Pith. sign in

REVIEW 3 major objections 5 minor 175 references

The unintended consequences of large language models as a labor-augmenting technology in science

T0 review · 3 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read Even flawless LLMs would reshape how scientists pick and polish projects.

desk verdict A clean optimal-foraging model with a genuinely useful phase distinction; the math mostly holds but the Discussion overreaches and the behavioral premise is unsecured. read the letter →

arxiv 2607.17397 v1 pith:QDMVOGS6 submitted 2026-07-19 physics.soc-ph

Eamon Duede , Kevin Gross , M.J. Crockett , Carl Bergstrom This is my paper
classification physics.soc-ph
keywords largelanguagemodelslabor-augmentingtechnologyoptimalforagingtheorymarginalvaluetheoremscientificpublishingopportunitycostresearcheffortallocationselectivity
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 tries to show that large language models, even if they work perfectly, will not simply give scientists more time to think deeply; they will change which projects get pursued and how carefully each is finished. Using a model of researchers as rate-maximizers, it argues that the effect depends on where in the research pipeline LLMs save time. If LLMs speed up the search for promising projects, scientists will publish a smaller share of projects and polish each one less. If LLMs speed up the mechanics of turning data into a paper, scientists will publish a larger share of projects and still polish less. The unifying cause is that faster work raises the opportunity cost of time, making thoroughness more expensive relative to moving on.

What carries the argument

The engine of the argument is the marginal-value theorem from optimal foraging theory, repurposed as a model of scientific labor. The researcher is modeled as choosing a policy that maximizes the long-run average benefit per unit time, Λ(τ)=E[V u(τ(V))]/(s+E[τ(V)]), where s is discovery time, τ(V) is time spent developing a project of value V, and u is a saturating function. This yields a unique optimal policy: a threshold project value v0 above which projects are developed, and the rule that marginal benefit of extra development equals λ, the opportunity cost of the researcher's time. All comparative statics about LLM speedups flow through changes in s, a, or the shape of u.

What would settle it

Track a field where LLMs are used mainly to generate hypotheses and identify projects. If, after adoption, the fraction of started projects that are published rises rather than falls, or the time spent polishing each accepted paper rises rather than falls, the model's predictions for shortened discovery are contradicted. The same observation serves across scenarios: measure the sign of the change in publication selectivity and in revision effort after a known speedup in a specific phase.

Watch

Extended reading notes

Core claim

The paper's central claim is that a researcher's response to LLM-driven speedups is governed by a single quantity: the shadow value of their time. A scientist should develop a project only if its expected value exceeds a threshold, and should invest in polishing until the marginal benefit of effort equals the long-run rate of return. Shortening the discovery phase raises the threshold (more selectivity) and lowers polishing per paper. Shortening the minimum time needed to produce a publishable manuscript lowers the threshold (less selectivity) and also lowers polishing. Accelerating the phase where extra effort improves the paper raises polishing but has an ambiguous effect on selectivity un

Load-bearing premise

The load-bearing premise is that scientists maximize the long-run average rate of benefit per unit time, with time as the only constraint; if researchers instead maximize total output, prestige, or per-paper quality, the predicted selectivity and thoroughness effects need not hold.

Editorial extensions

If this is right

  • In fields where LLMs help researchers find promising projects faster, the model predicts a smaller fraction of started projects will be published, and published papers will be developed less thoroughly.
  • In fields where LLMs expedite writing, figure-making, and submission, the model predicts more projects will reach publication, but each will receive less discretionary refinement.
  • When LLMs accelerate the discretionary phase (added analyses, robustness checks, polish), published work becomes more thorough, while the effect on how many projects are published depends on the shape of the benefit function.
  • The common prediction across scenarios is that time saved is reallocated to new work rather than to deepening existing work, so the hope that LLMs will buy scientists more time to think deeply is not supported.
  • Because faster manuscript production strains peer review and submissions, the model points to growing pressure on review capacity even under optimistic assumptions about LLM quality.

Reading between the lines

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

  • The model's logic extends beyond LLMs: any technology that shortens discovery or production time should trigger the same reallocation, making the paper a general framework for forecasting how automation reshapes scientific effort.
  • A testable extension is to compare two communities with similar LLM adoption but different primary uses: discovery-use should raise acceptance thresholds and shorten revision times, while write-up-use should raise submission volumes and also shorten revision times.
  • The paper treats institutional incentives as fixed; an extension would be to ask whether changes in credit structures (e.g., rewarding replication or deep revision) could offset the threshold and thoroughness effects, which the model suggests would require altering the opportunity cost of time itself.
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

3 major / 5 minor

Summary. The paper develops a formal model, inspired by optimal foraging theory, of how scientists allocate effort across projects when LLMs reduce time costs in different phases of the research cycle. A researcher first spends a fixed discovery time s, then observes project value v and chooses whether to develop it. Development requires a minimum time a and an optional discretionary time b, producing benefit v u(a+b), where u is increasing and concave. The researcher maximizes the long-run average benefit rate, equivalent to the renewal-reward ratio Λ = E[Vu(τ)] / (s + E[τ]) (Appendix §5.2, Eq. 1). The optimal policy has a threshold v0 and a development-time rule β(v). The paper derives comparative statics for three interventions: reducing s raises v0 and lowers thoroughness for developed projects (Props. 5–7); reducing a lowers v0 and lowers thoroughness (Props. 8–10); accelerating discretionary effort α raises thoroughness and, under a decreasing-elasticity condition, lowers v0 (Props. 11–12). The abstract and discussion frame these results as predictions about how LLMs will affect scientific selectivity and thoroughness.

