are indeterminate before a decision) is a postulate, not a derived result, so it cannot be circular in the derivation sense. Section 2 builds the formalism directly on quantum measurement postulates and checks that commuting observables reduce to classical Bayesian probabilities. Section 3.1 proposes an experiment but reports no data; the difference PII(coop)-PI(coop) is the interference term, and because the model parameters are free, the test is weak, but no fitted value is renamed as a prediction. Section 3.2's framing explanation is explicitly illustrative: the paper says the mental-representation descriptions are 'arbitrary and make no claim to psychological accuracy.' The displayed equations are conditional consequences of assuming a presentation-induced projection; the alleged inconsistency in the reader's take is not present because the formula subtracts the interference cross-term, which is precisely the projective-measurement effect. The only self-citation is Ref. [4], which shares an author; it is used to attribute the formalism, but the equations in this paper are re-derived independently, so that citation is not load-bearing. Overall, there is no equation-level circularity; the genuine weaknesses are empirical (the proposed experiment is not run) and underdetermination (free parameters), which are correctness/falsifiability concerns rather than circularity.

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

The model's explanatory engine is the choice of Hilbert-space basis coefficients and the assertion of non-commuting observables. Without data, these are unconstrained degrees of freedom, so the ledger is dominated by modeling assumptions rather than measurements.

free parameters (2)

Basis coefficients of presentation observables (alpha_i, beta_i, gamma_ij, delta_ij) = None specified
In Section 3.2, the mental representations a_i and b_i are introduced with arbitrary basis coefficients; the predicted framing effect depends entirely on these choices, which the paper admits are not tied to psychological measurement.
Initial superposition amplitudes (lambda_i, nu_j) = None specified
The probabilities of all choices derive from these amplitudes, but they are not measured or specified for any real decision situation, so they remain free parameters of the model.

assumptions (5)

domain assumption Preferences can be represented by vectors in a Hilbert space and choice probabilities follow the Born rule (Section 2.3).
No justification is given for why human preference states should obey quantum probability rules; this is the foundational postulate of the paper.
domain assumption Each decision situation corresponds to an observable whose eigenstates are the alternatives (Section 2.3).
This mapping between choices and measurements is stipulated without support from psychological evidence.
domain assumption There exist decision situations whose observables do not commute (Sections 2.5 and 3.1).
The experimental predictions require non-commuting observables, but no evidence or argument shows that, for example, answering a personality questionnaire and playing a prisoner's dilemma do not commute.
domain assumption The two populations in the proposed experiment have identical initial states (Section 3.1).
The interference prediction depends on identical initial superpositions for the two populations, which is plausible but unverified.
ad hoc to paper The framing presentation projects the agent's state onto an eigenstate of a representation observable (Section 3.2).
This is introduced to model framing, but it conflicts with the interference formulas that follow in the same section.

invented entities (2)

Indeterminate preference state (superposition of potential preferences) independent evidence
purpose: To explain why equivalent decision contexts can produce different choices without assuming preferences change.
The proposed two-population experiment (Section 3.1) is a falsifiable handle: it predicts a difference in cooperation rates that would violate classical conditional probability. However, the prediction is conditional on unspecified basis parameters and non-commutation, and no experimental data are provided.
Mental representation eigenstates associated with a framing observable
purpose: To model how the presentation of a decision situation alters the agent's state before the final choice.
The mental representations in Section 3.2 ('the game is seen as pure fun', 'as a test of generosity', etc.) are, by the paper's own footnote 7, arbitrary and not measured independently.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Quantum Decision Theory." pith.science (2026). https://pith.science/paper/PQHUC3QL

