pith. machine review for the scientific record. sign in

arxiv: 2604.26245 · v1 · submitted 2026-04-29 · 💻 cs.SI

Recognition: unknown

Institutional Floors and Partisan Lenses: Cross-National Online Discourse on Political Violence in France and the United States

Andrew Yen Chang

Authors on Pith no claims yet

Pith reviewed 2026-05-07 12:31 UTC · model grok-4.3

classification 💻 cs.SI
keywords political violenceonline discoursecivic floormoral evaluationFranceUnited Statessocial media analysispartisan polarization
0
0 comments X

The pith

France maintains a cross-partisan civic focus in online discussions of political violence, while the United States shows stronger partisan divisions.

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

This paper compares online social media discourse surrounding incidents of political violence in France and the United States to understand how moral evaluations differ by institutional context. Using classification of posts from Instagram and Facebook, it reveals that French discussions center on victims' civic roles irrespective of their political affiliations, supporting a civic floor that sets a shared baseline. American discourse, by contrast, tends to split along ideological and partisan lines in judging morality and assigning blame. The findings matter for understanding how national institutions might buffer or amplify polarization in digital public spheres. The work also points to new ways computational tools can analyze cross-national patterns in political perception.

Core claim

The analysis of discourse on the 2020 killing of Samuel Paty, the 2026 murder of Quentin Deranque in France, and the 2025 shooting of Charlie Kirk in the United States demonstrates that France's institutional framework upholds a cross-partisan civic baseline in moral evaluations, as shown by consistent focus on civic identity even for the politically-affiliated victim, unlike the ideologically divided patterns in the US.

What carries the argument

The civic floor hypothesis, tested through comparison of a civic victim and a politically-affiliated victim in France against a US case, using zero-shot GPT classification of moral evaluations and semantic network analysis.

If this is right

  • Discourse in France focuses on the victim's civic role rather than political affiliation.
  • US conversations reflect moral judgments aligned with partisan lines.
  • Cross-national differences appear in perceptions of moral values, emotional intensity, and institutional framing.
  • Semantic networks of discussion show distinct structures between the two countries.
  • The approach has implications for studying moral judgment in multilingual digital political discourse.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • This pattern may indicate that republican institutions in France create a stronger shared civic identity that resists partisan framing of violence.
  • Similar methods could test whether other countries with centralized civic education show comparable effects.
  • If the civic floor holds, it suggests policy efforts to emphasize shared national values could mitigate online polarization around political violence.

Load-bearing premise

That the zero-shot GPT-4o-mini classifications reliably and without bias detect moral evaluations, emotional intensity, and institutional framing in posts about political violence.

What would settle it

Observing equivalent levels of partisan moral division in French discourse about both Paty and Deranque, matching the US patterns, would falsify the civic floor hypothesis.

Figures

Figures reproduced from arXiv: 2604.26245 by Andrew Yen Chang.

Figure 1
Figure 1. Figure 1: Moral evaluation distributions across the three cases. Paty (FR) shows near-unanimous view at source ↗
Figure 2
Figure 2. Figure 2: Emotion distributions across all three cases. Both French cases (Paty and Deranque) are view at source ↗
Figure 3
Figure 3. Figure 3: Composite institutional framing (Civic + State) by case. Paty (FR) reaches 57.6%, Der view at source ↗
Figure 4
Figure 4. Figure 4: NRC full emotion profile by case using language-appropriate lexicons (French NRC for view at source ↗
read the original abstract