Significance. If the model's conclusions hold, the paper makes a valuable and non-obvious point: even a perfectly functioning, costless LLM does not simply create more free time; it changes the shadow value of researcher time and thereby reallocates effort across the research cycle. The sign of the effect on selectivity depends on which phase is accelerated, and the thoroughness result is exactly opposite for discretionary-acceleration versus discovery/writing-speed improvements. The paper is commendably transparent: the model has no fitted parameters, the propositions are stated precisely, and the proofs are self-contained. The central theoretical mechanism—the envelope-theorem comparative statics around the shadow value λ—is clean and correct in its main lines. The main weaknesses are that the headline claims in the abstract and discussion go beyond what the propositions establish, and the behavioral objective, while explicit, is not empirically grounded or tested against plausible alternatives. These are fixable in revision and do not undermine the internal mathematical logic.

major comments (3)
  1. [Abstract and Discussion, ¶4] The thoroughness result is overgeneralized. The abstract states that 'by allowing scientists to work more quickly, LLMs raise the opportunity cost of researcher time, creating incentives to refine papers less thoroughly before moving on.' This is true for decreases in s (Prop. 7) and decreases in a (Prop. 10), but Proposition 12 shows that accelerating discretionary development increases thoroughness, z(αβ(v)). The Discussion's claim that 'the common force uniting these results is that ... researchers become less willing to refine and further develop already-publishable results' ignores this opposite case. Since the abstract presents the less-thorough result as a general consequence, this is a load-bearing qualification that should be corrected.
  2. [Appendix §5.2, Eq. (1); Props. 6, 9, 10, 12] All comparative statics depend on the assumption that scientists maximize the long-run average benefit rate Λ(τ). This objective is stated in §2 and formalized in Eq. (1), but the paper gives no empirical or theoretical justification for it, and it does not discuss how the predictions would change under plausible alternative objectives such as maximizing total or discounted career output, maximizing publication count, or maximizing per-paper quality. Under a publication-count objective, for example, the threshold v0 would not respond to s in the same way. Because the abstract makes unconditional empirical predictions about researcher behavior, the paper should either defend the rate-maximization assumption as descriptively appropriate or explicitly frame the results as conditional on it, ideally with a robustness discussion.
  3. [Appendix §5.3.3, Prop. 11] The conclusion that accelerating discretionary development lowers v0 requires the assumption that the elasticity ε_z(b) = b z'(b)/z(b) is decreasing in b. This condition is stated in the proposition but is not discussed in the main text, and its economic plausibility is never assessed. The main-text claim in §3 that the net effect on v0 is 'ambiguous without further structure' is accurate, but readers may miss that the subsequent unambiguous statement in the discussion depends on this unexamined structural assumption. The paper should either justify this condition from microfoundations or present the threshold effect as conditional on a non-primitive assumption.
minor comments (5)
  1. [Appendix §5.3.3, Prop. 12 proof] There is a sign slip: the text says 'z′(αβ(v)) increases with α iff λ/α decreases with α.' The correct statement is that z′(αβ(v)) decreases with α when λ/α decreases. The conclusion still follows, because a decrease in z′ combined with concavity of z implies αβ(v) increases, but the sentence as written is incorrect.
  2. [§3, 'Reduction in minimum development time'] Typo: 'opportunity of cost of time' should be 'opportunity cost of time.' Also, the phrase 'Here the net effect on the threshold v0 is ambiguous' is correct, but the immediately following discussion should clarify that the unambiguous decrease in v0 in the appendix requires the elasticity condition.
  3. [§4, Discussion] Typo: 'nuanced affects' should be 'nuanced effects.' Also, the discussion of journal submissions and peer review would benefit from a citation to the model's distinction between phases, since the submission-pressure argument relates specifically to reductions in a rather than to all LLM uses.
  4. [Appendix §5.3.2, Prop. 8 proof] In the sentence 'here we explicitly derived λ/da', the notation should be dλ/da. Also, in the proof of Prop. 5, the symbol b1 is used where β1 (the policy) is meant; this is confusing because b is also discretionary time.
  5. [Figure 1] The panel references in the text are inconsistent: the text writes '(Figure 1)E)' and the caption lists [A]–[G], but the in-text citations use [E], [F], [G]. The figure caption also contains 'Benefit v(v)' which should likely be 'Benefit v u(t)'. Please clean up the labeling.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the model and comparative statics are derived self-contained from stated assumptions, with no fitted input, definitional equivalence, or load-bearing self-citation.

full rationale

The paper's derivation chain is self-contained. The objective is defined explicitly as the long-run average benefit rate Lambda(tau) = E[V u(tau(V))] / (s + E[tau(V)]) (Appendix Eq. 1), and Propositions 1-12 are proven from this objective plus the stated properties of u(t), F, s, a, and the acceleration parameter alpha. The headline predictions about selectivity (v0 increasing when s falls; v0 falling when a falls) and thoroughness (z(beta(v)) decreasing when s or a falls; increasing under acceleration) follow by envelope-theorem comparative statics from these assumptions rather than being assumed as outputs. No parameter is fitted to data and then renamed as a prediction. The self-citations in the introduction and discussion (e.g., literature on LLM capabilities and peer-review feedback cycles) are contextual or motivational; none is invoked as a uniqueness theorem, ansatz justification, or mathematical premise. The citation of Charnov's marginal value theorem supplies an external conceptual frame, but the appendix re-derives the optimal policy directly, so it is not load-bearing in the proof chain. The skeptic's concern that real researchers may not maximize the specified rate function is an external-validity or robustness issue, not a circularity: the paper states its objective explicitly and derives consequences from it. Thus the honest finding is no significant circularity, score 0.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The model is a stylized optimal-foraging model with no fitted parameters, invented entities, or empirical calibration. All results depend on the listed structural assumptions, the most consequential being the rate-maximization objective and the three-way decomposition of development time.

