Post-publication Comment · Critical AI
Comment on “Local US officials' views on the impacts and governance of AI: Evidence from 2022 and 2023 survey waves”
Critical AI · published 2026-06-28 · v1.0 · CRIT-000025
Concerning: Sophia Hatz, Noemi Dreksler, Kevin Wei, Baobao Zhang · PLOS One · 2025
Why this paper was selected
Scale-the-full-text-cohort batch (white-space domain): full-text critique span-grounded to the gold-OA full text via the source store.
AI/AGI centrality 3/5 · societal relevance 4/5 · source-journal note: Off-monitored: PLOS ONE is a peer-reviewed, gold open-access journal not in the journal's monitored top-tier determination; disclosed off-list. Critiqued at full text.
Summary
This is a competent, transparent two-wave survey (1028 completes; 524 in 2022, 504 in 2023) of US local elected officials' views on AI impacts and regulation, fielded by CivicPulse, using a planned-missing design with MICE imputation, survey weights, and multiplicity correction. Its descriptive findings (majorities expecting surveillance, misinformation, and data-security risks; 64% supporting regulation) are well supported. The problems are mostly interpretive overreach. The most consequential is that the headline narrative, that Republicans shifted to majority support between 2022 and 2023, rests on an unweighted descriptive jump (43% to 68%) computed across two independent samples of different people, not a tested differential change. Because the two waves are separate cross-sections (not a panel) and the abstract attributes the change to ChatGPT, the study cannot separate real attitude change from cohort and composition differences plus simultaneous national events. A second real gap: no response rate or contact denominator is reported, so nonresponse bias cannot be judged even though the authors invoke selection effects, and the 'somewhat representative' claim glosses over disclosed but unquantified skews toward larger, more Democratic, more educated districts. The authors disclose most issues candidly in a strong Limitations section, which lowers severity. (Two flaws in the draft critique cited sentences not present in the provided text and were re-anchored to verifiable abstract spans; a fourth, the I-dont-know coding issue, was dropped because it is author-disclosed and relied on numbers not in the provided sections.)
Central claims & evidence map
| Claim | Type | Evidence offered | Support | Overclaiming | Main weakness |
|---|---|---|---|---|---|
| The partisan-convergence story is foregrounded as a headline contribution, yet the only evidence carrying it is an unweighted, uncorrected descriptive cross-tab (43% to 68% among R | Methodological | the latter experienced a notable shift towards majority support between 2022 and 2023 | Moderate | Moderate | The partisan-convergence story is foregrounded as a headline contribution, yet the only evidence carrying it is an unweighted, uncorrected descriptive cross-tab (43% to 68% among Republicans) drawn fr |
| The design is two independent cross-sections of different respondents (Methods: 'two independent waves'; 524 in 2022 and 504 in 2023), not a panel, yet the abstract frames the 2022 | Causal | the survey captures attitudinal changes within the six months following the public release of ChatGPT | Strong | Moderate | The design is two independent cross-sections of different respondents (Methods: 'two independent waves'; 524 in 2022 and 504 in 2023), not a panel, yet the abstract frames the 2022-to-2023 difference |
| No response rate, contact count, or sampling-frame denominator is reported: Methods gives only completed responses (1028; 524 and 504 by wave) and the existence of a CivicPulse lis | Methodological | Sample medians are slightly skewed toward larger districts, are more Democratic, and are more educated than population median; generally, however, sample medians remain relatively close to population medians, suggesting that the samples are somewhat representative. | Moderate | Minor | No response rate, contact count, or sampling-frame denominator is reported: Methods gives only completed responses (1028; 524 and 504 by wave) and the existence of a CivicPulse list, but never states |
Per-claim assessment
C1. The partisan-convergence story is foregrounded as a headline contribution, yet the only evidence carrying it is an unweighted, uncorrected descriptive cross-tab (43% to 68% among R
The partisan-convergence story is foregrounded as a headline contribution, yet the only evidence carrying it is an unweighted, uncorrected descriptive cross-tab (43% to 68% among Republicans) drawn from two independent samples of different people across waves. A descriptive proportion difference between two cross-sections is not a tested differential-change effect, and the abstract escalates it to a 'notable shift' framed as a substantive over-time finding. The draft additionally asserts the year-by-party interaction is 'not statistically significant under the authors own corrected analysis,' but no such interaction result appears anywhere in the provided Methods/Results/Limitations/Abstract, so that specific claim cannot be verified from the text and the critique is held only to what the groundable spans support.
C2. The design is two independent cross-sections of different respondents (Methods: 'two independent waves'; 524 in 2022 and 504 in 2023), not a panel, yet the abstract frames the 2022
The design is two independent cross-sections of different respondents (Methods: 'two independent waves'; 524 in 2022 and 504 in 2023), not a panel, yet the abstract frames the 2022-to-2023 difference as ChatGPT-induced attitudinal change. With no within-person measurement and many co-occurring 2023 developments plus differing sample composition across waves, the difference cannot be causally attributed to ChatGPT nor separated from cohort/composition effects. The temporal-window phrasing ('within the six months following the public release of ChatGPT') imports a causal reading the design cannot identify.