This paper studies how online discussion shapes and assesses political violence across different settings, particularly how moral evaluation, as a social perception, varies across institutional contexts. We take France and the United States as case studies, both democracies, and three incidents of political violence: the 2020 killing of Samuel Paty in France, the 2025 shooting of Charlie Kirk in the United States, and the 2026 murder of Quentin Deranque in France. Using publicly available posts on Instagram and Facebook, we use GPT-4o-mini for zero-shot classification and social network analysis. Our research demonstrates clear cross-national differences in how moral values are perceived, the emotional intensity expressed, the framing of institutions, and the structure of semantic networks. In France, the discourse tends to focus on the victim's civic role rather than their political affiliation, whilst in the U.S., the conversation is more ideologically divided, with moral judgments frequently reflecting partisan lines. By comparing the two French cases -- a civic victim (Paty) versus the politically-affiliated victim (Deranque) -- we find evidence consistent with the \textit{civic floor hypothesis}, which demonstrates France's institutional framework upholds a cross-partisan civic baseline regardless of the victim's political ties. We conclude by analyzing the implications of computational social perception for multilingual NLP and by exploring moral judgment in cross-national digital political discourse.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper analyzes online discourse on three incidents of political violence (Samuel Paty and Quentin Deranque in France; Charlie Kirk in the US) using zero-shot GPT-4o-mini classification of Instagram and Facebook posts to compare moral evaluations, emotional intensity, institutional framing, and semantic networks. It reports cross-national differences, with French discourse emphasizing victims' civic roles over partisan ties, and presents evidence from the two French cases as consistent with a 'civic floor hypothesis' that France's institutions maintain a cross-partisan civic baseline.

Significance. If the LLM-based classifications prove reliable, the work offers a computational approach to studying how institutional contexts shape moral perceptions in digital political discourse, with potential implications for multilingual NLP and cross-national comparisons of social media responses to violence. The use of real public post data and network analysis provides a concrete empirical basis for the interpretive claims.

major comments (2)
  1. [Methods] Methods section: The zero-shot GPT-4o-mini classification of moral evaluations, emotional intensity, institutional framing, and semantic structures lacks any reported validation (e.g., human annotation agreement, accuracy metrics on a held-out set, or bias checks for multilingual political content). This is load-bearing for the central claim, as the observed differences between the Paty and Deranque cases—and thus support for the civic floor hypothesis—rest entirely on these unvalidated labels.
  2. [Results/Discussion] Results and Discussion sections: The cross-partisan civic baseline in France is inferred from differences in LLM outputs between the two French cases, but no analysis addresses potential systematic biases in zero-shot prompting on short, context-poor, multilingual social media text (e.g., sarcasm, local idioms, or training-data skews toward English-centric moral framing).
minor comments (2)
  1. [Abstract] Abstract and introduction: The dates for the Kirk (2025) and Deranque (2026) incidents appear to be in the future relative to the current date; clarify if these are hypothetical or correct the timeline.
  2. [Introduction] The paper introduces the 'civic floor hypothesis' as an interpretive frame but does not specify how it was pre-registered or distinguished from post-hoc interpretation of the LLM results.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments, which highlight important areas for improvement in our methods and discussion of limitations. We provide point-by-point responses below and commit to revisions that address the concerns raised.

read point-by-point responses
  1. Referee: [Methods] Methods section: The zero-shot GPT-4o-mini classification of moral evaluations, emotional intensity, institutional framing, and semantic structures lacks any reported validation (e.g., human annotation agreement, accuracy metrics on a held-out set, or bias checks for multilingual political content). This is load-bearing for the central claim, as the observed differences between the Paty and Deranque cases—and thus support for the civic floor hypothesis—rest entirely on these unvalidated labels.

    Authors: We recognize the critical importance of validating the zero-shot classifications, as they underpin our comparative findings and the civic floor hypothesis. Although the original manuscript did not include such validation, we will perform a post-hoc validation by randomly sampling 300 posts across the three cases and having them independently annotated by two human coders fluent in the respective languages. We will report Cohen's kappa for inter-annotator agreement and accuracy relative to the LLM labels in a new subsection of the Methods. This addresses the load-bearing nature of the classifications. revision: yes

  2. Referee: [Results/Discussion] Results and Discussion sections: The cross-partisan civic baseline in France is inferred from differences in LLM outputs between the two French cases, but no analysis addresses potential systematic biases in zero-shot prompting on short, context-poor, multilingual social media text (e.g., sarcasm, local idioms, or training-data skews toward English-centric moral framing).