assumptions (5)
  • domain assumption Project value v is drawn iid from an atomless distribution F with support [0, v̄]; value is revealed only after a fixed discovery phase of length s.
    Appendix §5.1; all threshold results depend on this information structure. It excludes learning, adaptive search, changing project values, or non-iid project portfolios.
  • domain assumption Benefit function u(t)=0 for t≤a, u'>0, u''<0 for t>a, u(t)→1; development time splits into required a and discretionary b.
    Appendix §5.1; concavity and the a-threshold drive the first-order conditions and uniqueness.
  • domain assumption The researcher maximizes long-run average rate of benefit, Λ(τ)=E[V u(τ(V))]/(s+E[τ(V)]), via the renewal-reward theorem.
    Appendix §5.2, Eq. (1); this is the optimal-foraging objective and is the load-bearing assumption that converts all speed-ups into opportunity-cost effects.
  • ad hoc to paper For Proposition 11, the elasticity ε_z(b)=b z'(b)/z(b) is decreasing in b.
    Appendix §5.3.3; needed to sign the threshold effect of α. The paper notes the effect is ambiguous without this structure but gives no independent justification that z has decreasing elasticity.
  • domain assumption LLMs are modeled as time-saving only, with no errors, no change in output quality per unit effort, and negligible financial cost.
    Introduction, paragraph 2; the authors state this is 'as imagined by their most enthusiastic proponents.' The conclusions are conditional on this favorable assumption.

how reviews work

0 comments
Cite this review

Pith. "Pith review of The unintended consequences of large language models as a labor-augmenting technology in science." pith.science (2026). https://pith.science/paper/QDMVOGS6

@misc{pith2026260717397,
  author       = {Pith},
  title        = {Pith review of: The unintended consequences of large language models as a labor-augmenting technology in science},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QDMVOGS6}},
  note         = {Machine review of arXiv:2607.17397}
}
read the original abstract

As a labor-augmenting technology, large language models (LLMs) have the potential to accelerate scientific activity across the research pipeline. But even if LLMs perform on par with human experts at selected tasks, their use will bring unintended consequences as they alter the balance of frictions and inducements that steer the allocation of research effort across projects. Here we develop a simple mathematical model to illustrate. In fields where LLMs are useful primarily as tools for discovering promising projects, researchers will become more selective about what they publish; where they facilitate the process of publishing existing data, researchers will become less selective. By allowing scientists to work more quickly, LLMs raise the opportunity cost of researcher time, creating incentives to refine papers less thoroughly before moving on. Enticing as it is to imagine that, by saving us time on mundane tasks, LLMs will provide us with more time to think deeply and develop projects completely, our results temper such hopes.

Figures

Figures reproduced from arXiv: 2607.17397 by the authors.

Figure 1
Figure 1. Modeling optimal policies. [A] Schematic diagram of the model with a discovery [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

175 extracted references · 1 canonical work pages

  1. [2]

    The Journal of Philosophy , author =

    The. The Journal of Philosophy , author =. 2003 , pages =. doi:10.5840/jphil2003100224 , language =

    work page doi:10.5840/jphil2003100224 2003

  2. [3]

    Readings in industrial economics: Volume two: Private enterprise and state intervention , pages=

    Economic welfare and the allocation of resources for invention , author=. Readings in industrial economics: Volume two: Private enterprise and state intervention , pages=. 1962 , publisher=

    1962

  3. [4]

    Economic theory , volume=

    A formal model of theory choice in science , author=. Economic theory , volume=. 1999 , publisher=

    1999

  4. [5]

    Episteme , volume=

    Optimal publishing strategies , author=. Episteme , volume=. 2009 , publisher=

    2009

  5. [6]

    Synthese , volume=

    Is a bird in the hand worth two in the bush? Or, whether scientists should publish intermediate results , author=. Synthese , volume=. 2014 , publisher=

    2014

  6. [7]

    Annals of internal medicine , volume=

    Manuscript quality before and after peer review and editing at Annals of Internal Medicine , author=. Annals of internal medicine , volume=. 1994 , publisher=

    1994

  7. [8]

    American sociological review , volume=

    Tradition and innovation in scientists’ research strategies , author=. American sociological review , volume=. 2015 , publisher=

    2015

  8. [9]

    1979 , publisher=

    TEX and METAFONT: New directions in typesetting , author=. 1979 , publisher=

    1979

Show all 175 references
  1. [10]

    Science , volume=

    Primer-directed enzymatic amplification of DNA with a thermostable DNA polymerase , author=. Science , volume=. 1988 , publisher=

    1988

  2. [11]

    Journal of the American Chemical Society , volume=

    The ultra-centrifuge, a new instrument for the determination of size and distribution of size of particle in amicroscopic colloids , author=. Journal of the American Chemical Society , volume=. 1924 , publisher=

