Pith. sign in

REVIEW 2 major objections 1 minor

War in the Abstract: The Rise and Consequences of Militarized Language in Scientific Communication

T0 review · 2 major / 1 minor · reviewed 2026-06-26 · grok-4.3

Pith's one-line read Militarized terms in scientific abstracts rose 48% from 2010 to 2025 and reduce credibility when used.

desk verdict Large-scale count of militaristic terms in 21M abstracts plus a framing experiment, but the term list lacks validation so the trends rest on shaky ground. read the letter →

arxiv 2606.23462 v2 pith:UK2JTUWM submitted 2026-06-22 cs.CL cs.AIcs.CYcs.DL

classification cs.CLcs.AIcs.CYcs.DL
keywords militarizedlanguagescientificabstractswarframingcredibilityconflictcorrelationpersuasionexperiment
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

The paper tracks the growing use of militaristic vocabulary across 21.4 million scientific abstracts in two large databases. It reports steady increases that speed up after 2019 and line up with real-world conflict data at both country and yearly levels. A controlled experiment then tests the effects of this language by presenting the same scientific content with or without war framing. Results show drops in how credible the work seems along with smaller drops in willingness to fund it or back related policies.

What carries the argument

Corpus analysis of a predefined list of militaristic terms applied to 21.4 million abstracts, paired with a within-subject experiment measuring framing effects on credibility and support measures.

What would settle it

Re-analysis of the same 21.4 million papers with a different term list or different conflict dataset that finds no rise or no correlation with conflicts would undermine the prevalence and alignment results.

Watch

Extended reading notes

Core claim

Between 2010 and 2025 the presence of militaristic terms in scientific abstracts rose 48% in OpenAlex and 32% in PubMed, with the rise accelerating sharply after 2019 and aligning with conflict data at country and annual scales; war framing reduced credibility by a mean shift of -0.18 Likert units, funding willingness, and policy support.

Load-bearing premise

The predefined list of militaristic terms used to label abstracts accurately captures the intended construct without substantial false positives or context-dependent misclassification.

Editorial extensions

If this is right

  • Social sciences show the highest levels of militaristic language while engineering and computer science show the fastest growth.
  • Abstracts from the Global South display the fastest rise.
  • The COVID and post-2022 large-language-model periods saw both a rise in use and a narrowing of the gap between native-English and non-English authors.
  • War framing produces a small trend-level increase in sense of urgency alongside the negative effects on credibility and support.

Reading between the lines

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

  • Continued growth in such language could gradually lower public willingness to act on scientific findings.
  • Disciplines might develop style guidelines that discourage militaristic metaphors in abstracts to protect perceived trustworthiness.
  • Similar experiments could test whether other loaded metaphors produce comparable drops in support.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 1 minor

Summary. The paper analyzes 21.4 million scientific abstracts (2010-2025) from OpenAlex and PubMed, reporting a 48% and 32% rise in militaristic terms respectively, with post-2019 acceleration (cross-database r=0.96) and correlations to real-world conflict data (r=0.77-0.84). It additionally reports results from a within-subject experiment (N=801, 32,040 trials) in which war framing reduced credibility (d_z=-0.28), funding willingness (d_z=-0.12), and policy support (d_z=-0.08).

Significance. Large sample sizes, cross-database consistency, conflict-aligned correlations, and effect sizes reported with confidence intervals provide directional support for the claims if the measurement is valid. The experimental component supplies causal evidence on potential downstream effects. If the term-list measurement holds, the work identifies a measurable shift in scientific communication with possible implications for credibility and persuasion.

major comments (2)
  1. [Methods (corpus analysis)] Corpus analysis / term-list construction (described in the methods for the 21.4M-abstract study): no validation, human coding, precision/recall, or false-positive audit is reported for the fixed militaristic term list. Polysemous terms (e.g., 'campaign', 'target', 'attack', 'defense', 'strategy') occur frequently in non-militaristic technical contexts; without evidence that such usages were excluded or quantified, the 48%/32% prevalence rises, post-2019 acceleration, Global-South patterns, and conflict correlations (r=0.77-0.84) cannot be unambiguously attributed to militarized language rather than topic or database shifts.
  2. [Methods (experiment)] Experimental methods (N=801 study): the exact framing stimuli, control sentences, and any pre-registration or attention-check details are not supplied. These omissions are load-bearing for interpreting the reported effect sizes (credibility d_z=-0.28, funding d_z=-0.12) as specifically due to militaristic language rather than other lexical or affective differences between conditions.
minor comments (1)
  1. [Abstract] Abstract: the size of the militaristic term list and one or two concrete examples would help readers assess face validity without needing the full methods.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for these constructive comments on methodological transparency. We agree that both the corpus term-list validation and the experimental stimuli details require fuller reporting. We will revise the manuscript to address these points directly.