    Authors: We concur that systematic biases in zero-shot prompting for social media text, especially multilingual and abbreviated content, could influence the results. To mitigate this in the revision, we will include an analysis of potential biases by manually reviewing a sample of posts for sarcasm and idioms, and discuss how the French and English contexts might interact with the model's training data. Additionally, we will test the robustness by re-running classifications with varied prompts and report if the cross-national patterns hold. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical test of interpretive hypothesis on external data

full rationale

The paper conducts an empirical study by applying zero-shot GPT-4o-mini classification to public Instagram and Facebook posts about three specific incidents, then compares moral evaluations, emotional intensity, institutional framing, and semantic networks across cases and countries to assess consistency with the civic floor hypothesis. No equations, parameters fitted to subsets of the target data, self-citations, or ansatzes are invoked in a load-bearing manner. The hypothesis functions as an interpretive frame tested against independent observations rather than being defined by or derived from the classification outputs themselves. The derivation chain remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

The work depends on the reliability of LLM-based social perception classification and the assumption that the three selected incidents are comparable proxies for broader discourse patterns.

axioms (2)
  • domain assumption Zero-shot GPT-4o-mini classification produces valid labels for moral values, emotional intensity, institutional framing, and semantic networks in social media text across English and French.
    Invoked throughout the methods description as the core measurement tool without reported validation.
  • domain assumption The three chosen incidents and their associated public posts are representative of national-level discourse on political violence.
    Underlies the cross-national and within-France comparisons.
invented entities (1)
  • civic floor hypothesis no independent evidence
    purpose: Interpretive framework explaining why French discourse maintains civic focus independent of victim partisanship.
    Introduced to account for the observed pattern in the two French cases.

pith-pipeline@v0.9.0 · 5546 in / 1397 out tokens · 50040 ms · 2026-05-07T12:31:26.762337+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

25 extracted references

  1. [1]

    Almond and Sidney Verba.The Civic Culture: Political Attitudes and Democracy in Five Nations

    Gabriel A. Almond and Sidney Verba.The Civic Culture: Political Attitudes and Democracy in Five Nations. Princeton University Press, Princeton, NJ, 1963

  2. [2]

    Conservative influencer charlie kirk shot dead in ‘political assassination’.Reuters, September 2025

    Joseph Ax, Brian Brooks, and Andrea Shalal. Conservative influencer charlie kirk shot dead in ‘political assassination’.Reuters, September 2025

  3. [3]

    Charlie kirk shot dead at utah university event.BBC News, September 2025

    BBC News. Charlie kirk shot dead at utah university event.BBC News, September 2025

  4. [4]

    France’s ideals, and what it saw as an attack on them.The New York Times, October 2020

    Aurelien Breeden and Constant M ´eheut. France’s ideals, and what it saw as an attack on them.The New York Times, October 2020

  5. [5]

    A coefficient of agreement for nominal scales.Educational and Psychological Measurement, 20(1):37–46, 1960

    Jacob Cohen. A coefficient of agreement for nominal scales.Educational and Psychological Measurement, 20(1):37–46, 1960. 12

  6. [6]

    The indivisibility of the french republic as political theory and constitutional doctrine.European Constitutional Law Review, 11(3):458–481, 2015

    Eoin Daly. The indivisibility of the french republic as political theory and constitutional doctrine.European Constitutional Law Review, 11(3):458–481, 2015

  7. [7]

    Shades of american identity: Implicit relations between ethnic and national identities.Social and Personality Psychology Compass, 8(12):739–754, 2014

    Thierry Devos and Hafsa Mohamed. Shades of american identity: Implicit relations between ethnic and national identities.Social and Personality Psychology Compass, 8(12):739–754, 2014

  8. [8]

    Susan T. Fiske. Social cognition and social perception.Annual Review of Psychology, 44:155–194, 1993

  9. [9]

    Eight people go on trial in Paris over beheading of teacher Samuel Paty.France 24, November 2024