    1924

  3. [12]

    nature , volume=

    Highly accurate protein structure prediction with AlphaFold , author=. nature , volume=. 2021 , publisher=

    2021

  4. [13]

    2012 , publisher=

    Geometry of voting , author=. 2012 , publisher=

    2012

  5. [14]

    2004 , publisher=

    Auctions: theory and practice , author=. 2004 , publisher=

    2004

  6. [15]

    2009 , publisher=

    Theory of decision under uncertainty , author=. 2009 , publisher=

    2009

  7. [16]

    2002 , publisher=

    Political economics: explaining economic policy , author=. 2002 , publisher=

    2002

  8. [17]

    1991 , publisher=

    Game theory , author=. 1991 , publisher=

    1991

  9. [18]

    2009 , publisher=

    The theory of incentives: the principal-agent model , author=. 2009 , publisher=

    2009

  10. [19]

    1995 , publisher=

    Microeconomic theory , author=. 1995 , publisher=

    1995

  11. [20]

    Journal of Economic Perspectives , volume=

    Optimal taxation in theory and practice , author=. Journal of Economic Perspectives , volume=. 2009 , publisher=

    2009

  12. [21]

    Encyclopedia of Complexity and Systems Science , editor=

    Signaling Games , author=. Encyclopedia of Complexity and Systems Science , editor=

  13. [22]

    2015 , publisher=

    Contest theory: Incentive mechanisms and ranking methods , author=. 2015 , publisher=

    2015

  14. [23]

    1990 , publisher=

    Two-Sided Matching: A Study in Game-Theoretic Modeling and Analysis , author=. 1990 , publisher=

    1990

  15. [24]

    2020 , publisher=

    The knowledge machine: How irrationality created modern science , author=. 2020 , publisher=

    2020

  16. [25]

    Journal of Economic Literature , pages=

    The economics of science , author=. Journal of Economic Literature , pages=

  17. [26]

    Research policy , volume=

    Toward a new economics of science , author=. Research policy , volume=. 1994 , publisher=

    1994

  18. [27]

    2019 , journal=

    Identification of and correction for publication bias , author=. 2019 , journal=

    2019

  19. [28]

    The annals of statistics , pages=

    Agreeing to disagree , author=. The annals of statistics , pages=

  20. [29]

    Nature Human Behaviour , volume=

    Redefine statistical significance , author=. Nature Human Behaviour , volume=

  21. [30]

    2019 , institution=

    Publication Bias and Editorial Statement on Negative Findings , author=. 2019 , institution=

    2019

  22. [31]

    The American Statistician , volume=

    An Introduction to Second-Generation p -Values , author=. The American Statistician , volume=

  23. [32]

    The American Statistician , volume=

    P-value precision and reproducibility , author=. The American Statistician , volume=

  24. [33]

    USSR computational mathematics and mathematical physics , volume=

    The relaxation method of finding the common point of convex sets and its application to the solution of problems in convex programming , author=. USSR computational mathematics and mathematical physics , volume=

  25. [34]

    Monthly Weather Review , volume=

    Verification of forecasts expressed in terms of probability , author=. Monthly Weather Review , volume=

  26. [35]

    American Economic Journal: Applied Economics , volume=

    Star wars: The empirics strike back , author=. American Economic Journal: Applied Economics , volume=

  27. [36]

    The Lancet , volume=

    No level of alcohol consumption improves health , author=. The Lancet , volume=

  28. [37]

    Greater Statistical Stringency

    The World of Research Has Gone Berserk: Modeling the Consequences of Requiring “Greater Statistical Stringency” for Scientific Publication , author=. The American Statistician , volume=

  29. [38]

    Journal of Economic Literature , volume=

    Transparency, reproducibility, and the credibility of economics research , author=. Journal of Economic Literature , volume=

  30. [39]

    2012 , publisher=

    Elements of information theory , author=. 2012 , publisher=

    2012

  31. [40]

    Journal of the American Medical Association , volume=

    Factors influencing publication of research results: follow-up of applications submitted to two institutional review boards , author=. Journal of the American Medical Association , volume=

  32. [41]

    Biometrics , volume=

    Trim and fill: a simple funnel-plot--based method of testing and adjusting for publication bias in meta-analysis , author=. Biometrics , volume=. 2000 , publisher=

    2000

  33. [42]

    Frontiers in psychology , volume=

    Replication, falsification, and the crisis of confidence in social psychology , author=. Frontiers in psychology , volume=

  34. [43]

    Working paper available at SSRN , year=

    What you see is all there is , author=. Working paper available at SSRN , year=

  35. [44]

    Science , volume=

    Science of science , author=. Science , volume=. 2018 , publisher=

    2018

  36. [45]

    Which findings should be published? , year =

    Alexander Frankel and Kasy, Maximilian , note =. Which findings should be published? , year =

  37. [46]

    American Economic Review (conditionally accepted) , year=

    Quantifying information and uncertainty , author=. American Economic Review (conditionally accepted) , year=

  38. [47]

    The American Scientist , volume=

    The statistical crisis in science , author=. The American Scientist , volume=

  39. [48]

    Researcher incentives and empirical methods , author=

  40. [49]

    Grantham , title =

    Neal S. Grantham , title =

  41. [50]

    2019 , publisher=

    Grantham, Neal S and Guan, Yawen and Reich, Brian J and Borer, Elizabeth T and Gross, Kevin , journal=. 2019 , publisher=