@misc{pith2026241202165,
  author       = {Pith},
  title        = {Pith review of: Quantum Decision Theory},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PQHUC3QL}},
  note         = {Machine review of arXiv:2412.02165}
}

read the original abstract

In this article, we propose to use the formalism of quantum mechanics to describe and explain the so-called "abnormal" behaviour of agents in certain decision or choice contexts. The basic idea is to postulate that the preferences of these agents are indeterminate (in the quantum sense of the term) before the choice is made or the decision is taken. An agent's state before the decision is represented by a superposition of potential preferences. The decision is assimilated to a measure of the agent's state and leads to a projection of the state onto one of the particular preferences. We therefore consider that uncertainty about preferences is not linked to incomplete information but to essential indeterminacy. We explore the consequences of these hypotheses on the usual concepts of decision theory and apply the formalism to the problem of the so-called "framing" effect.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

8 extracted references · 8 canonical work pages

[4]

and Zwirn H., Type indeterminacy: A model of the KT(Kahneman_Tversky)-man

Lambert Mogiliansky, A., Zamir S. and Zwirn H., Type indeterminacy: A model of the KT(Kahneman_Tversky)-man. Journal of Mathematical Psychology, (2009)

work page 2009
[1]

and TverskyA., Choice, Values and Frames, Cambridge University Press (2000)

Kahneman D. and TverskyA., Choice, Values and Frames, Cambridge University Press (2000)

work page 2000
[2]

Harsanyi JC., Games of Incomplete Information Played by Bayesian Players, Part I, II, II Management Sciences (1967). 14

work page 1967
[3]

and Simonson I., Context-Dependent Preferences, Management Sciences 39 : 85-117 (1993)

Tversky A. and Simonson I., Context-Dependent Preferences, Management Sciences 39 : 85-117 (1993)

work page 1993
[5]

Mackey G.W., Mathematical Foundations of Quantum Mechanics, New York: Benjamin (1963)

work page 1963
[6]

The character of Physical Laws, The M.I.T

Feynman R., 1. The character of Physical Laws, The M.I.T. Press. (1965)

work page 1965
[7]

, Reward Structur e of Cooperation: the Decomposed Prisoner's Dilemma Game., Journal of Personality and Psychology 7: 21-27 (1970)

Pruitt D.G. , Reward Structur e of Cooperation: the Decomposed Prisoner's Dilemma Game., Journal of Personality and Psychology 7: 21-27 (1970)

work page 1970
[8]

European Economic Review: 413-436 (1998)

Selten R., Features of Experimentally Observed Bounded Rationality. European Economic Review: 413-436 (1998)

work page 1998

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [4]

and Zwirn H., Type indeterminacy: A model of the KT(Kahneman_Tversky)-man

Lambert Mogiliansky, A., Zamir S. and Zwirn H., Type indeterminacy: A model of the KT(Kahneman_Tversky)-man. Journal of Mathematical Psychology, (2009)

work page 2009

[2] [1]

and TverskyA., Choice, Values and Frames, Cambridge University Press (2000)

Kahneman D. and TverskyA., Choice, Values and Frames, Cambridge University Press (2000)

work page 2000

[3] [2]

Harsanyi JC., Games of Incomplete Information Played by Bayesian Players, Part I, II, II Management Sciences (1967). 14

work page 1967

[4] [3]

and Simonson I., Context-Dependent Preferences, Management Sciences 39 : 85-117 (1993)

Tversky A. and Simonson I., Context-Dependent Preferences, Management Sciences 39 : 85-117 (1993)

work page 1993

[5] [5]

Mackey G.W., Mathematical Foundations of Quantum Mechanics, New York: Benjamin (1963)

work page 1963

[6] [6]

The character of Physical Laws, The M.I.T

Feynman R., 1. The character of Physical Laws, The M.I.T. Press. (1965)

work page 1965

[7] [7]

, Reward Structur e of Cooperation: the Decomposed Prisoner's Dilemma Game., Journal of Personality and Psychology 7: 21-27 (1970)

Pruitt D.G. , Reward Structur e of Cooperation: the Decomposed Prisoner's Dilemma Game., Journal of Personality and Psychology 7: 21-27 (1970)

work page 1970

[8] [8]

European Economic Review: 413-436 (1998)

Selten R., Features of Experimentally Observed Bounded Rationality. European Economic Review: 413-436 (1998)

work page 1998