    France 24. Eight people go on trial in Paris over beheading of teacher Samuel Paty.France 24, November 2024

  10. [10]

    Accounting for change in diverse societies: France

    Global Centre for Pluralism. Accounting for change in diverse societies: France. Technical report, Global Centre for Pluralism, Ottawa, 2018

  11. [11]

    Democracy and the deferral of justice in France and the United States.Yale French Studies, 100:65–87, 2001

    Daniel Gordon. Democracy and the deferral of justice in France and the United States.Yale French Studies, 100:65–87, 2001

  12. [12]

    Wojcik, and Pe- ter H

    Jesse Graham, Jonathan Haidt, Sena Koleva, Matt Motyl, Ravi Iyer, Sean P. Wojcik, and Pe- ter H. Ditto. Moral foundations theory: The pragmatic validity of moral pluralism. In Patricia Devine and Ashby Plant, editors,Advances in Experimental Social Psychology, volume 47, pages 55–130. Academic Press, 2013

  13. [13]

    Assassination of Charlie Kirk

    Ted Grant. Assassination of Charlie Kirk. Encyclopædia Britannica, 2025

  14. [14]

    The public sphere: An encyclopedia article.New German Critique, 3:49– 55, 1974

    J ¨urgen Habermas. The public sphere: An encyclopedia article.New German Critique, 3:49– 55, 1974. Trans. S. Lennox and F. Lennox

  15. [15]

    Hopp, Jacob T

    Frederic R. Hopp, Jacob T. Fisher, Devin Cornell, Richard Huskey, and Ren ´e Weber. The extended moral foundations dictionary (eMFD): Development and applications of a crowd- sourced approach to extracting moral intuitions from text.Behavior Research Methods, 53:232–246, 2021

  16. [16]

    Hutto and Eric Gilbert

    Clayton J. Hutto and Eric Gilbert. V ADER: A parsimonious rule-based model for sentiment analysis of social media text. InProceedings of the Eighth International Conference on Weblogs and Social Media (ICWSM). AAAI Press, 2014

  17. [17]

    Oxford University Press, New York, 2008

    C ´ecile Laborde.Critical Republicanism: The Hijab Controversy and Political Philosophy. Oxford University Press, New York, 2008

  18. [18]

    Color-blind: France’s approach to race.Harvard International Review, Febru- ary 2021

    Anna LaBreck. Color-blind: France’s approach to race.Harvard International Review, Febru- ary 2021

  19. [19]

    Samuel Paty’s beheading: The limits of speech and religion.Religion Un- plugged, December 2024

    Paul Marshall. Samuel Paty’s beheading: The limits of speech and religion.Religion Un- plugged, December 2024

  20. [20]

    Content library and API

    Meta Platforms, Inc. Content library and API. Meta Platforms, Inc., 2024. 13

  21. [21]

    Mohammad and Peter D

    Saif M. Mohammad and Peter D. Turney. Crowdsourcing a word-emotion association lexi- con.Computational Intelligence, 29(3):436–465, 2013

  22. [22]

    Pietro S. Nivola. Why federalism matters. Technical Report 150, Brookings Institution, Washington, DC, 2005

  23. [23]

    GPT-4o mini

    OpenAI. GPT-4o mini. OpenAI, 2024

  24. [24]

    Minute de silence pour Quentin Deranque: Les d ´eput´es ont-ils compris `a qui ils rendaient aussi hommage?Ouest-France, 2026

    Ouest-France. Minute de silence pour Quentin Deranque: Les d ´eput´es ont-ils compris `a qui ils rendaient aussi hommage?Ouest-France, 2026

  25. [25]

    frame":

    Elisabeth Zoller. La¨ıcit´e in the United States or the separation of church and state in a pluralist society.Indiana Journal of Global Legal Studies, 13(2):561–594, 2006. A GPT-4o-mini Classification Prompt The following prompt was used for all zero-shot classifications across all three corpora. Tempera- ture was set to 0 for full reproducibility. System...