    2019

  42. [51]

    The American Naturalist , volume=

    A general, synthetic model for predicting biodiversity gradients from environmental geometry , author=. The American Naturalist , volume=. 2016 , publisher=

    2016

  43. [52]

    PLoS Biology , volume=

    Contest models highlight inherent inefficiencies of scientific funding competitions , author=. PLoS Biology , volume=. 2019 , publisher=

    2019

  44. [53]

    Nature Human Behaviour , volume=

    Mapping the universe of registered reports , author=. Nature Human Behaviour , volume=. 2018 , publisher=

    2018

  45. [54]

    PLoS Medicine , volume=

    Why most published research findings are false , author=. PLoS Medicine , volume=

  46. [55]

    2012 , publisher=

    Forecast verification: a practitioner's guide in atmospheric science , author=. 2012 , publisher=

    2012

  47. [56]

    The American Statistician , volume=

    Before p< 0.05 to Beyond p< 0.05 : Using History to Contextualize p -Values and Significance Testing , author=. The American Statistician , volume=

  48. [57]

    The Journal of Philosophy , volume=

    The division of cognitive labor , author=. The Journal of Philosophy , volume=

  49. [58]

    Rebecca Anne Lyzinski , title =

  50. [59]

    PloS one , volume=

    Conservative tests under satisficing models of publication bias , author=. PloS one , volume=

  51. [60]

    Ecology , volume=

    Controls over native perennial grass exclusion and persistence in California grasslands invaded by annuals , author=. Ecology , volume=. 2015 , publisher=

    2015

  52. [61]

    The American Naturalist , volume=

    Within-host niche differences and fitness trade-offs promote coexistence of plant viruses , author=. The American Naturalist , volume=. 2016 , publisher=

    2016

  53. [62]

    On the problem of the most efficient tests of statistical hypotheses , author=

    IX. On the problem of the most efficient tests of statistical hypotheses , author=. Philosophical Transactions of the Royal Society of London. Series A, Containing Papers of a Mathematical or Physical Character , volume=

  54. [63]

    Publication bias and the canonization of false facts , author=. e

  55. [64]

    Scientific utopia:

    Nosek, Brian A and Spies, Jeffrey R and Motyl, Matt , journal=. Scientific utopia:

  56. [65]

    2014 , volume =

    Registered reports: A method to increase the credibility of published results , author=. 2014 , volume =

    2014

  57. [66]

    Perspectives on Psychological Science , volume=

    Is the replicability crisis overblown? Three arguments examined , author=. Perspectives on Psychological Science , volume=

  58. [67]

    Journal of Clinical Epidemiology , volume=

    The distribution of P-values in medical research articles suggested selective reporting associated with statistical significance , author=. Journal of Clinical Epidemiology , volume=. 2017 , publisher=

    2017

  59. [68]

    Synthese , volume=

    Aggregating incoherent agents who disagree , author=. Synthese , volume=

  60. [69]

    2016 , journal =

    Introducing cross-referee commenting in peer review , author=. 2016 , journal =

    2016

  61. [70]

    The EMBO Journal , volume=

    A transparent black box , author=. The EMBO Journal , volume=

  62. [71]

    , author=

    The file drawer problem and tolerance for null results. , author=. Psychological bulletin , volume=

  63. [72]

    2005 , publisher=

    Rothstein, Hannah R and Sutton, Alexander J and Borenstein, Michael , title=. 2005 , publisher=

    2005

  64. [73]

    Scientific publishing: the e

    Schekman, Randy and Watt, Fiona and Weigel, Detlef , journal=. Scientific publishing: the e

  65. [74]

    Ecology Letters , volume=

    The community ecology of pathogens: coinfection, coexistence and community composition , author=. Ecology Letters , volume=. 2015 , publisher=

    2015

  66. [75]

    2019 , volume =

    Open science and modified funding lotteries can impede the natural selection of bad science , author=. 2019 , volume =

    2019

  67. [76]

    Evolutionary Applications , volume=

    Environmental influences on mosquito foraging and integrated vector management can delay the evolution of behavioral resistance , author=. Evolutionary Applications , volume=. 2016 , publisher=

    2016

  68. [77]

    Malaria Journal , volume=

    Evolution of host preference in anthropophilic mosquitoes , author=. Malaria Journal , volume=. 2018 , publisher=

    2018

  69. [78]

    The Journal of Philosophy , volume=

    The role of the priority rule in science , author=. The Journal of Philosophy , volume=

  70. [79]

    Basic and Applied Social Psychology , volume =

    David Trafimow and Michael Marks , title =. Basic and Applied Social Psychology , volume =

  71. [80]

    Wasserstein, Ronald L and Lazar, Nicole A , journal=. The

  72. [81]

    Wasserstein and Allen L

    Ronald L. Wasserstein and Allen L. Schirm and Nicole A. Lazar , title =. The American Statistician , volume =

  73. [82]

    Monthly Weather Review , volume=

    Kullback--Leibler divergence as a forecast skill score with classic reliability--resolution--uncertainty decomposition , author=. Monthly Weather Review , volume=

  74. [83]

    Erkenntnis , volume=

    The epistemic benefit of transient diversity , author=. Erkenntnis , volume=

  75. [84]

    Journal of the Royal Society of Medicine , volume=

    Peer review: a flawed process at the heart of science and journals , author=. Journal of the Royal Society of Medicine , volume=. 2006 , publisher=