C3. No response rate, contact count, or sampling-frame denominator is reported: Methods gives only completed responses (1028; 524 and 504 by wave) and the existence of a CivicPulse lis
No response rate, contact count, or sampling-frame denominator is reported: Methods gives only completed responses (1028; 524 and 504 by wave) and the existence of a CivicPulse list, but never states how many officials were sampled or invited. This makes nonresponse bias impossible to gauge, which directly undercuts the authors own appeal to selection effects, and the representativeness conclusion is hedged as 'somewhat representative' despite self-reported skews toward larger, more Democratic, and more educated districts that are left unquantified.
Scorecard
Sub-scores are 0–5 editorial judgements on fixed scales (higher is better, except methodological risk and overclaiming where higher is worse). They are contestable and open to a severity challenge from authors.
What the paper does
This is a competent, transparent two-wave survey (1028 completes; 524 in 2022, 504 in 2023) of US local elected officials' views on AI impacts and regulation, fielded by CivicPulse, using a planned-missing design with MICE imputation, survey weights, and multiplicity correction. Its descriptive findings (majorities expecting surveillance, misinformation, and data-security risks; 64% supporting regulation) are well supported. The problems are mostly interpretive overreach. The most consequential is that the headline narrative, that Republicans shifted to majority support between 2022 and 2023,
Statistical Inference
The partisan-convergence story is foregrounded as a headline contribution, yet the only evidence carrying it is an unweighted, uncorrected descriptive cross-tab (43% to 68% among Republicans) drawn from two independent samples of different people across waves. A descriptive proportion difference between two cross-sections is not a tested differential-change effect, and the abstract escalates it to a 'notable shift' framed as a substantive over-time finding. The draft additionally asserts the year-by-party interaction is 'not statistically significant under the authors own corrected analysis,' but no such interaction result appears anywhere in the provided Methods/Results/Limitations/Abstract, so that specific claim cannot be verified from the text and the critique is held only to what the groundable spans support.
Identification
The design is two independent cross-sections of different respondents (Methods: 'two independent waves'; 524 in 2022 and 504 in 2023), not a panel, yet the abstract frames the 2022-to-2023 difference as ChatGPT-induced attitudinal change. With no within-person measurement and many co-occurring 2023 developments plus differing sample composition across waves, the difference cannot be causally attributed to ChatGPT nor separated from cohort/composition effects. The temporal-window phrasing ('within the six months following the public release of ChatGPT') imports a causal reading the design cannot identify.
Sample Data
No response rate, contact count, or sampling-frame denominator is reported: Methods gives only completed responses (1028; 524 and 504 by wave) and the existence of a CivicPulse list, but never states how many officials were sampled or invited. This makes nonresponse bias impossible to gauge, which directly undercuts the authors own appeal to selection effects, and the representativeness conclusion is hedged as 'somewhat representative' despite self-reported skews toward larger, more Democratic, and more educated districts that are left unquantified.
What the paper does well
This is a transparent descriptive survey, and on the dimensions that matter most for a descriptive contribution it is solid: a planned-missing design handled with MICE (120 imputations, 200 iterations, predictive mean matching), survey weights folded into the imputation model, and Benjamini-Hochberg correction guarding against false positives. The core descriptive findings, majorities anticipating surveillance (83%), misinformation (69%), and reduced data security (64%), and 64% supporting government regulation, are simple proportions that need no causal identification and are well supported. The authors do not hide the weaknesses: the Limitations section explicitly flags selection effects, the IDK/Likert coding problem, missingness up to 69%, the power cost of multiplicity correction, and subgroup imprecision, and the partisan language in the abstract is hedged as a 'shift' rather than asserted as a ChatGPT-caused effect, so much of the overreach lives in framing emphasis rather than in unqualified causal assertion.
Strongest critique
The paper's most-promoted contribution, a Republican shift toward majority support for AI regulation between 2022 and 2023, is carried entirely by an unweighted descriptive cross-tab (43% to 68%) computed across two independent samples of different people, not by a tested differential-change effect. Because there is no panel (Methods: 'two independent waves'; separate 2022 and 2023 respondents), the over-time difference confounds genuine attitude updating with sample composition and a dense cluster of simultaneous 2023 developments, so neither the ChatGPT attribution nor the convergence claim is identified. Leading the abstract with this temporal-causal story ('captures attitudinal changes within the six months following the public release of ChatGPT' and a 'notable shift towards majority support') overstates what two cross-sections can bear.
Strongest fair defence
This is a transparent descriptive survey, and on the dimensions that matter most for a descriptive contribution it is solid: a planned-missing design handled with MICE (120 imputations, 200 iterations, predictive mean matching), survey weights folded into the imputation model, and Benjamini-Hochberg correction guarding against false positives. The core descriptive findings, majorities anticipating surveillance (83%), misinformation (69%), and reduced data security (64%), and 64% supporting government regulation, are simple proportions that need no causal identification and are well supported. The authors do not hide the weaknesses: the Limitations section explicitly flags selection effects, the IDK/Likert coding problem, missingness up to 69%, the power cost of multiplicity correction, and subgroup imprecision, and the partisan language in the abstract is hedged as a 'shift' rather than asserted as a ChatGPT-caused effect, so much of the overreach lives in framing emphasis rather than in unqualified causal assertion.