read point-by-point responses
  1. Referee: [Methods (corpus analysis)] Corpus analysis / term-list construction (described in the methods for the 21.4M-abstract study): no validation, human coding, precision/recall, or false-positive audit is reported for the fixed militaristic term list. Polysemous terms (e.g., 'campaign', 'target', 'attack', 'defense', 'strategy') occur frequently in non-militaristic technical contexts; without evidence that such usages were excluded or quantified, the 48%/32% prevalence rises, post-2019 acceleration, Global-South patterns, and conflict correlations (r=0.77-0.84) cannot be unambiguously attributed to militarized language rather than topic or database shifts.

    Authors: We agree this is a substantive limitation in the current reporting. The methods section does not include a validation study, precision/recall metrics, or false-positive audit for the fixed term list, and polysemy in terms such as 'campaign' or 'target' is a genuine concern that could inflate counts. In the revised manuscript we will add a dedicated subsection on term-list construction that reports any pilot human coding performed, quantifies false-positive rates on a sampled subset of abstracts, and discusses mitigation steps for polysemy. We will also note that while cross-database consistency (r=0.96) and alignment with external conflict data provide convergent support, these do not substitute for direct validation; the added audit will allow readers to assess the degree of contamination. revision: yes

  2. Referee: [Methods (experiment)] Experimental methods (N=801 study): the exact framing stimuli, control sentences, and any pre-registration or attention-check details are not supplied. These omissions are load-bearing for interpreting the reported effect sizes (credibility d_z=-0.28, funding d_z=-0.12) as specifically due to militaristic language rather than other lexical or affective differences between conditions.

    Authors: We concur that the stimuli, controls, pre-registration status, and attention-check procedures must be supplied for the effects to be interpretable. The revised manuscript will include the full set of war-framed and control sentences in an appendix, along with all attention-check wording, exclusion criteria, and data-quality metrics. The study was not pre-registered; we will state this explicitly as a limitation. These additions will enable direct evaluation of whether the observed shifts (e.g., d_z = -0.28 on credibility) are attributable to militaristic framing rather than other lexical differences. revision: yes

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: direct corpus counts, external correlations, and independent experiment

full rationale

The paper reports prevalence trends from applying a fixed term list to 21.4 million external abstracts (OpenAlex/PubMed), correlations against the independent Uppsala Conflict Data Program, and effect sizes from a separate N=801 within-subject experiment. No equations, fitted parameters, self-definitional constructs, or load-bearing self-citations are present that would make any reported quantity reduce to the authors' own inputs by construction. The measurement pipeline and causal test remain independent of the target statistics.

Assumptions & free parameters 0 free parameters · 1 assumptions · 0 invented entities

Based solely on the abstract, the central claims rest on an unvalidated dictionary of militaristic terms and standard statistical assumptions for correlations and effect-size calculations; no free parameters, invented entities, or additional axioms are explicitly described.

assumptions (1)
  • domain assumption A fixed list of militaristic terms validly and exhaustively measures the construct of militarized language in abstracts
    Required to interpret all prevalence, correlation, and experimental results

how reviews work

0 comments
Cite this review

Pith. "Pith review of War in the Abstract: The Rise and Consequences of Militarized Language in Scientific Communication." pith.science (2026). https://pith.science/paper/UK2JTUWM

@misc{pith2026260623462,
  author       = {Pith},
  title        = {Pith review of: War in the Abstract: The Rise and Consequences of Militarized Language in Scientific Communication},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UK2JTUWM}},
  note         = {Machine review of arXiv:2606.23462}
}
abstract