    2006

  76. [85]

    Science , volume=

    Chance and consensus in peer review , author=. Science , volume=. 1981 , publisher=

    1981

  77. [86]

    BioScience , volume=

    Is peer review a game of chance? , author=. BioScience , volume=. 2006 , publisher=

    2006

  78. [87]

    1990 , publisher=

    Peerless science: Peer review and US science policy , author=. 1990 , publisher=

    1990

  79. [88]

    Journal of the American Society for Information Science and Technology , volume=

    Bias in peer review , author=. Journal of the American Society for Information Science and Technology , volume=. 2013 , publisher=

    2013

  80. [89]

    An analysis of preliminary and post-discussion priority scores for grant applications peer reviewed by the Center for Scientific Review at the

    Martin, Michael R and Kopstein, Andrea and Janice, Joy M , journal=. An analysis of preliminary and post-discussion priority scores for grant applications peer reviewed by the Center for Scientific Review at the. 2010 , publisher=

    2010

  81. [90]

    Journal of the American Medical Association , volume=

    Effects of editorial peer review: A systematic review , author=. Journal of the American Medical Association , volume=. 2002 , publisher=

    2002

  82. [91]

    The Journal of Economic Education , volume=

    Choosing partners: A classroom experiment , author=. The Journal of Economic Education , volume=. 2013 , publisher=

    2013

  83. [92]

    The transmission sense of information

    Response to commentaries on “The transmission sense of information” , author=. Biology & Philosophy , volume=. 2011 , publisher=

    2011

  84. [93]

    The American Naturalist , volume=

    When unreliable cues are good enough , author=. The American Naturalist , volume=. 2013 , publisher=

    2013

  85. [94]

    Evolution, Medicine, and Public Health , volume=

    Adaptive behavior can produce maladaptive anxiety due to individual differences in experience , author=. Evolution, Medicine, and Public Health , volume=. 2016 , publisher=

    2016

  86. [95]

    Science , volume=

    Can ignorance promote democracy? , author=. Science , volume=. 2011 , publisher=

    2011

  87. [96]

    Proceedings of the Royal Society B: Biological Sciences , volume=

    Between cheap and costly signals: the evolution of partially honest communication , author=. Proceedings of the Royal Society B: Biological Sciences , volume=. 2013 , publisher=

    2013

  88. [97]

    Journal of The Royal Society Interface , volume=

    Honest signalling with costly gambles , author=. Journal of The Royal Society Interface , volume=. 2013 , publisher=

    2013

  89. [98]

    Evolution, Medicine, and Public Health , volume=

    Depression and anxiety: Maladaptive byproducts of adaptive mechanisms , author=. Evolution, Medicine, and Public Health , volume=. 2016 , publisher=

    2016

  90. [99]

    2023 , eprint=

    Rationalizing risk aversion in science , author=. 2023 , eprint=

    2023

  91. [100]

    The Annals of Mathematical Statistics , pages=

    Upper and lower probabilities induced by a multivalued mapping , author=. The Annals of Mathematical Statistics , pages=. 1967 , volume =

    1967

  92. [101]

    1976 , publisher=

    A mathematical theory of evidence , author=. 1976 , publisher=

    1976

  93. [102]

    The Annals of Statistics , pages=

    Prior envelopes based on belief functions , author=. The Annals of Statistics , pages=

  94. [103]

    2009 , publisher=

    Robust statistics , author=. 2009 , publisher=

    2009

  95. [104]

    Ambiguity and the [B]ayesian paradigm

    Gilboa, Itzhak and Marinacci, Massimo. Ambiguity and the [B]ayesian paradigm. Advances in Economics and Econometrics: Theory and Applications, Tenth World Congress of the Econometric Society. 2013

    2013

  96. [105]

    2014 , publisher=

    Introduction to imprecise probabilities , author=. 2014 , publisher=

    2014

  97. [106]

    The review of economic studies , volume=

    Learning from neighbours , author=. The review of economic studies , volume=. 1998 , publisher=

    1998

  98. [107]

    Proceedings of the National Academy of Sciences , volume=

    Partial connectivity increases cultural accumulation within groups , author=. Proceedings of the National Academy of Sciences , volume=. 2016 , publisher=

    2016

  99. [108]

    Philosophy of Science , volume=

    In epistemic networks, is less really more? , author=. Philosophy of Science , volume=. 2017 , publisher=

    2017

  100. [109]

    Journal of Statistical Theory and Practice , volume=

    Imprecision and prior-data conflict in generalized Bayesian inference , author=. Journal of Statistical Theory and Practice , volume=. 2009 , publisher=

    2009

  101. [110]

    Journal of Economic Theory , volume=

    Dynamics of inductive inference in a unified framework , author=. Journal of Economic Theory , volume=

  102. [111]

    Journal of Economic Theory , volume=

    Learning (to disagree?) in large worlds , author=. Journal of Economic Theory , volume=. 2022 , publisher=

    2022

  103. [112]

    Bayes' theorem for

    Wasserman, Larry A and Kadane, Joseph B , journal=. Bayes' theorem for

  104. [113]

    1988 , publisher=

    Possibility theory , author=. 1988 , publisher=

    1988

  105. [114]

    1991 , publisher=

    Statistical reasoning with imprecise probabilities , author=. 1991 , publisher=

    1991

  106. [115]

    Cognitive Science , volume=

    Languages and designs for probability judgment , author=. Cognitive Science , volume=