Conclusion
A methodologically careful, transparency-forward descriptive survey whose value lies in mapping local US officials' AI attitudes, undermined chiefly by interpretive overreach: the headline partisan-convergence-after-ChatGPT narrative is foregrounded in the abstract despite resting on an unweighted descriptive cross-tab and an independent-cross-section design that cannot identify change or its cause. The descriptive results (risk anticipations, 64% regulation support, specific-policy majorities) are trustworthy; the temporal, causal, and partisan-shift claims should be read as suggestive at best. Severity is moderate because the authors disclose most design weaknesses candidly in a strong Limitations section and the under-supported claims are largely confined to abstract framing rather than the proportions themselves, but the mismatch between what the abstract emphasizes and what two cross-sections support is real and would mislead a casual reader. Note on the draft under review: two of its four proposed flaws cited verbatim spans (the 'interaction between year and party is not statistically significant' sentence and the 'design was not a longitudinal panel' sentence) that do not appear in the provided Methods/Results/Limitations/Abstract; those spans were not groundable, so the affected critiques were re-anchored to verifiable abstract spans and the unverifiable non-significance assertion was removed. The measurement/IDK flaw was dropped as author-disclosed and dependent on ungroundable specifics (26% IDK, '21 additional significant coefficients'), leaving three defensible, span-exact flaws.
Reply from the authors
Following the practice of Nature Matters Arising, Science Technical Comments and PNAS Letters, this Comment is published as one half of a Comment + Reply pair: the authors of the original article are invited to respond, and any reply is published here verbatim alongside the Comment as part of the record.
Reply: not yet invited. No reply has been received for publication.
The authors have a right of reply and no veto. A reply may request a factual correction, a methodological rebuttal, a clarification, a data/code update, or a severity challenge, and is published unedited. See the right-of-reply policy.
Automated re-evaluation after reply: Authors may reply at any time; this critique addresses claims, methods and inference only, never the authors.
References
Every external source this Comment cites, each with a verified link. 0 fabricated.
Source-grounding attestation
- ✓Verbatim source spans present in the critique — 3/3 provenance spans re-derived in the critique prose
- ✓Passes the publication validator — no errors
- ✓Zero fabricated citations — 0 fabricated
- ✓Severity within the access-basis cap — severity "moderate" ≤ cap "high" for open_access
Every verbatim span the critique relies on is re-derived in the prose in-app; span-in-source is re-verifiable offline (the abstract is re-fetched, not stored, per the no-reproduce policy).
Re-verify span-in-source offline: python3 scripts/verify-queue-critiques.py
Independent faithfulness review
A refute-by-default adversarial panel (two independent reviewers — an overreach lens and a mischaracterization lens — that fetched the real source) tried to prove this critique misread the paper. This is an AI adversarial review recorded with its reasoning, not a deterministic check.
The SHARPENED critique holds against the provided text under a strict-neutral standard. All three kept flaws are anchored to spans that are verbatim present in the source, and I verified each: 1. statistical_inference/overclaim — Span "the latter experienced a notable shift towards majority support between 2022 and 2023" is verbatim in the Abstract. The supporting figures (Republicans 43%->68%) are verbatim in Results ("shifted from minority agreement in 2022 (43%) to majority agreement in 2023 (68%)"). Methods confirms "two independent waves" with 524 (2022) and 504 (2023) responses — different respondents, no panel. So a descriptive proportion difference across two cross-sections is correctly characterized as not a tested differential-change effect, and the abstract does foreground it as a substantive over-time finding. GROUNDED. 2. identification/causal — Span "the survey captures attitudinal changes within the six months following the public release of ChatGPT" is verbatim in the Abstract (full sentence: "...within the six months following the public release of ChatGPT and the
Version & correction history
| Version | Date | Change |
|---|---|---|
| v1.0 | 2026-06-28 | Initial publication (scale-the-full-text-cohort batch). |
No silent substantive corrections — every change is versioned and visible.
How to cite this Comment
Critical AI. Comment on “Local US officials' views on the impacts and governance of AI: Evidence from 2022 and 2023 survey waves” (Sophia Hatz et al., PLOS One, 2025). Critical AI; 2026. https://policywindow.org/critique/c/local-us-officials-views-ai-governance
A registered DOI will replace the URL once minted; until then the canonical URL is the persistent identifier. Highwire/Dublin-Core citation tags and a schema.org Review record are embedded in this page for Google Scholar and reference managers.
Verify this Comment. Its checkable facts (target DOI, access-basis severity cap, zero fabricated citations) are served — as the app’s self-report — at /critique/api/critiques/local-us-officials-views-ai-governance/verify; to confirm them independently of this site, re-derive the same checks (and resolve the target DOI) with npx tsx scripts/verify-critical-ai.ts --critique local-us-officials-views-ai-governance --live.
Content fingerprint 759a1e086ccbb244 (v1.0) — this Comment’s substantive content is content-addressed; a silent post-publication edit would change it.