Scientists do not, by profession, wage war. Yet warfare's vocabulary consistently appears in their abstracts. To quantify the extent to which warfare's vocabulary pervades scientific abstracts, we analyze 21.4 million papers (2010-2025; OpenAlex, PubMed). We additionally run a within-subject war-framing experiment ($N = 801$; 32{,}040 trials) designed to provide causal insight into the effects of militaristic language on persuasion. Between 2010 and 2025, the presence of militaristic terms in scientific abstracts rose 48\% in OpenAlex and 32\% in PubMed, with the rise accelerating sharply after 2019 (cross-database $r = 0.96$, $p < 10^{-8}$). The prevalence of militaristic language is conflict-aligned at both country and annual scales (Uppsala Conflict Data Program; $r = 0.77$-$0.84$), with the abstracts from the Global South displaying the fastest rise in militaristic language. Among disciplines, social sciences leads in level of such language while engineering and computer science lead in growth. The COVID and post-2022 large-language-model eras also accelerated the rise and narrowed the language gap between native-English-speaking and non-English-speaking countries. In our follow-up experiment, we found that war framing reduced credibility (mean shift $-0.176$ Likert units, 95\% CI $[-0.21, -0.14]$; $d_z = -0.28$, $p < 10^{-20}$), funding willingness ($d_z = -0.12$) and policy support ($d_z = -0.08$), with a trend-level increase in sense of urgency ($d_z = +0.07$). Collectively, findings reveal that while scientific abstracts drift toward warfare, the use of militaristic language may erode credibility, funding willingness, and policy support.

Figures

Figures reproduced from arXiv: 2606.23462 by the authors.

Figure 1
Figure 1. Study design and analytical pipeline for quantifying militarized language in scientific publications. We assembled a corpus of approximately 29 million peer-reviewed articles published between 2010 and 2025 from two complementary bibliographic databases: OpenAlex (multidisciplinary) and PubMed (biomedical). After harmonization, deduplication and removal of records lacking abstracts or titles, N ≈ 21.4 million papers… view at source ↗
Figure 2
Figure 2. Temporal dynamics of use of militarized terms in scientific publications, 2010-2025. A. Prevalence of Tier 1 (T1, direct military; crimson), Tier 2 (T2, combat; amber), and Tier 3 (T3, ambiguous; blue) terms in paper titles and abstracts across OpenAlex (solid lines) and PubMed (dashed lines). B. Title versus abstract prevalence for each tier, pooled across both databases. Solid lines denote titles; dashed lines den… view at source ↗
Figure 3
Figure 3. Geographic variation in the prevalence of militarized scientific language. A. Side-by-side world choropleths of the prevalence of militaristic terms by country of author affiliation in 2010 (left) and 2024 (right). Countries with fewer than 10,000 cumulative papers across 2010-2025 are shown in gray. B. Regional heatmap showing union prevalence by year for ten geopolitical groupings, ordered by overall prevalence. C… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Disciplinary patterns of militarized language adoption. A. Lollipop plot of mean prevalence of militaristic terms by discipline, with horizontal bars showing the range across years (2010-2025). B. Stacked horizontal bars decomposing mean prevalence into T1 (crimson), T…
Figure 5
Figure 5. Figure 5: Association between armed conflict and militarized scientific language. A. Global overlay of the number of active armed conflicts per year (UCDP/PRIO, gray bars; left axis) and pooled tier-specific prevalence (colored lines; right axis). B. Violin and box plots compari…
Figure 6
Figure 6. Figure 6: Sensitivity analyses and confound controls. A. Pandemic sensitivity analy￾sis. Tier-specific prevalence for non-pandemic papers (solid) versus pandemic-related papers (dashed). B. Pre/post-2020 structural break. Scatter points show pooled yearly prevalence for T1 (circ…
Figure 7
Figure 7. Figure 7: Survey experiment: effect of militarized framing on public perceptions of science. A. Forest plot of the within-subject effect of war-framed versus neutrally-framed scientific abstracts on four outcome measures (7-point Likert scales). Points show mean dif￾ferences; ho…

Discussion (0). Continue with ORCID to comment.

Pith tools

Reviewed June 26, 2026 · model on record in the stance chip above.