  107. [116]

    2021 , journal=

    Judicious judgment meets unsettling updating: dilation, sure loss and Simpson’s paradox , author=. 2021 , journal=

    2021

  108. [117]

    Statistical decision theory and

    Berger, James O , year=. Statistical decision theory and

  109. [118]

    Administrative science quarterly , volume=

    The network structure of exploration and exploitation , author=. Administrative science quarterly , volume=. 2007 , publisher=

    2007

  110. [119]

    Philosophy of science , volume=

    The communication structure of epistemic communities , author=. Philosophy of science , volume=. 2007 , publisher=

    2007

  111. [120]

    Synthese , volume=

    How should we promote transient diversity in science? , author=. Synthese , volume=. 2023 , publisher=

    2023

  112. [121]

    Journal of mathematical economics , volume=

    Maxmin expected utility with non-unique prior , author=. Journal of mathematical economics , volume=. 1989 , publisher=

    1989

  113. [122]

    The quarterly journal of economics , volume=

    Risk, ambiguity, and the Savage axioms , author=. The quarterly journal of economics , volume=. 1961 , publisher=

    1961

  114. [123]

    Journal of risk and uncertainty , volume=

    Recent developments in modeling preferences: Uncertainty and ambiguity , author=. Journal of risk and uncertainty , volume=. 1992 , publisher=

    1992

  115. [124]

    Annales de l'Institut Fourier , volume=

    Theory of capacities , author=. Annales de l'Institut Fourier , volume=

  116. [125]

    Theory and Decision , volume=

    A minimal extension of Bayesian decision theory , author=. Theory and Decision , volume=. 2016 , publisher=

    2016

  117. [126]

    2009 , publisher=

    Rational decisions , author=. 2009 , publisher=

    2009

  118. [127]

    Games and Economic Behavior , volume=

    Merging, reputation, and repeated games with incomplete information , author=. Games and Economic Behavior , volume=. 1999 , publisher=

    1999

  119. [128]

    Econometrica , pages=

    Subjective probability and expected utility without additivity , author=. Econometrica , pages=. 1989 , volume=

    1989

  120. [129]

    2002 , booktitle=

    Peirce, Charles S , title =. 2002 , booktitle=

    2002

  121. [130]

    American Sociological Review , author =

    Priorities in. American Sociological Review , author =. 1957 , pages =. doi:10.2307/2089193 , language =

    1957 doi

  122. [131]

    The Bell Journal of Economics , volume=

    Moral Hazard and Observability , author=. The Bell Journal of Economics , volume=. 1979 , publisher=

    1979

  123. [132]

    2010 , author =

    Science as a process: an evolutionary account of the social and conceptual development of science , publisher =. 2010 , author =

    2010

  124. [133]

    Arrow and the ascent of modern economic theory , pages=

    Information disclosure and the economics of science and technology , author=. Arrow and the ascent of modern economic theory , pages=. 1987 , publisher=

    1987

  125. [134]

    Manuscript

    Further properties of the priority rule , author=. Manuscript. http://www.strevens.org/research/scistruc/MorePrior.pdf , year=

  126. [135]

    Journal of Economic Methodology , volume=

    Herding and the quest for credit , author=. Journal of Economic Methodology , volume=. 2013 , publisher=

    2013

  127. [136]

    arXiv preprint arXiv:1605.05822 , year=

    Why scientists chase big problems: individual strategy and social optimality , author=. arXiv preprint arXiv:1605.05822 , year=

    arXiv

  128. [137]

    The quarterly journal of economics , volume=

    Market structure and innovation , author=. The quarterly journal of economics , volume=. 1979 , publisher=

    1979

  129. [138]

    The Quarterly Journal of Economics , volume=

    Market structure and innovation: A reformulation , author=. The Quarterly Journal of Economics , volume=. 1980 , publisher=

    1980

  130. [139]

    Proceedings of the forty-third annual ACM symposium on Theory of computing , pages=

    Mechanisms for (mis) allocating scientific credit , author=. Proceedings of the forty-third annual ACM symposium on Theory of computing , pages=

  131. [140]

    The Journal of Philosophy , volume=

    The credit economy and the economic rationality of science , author=. The Journal of Philosophy , volume=. 2018 , publisher=

    2018

  132. [141]

    Nature , volume=

    Artificial intelligence and illusions of understanding in scientific research , author=. Nature , volume=. 2024 , publisher=

    2024

  133. [142]

    arXiv preprint arXiv:2107.00061 , year=

    All that's' human'is not gold: Evaluating human evaluation of generated text , author=. arXiv preprint arXiv:2107.00061 , year=

    arXiv

  134. [143]

    Proceedings of the National Academy of Sciences , volume=

    Human heuristics for AI-generated language are flawed , author=. Proceedings of the National Academy of Sciences , volume=. 2023 , publisher=

    2023

  135. [144]

    Scientific Reports , volume=

    AI-generated poetry is indistinguishable from human-written poetry and is rated more favorably , author=. Scientific Reports , volume=. 2024 , publisher=

    2024

  136. [145]

    arXiv preprint arXiv:2407.14452 , year=

    Undermining Mental Proof: How AI Can Make Cooperation Harder by Making Thinking Easier , author=. arXiv preprint arXiv:2407.14452 , year=

    arXiv

  137. [146]

    , author=

    The file drawer problem and tolerance for null results. , author=. Psychological bulletin , volume=. 1979 , publisher=

    1979

  138. [147]

    2025 , publisher=

    More than words: How to think about writing in the age of AI , author=. 2025 , publisher=

    2025

  139. [148]

    Convenience AI , author=

  140. [149]

    Proceedings of the National Academy of Sciences , volume=

    How should the advancement of large language models affect the practice of science? , author=. Proceedings of the National Academy of Sciences , volume=. 2025 , publisher=

    2025

  141. [150]

    ACM computing surveys , volume=

    Survey of hallucination in natural language generation , author=. ACM computing surveys , volume=. 2023 , publisher=

    2023

  142. [151]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Undermining mental proof: How ai can make cooperation harder by making thinking easier , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  143. [152]

    International Conference on Learning Representations , volume=

    Can llms generate novel research ideas? a large-scale human study with 100+ nlp researchers , author=. International Conference on Learning Representations , volume=

  144. [153]

    Proceedings of the 1st Workshop on NLP for Science (NLP4Science) , pages=

    Hypothesis generation with large language models , author=. Proceedings of the 1st Workshop on NLP for Science (NLP4Science) , pages=

  145. [154]

    Nature , volume=

    Autonomous chemical research with large language models , author=. Nature , volume=. 2023 , publisher=

    2023

  146. [155]

    International Conference on Learning Representations , volume=

    Discoverybench: Towards data-driven discovery with large language models , author=. International Conference on Learning Representations , volume=

  147. [156]

    Advances in neural information processing systems , volume=

    Autosurvey: Large language models can automatically write surveys , author=. Advances in neural information processing systems , volume=

  148. [157]

    Nature , pages=

    Synthesizing scientific literature with retrieval-augmented language models , author=. Nature , pages=. 2026 , publisher=

    2026

  149. [158]

    Nature Human Behaviour , pages=

    Quantifying large language model usage in scientific papers , author=. Nature Human Behaviour , pages=. 2025 , publisher=

    2025

  150. [159]

    arXiv preprint arXiv:2408.06292 , year=

    The ai scientist: Towards fully automated open-ended scientific discovery , author=. arXiv preprint arXiv:2408.06292 , year=

    arXiv

  151. [160]

    Findings of the Association for Computational Linguistics: EMNLP 2024 , pages=

    Blade: Benchmarking language model agents for data-driven science , author=. Findings of the Association for Computational Linguistics: EMNLP 2024 , pages=

    2024

  152. [161]

    ACM Transactions on Information Systems , volume=

    A survey on hallucination in large language models: Principles, taxonomy, challenges, and open questions , author=. ACM Transactions on Information Systems , volume=. 2025 , publisher=

    2025

  153. [162]

    Proceedings of the 60th annual meeting of the association for computational linguistics (volume 1: long papers) , pages=

    Truthfulqa: Measuring how models mimic human falsehoods , author=. Proceedings of the 60th annual meeting of the association for computational linguistics (volume 1: long papers) , pages=

  154. [163]

    Scientific Reports , volume=

    Fabrication and errors in the bibliographic citations generated by ChatGPT , author=. Scientific Reports , volume=. 2023 , publisher=

    2023

  155. [164]

    International Conference on Learning Representations , volume=

    Towards understanding sycophancy in language models , author=. International Conference on Learning Representations , volume=

  156. [165]

    Nature , volume=

    AI models collapse when trained on recursively generated data , author=. Nature , volume=. 2024 , publisher=

    2024

  157. [166]

    Advances in Neural Information Processing Systems , volume=

    Language models don't always say what they think: Unfaithful explanations in chain-of-thought prompting , author=. Advances in Neural Information Processing Systems , volume=

  158. [167]

    Proceedings of the National Academy of Sciences , volume=

    The diffusion of large language models in published academic articles , author=. Proceedings of the National Academy of Sciences , volume=. 2026 , publisher=

    2026

  159. [168]

    arXiv preprint arXiv:2605.07723 , year=

    LLM hallucinations in the wild: Large-scale evidence from non-existent citations , author=. arXiv preprint arXiv:2605.07723 , year=

    arXiv

  160. [169]

    , title =

    Ross, Sheldon M. , title =. 2014 , isbn =

    2014

  161. [170]

    Royal Society Open Science , year =

    Peters, Uwe and Chin-Yee, Benjamin , title =. Royal Society Open Science , year =

  162. [171]

    Theoretical population biology , volume=

    Optimal foraging, the marginal value theorem , author=. Theoretical population biology , volume=. 1976 , publisher=

    1976

  163. [172]

    The Lancet Gastroenterology & Hepatology , volume=

    Endoscopist deskilling risk after exposure to artificial intelligence in colonoscopy: a multicentre, observational study , author=. The Lancet Gastroenterology & Hepatology , volume=. 2025 , publisher=

    2025

  164. [173]

    PLoS Biology , volume=

    Screening, sorting, and the feedback cycles that imperil peer review , author=. PLoS Biology , volume=. 2026 , publisher=

    2026

  165. [174]

    Science , volume=

    Scientific production in the era of large language models , author=. Science , volume=. 2025 , publisher=

    2025

  166. [175]

    Science , volume=

    After science , author=. Science , volume=. 2025 , publisher=

    2025

  167. [176]

    Nature , volume=

    The uncritical adoption of AI in science is alarming—we urgently need guard rails , author=. Nature , volume=. 2026 , publisher=

    2026

Pith tools

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