pith. sign in

arxiv: 2510.01757 · v3 · submitted 2025-10-02 · 💻 cs.CY · cs.SI· physics.soc-ph

Framing Unionization on Facebook: Communication around Representation Elections in the United States

Arianna Pera , Veronica Jude , Ceren Budak , Luca Maria Aiello This is my paper

Pith reviewed 2026-05-18 11:01 UTC · model grok-4.3

classification 💻 cs.CY cs.SIphysics.soc-ph
keywords labor unionsfacebookdiscourse framesrepresentation electionssocial mediaNLRBframing analysisunion organizing
0
0 comments X

The pith

Unions using more diagnostic, prognostic, and community frames on Facebook before elections had higher odds of winning representation votes.

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

The study pairs official National Labor Relations Board election records with 158,000 Facebook posts written by U.S. labor unions from 2015 to 2024. A fine-tuned RoBERTa model labeled each post according to five standard discourse frames drawn from labor and social-movement research: diagnostic frames that name problems, prognostic frames that offer solutions, motivational frames that urge action, community frames that stress solidarity, and engagement frames that prompt online interaction. Posts showed heavy overall use of diagnostic and community frames, yet higher shares of diagnostic, prognostic, and community language before an election correlated with greater likelihood of a union victory. After elections the pattern reversed depending on the result, with winning unions cutting back on prognostic and motivational language and losing unions increasing prognostic and engagement language.

Core claim

Greater use of diagnostic, prognostic, and community frames prior to an election was associated with higher odds of a successful outcome. After elections, framing patterns diverged depending on results: after wins, the use of prognostic and motivational frames decreased, whereas after losses, the use of prognostic and engagement frames increased. Diagnostic and community frames dominated union communication overall, but frame usage varied substantially across organizations.

What carries the argument

The five discourse frames (diagnostic, prognostic, motivational, community, engagement) labeled by a fine-tuned RoBERTa classifier applied to Facebook posts, then compared with NLRB election timing and results.

If this is right

  • Unions can raise their chances in representation elections by increasing diagnostic, prognostic, and community framing in posts issued before the vote.
  • After a win, unions typically reduce their use of prognostic and motivational frames in later posts.
  • After a loss, unions typically increase their use of prognostic and engagement frames in later posts.
  • Frame choices differ enough across unions that organizations could adopt more deliberate messaging strategies tailored to their own patterns.
  • Linking social-media text to official election data can surface measurable correlations between online discourse and concrete organizing results.

Where Pith is reading between the lines

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

  • The same frame-analysis method could be applied to other collective-action settings such as contract campaigns or strike calls to test whether the same pre-event patterns predict success.
  • Repeating the study on platforms other than Facebook would show whether the observed frame-outcome links are specific to one site or hold more broadly.
  • Unions could experiment with training or templates that encourage diagnostic and community language in the months before scheduled elections and measure resulting changes in win rates.
  • The post-election divergence suggests unions treat wins and losses as signals that prompt different adjustments in future messaging, a dynamic that could be tracked in real time.

Load-bearing premise

The fine-tuned RoBERTa classifier accurately and consistently annotates posts with the five theoretically defined discourse frames without substantial misclassification that would distort the reported associations with election outcomes.

What would settle it

A hand-coded sample of several hundred posts shows large disagreement with the model's frame labels, or the statistical links between frames and election outcomes vanish when the analysis is rerun on only high-confidence model predictions.

Figures

Figures reproduced from arXiv: 2510.01757 by Arianna Pera, Ceren Budak, Luca Maria Aiello, Veronica Jude.

Figure 2
Figure 2. Figure 2: Pearson’s correlation between frames and emo￾tions in Facebook posts. Positive values (red) indicate pos￾itive correlations; negative values (blue) indicate negative correlations. (UNITE HERE), all located toward the bottom of the den￾drogram (union acronyms are expanded in Table E1). By contrast, lower-than-baseline usage of diagnostic and prog￾nostic frames was more characteristic of craft unions. This d… view at source ↗
Figure 3
Figure 3. Figure 3: Posterior mean log-odds estimates for selected variables, averaged across ten random seeds. Black dots in￾dicate posterior means; grey lines show 90% Highest Den￾sity Intervals (HDIs) for each seed. Ratios denote the num￾ber of seeds (out of ten) with HDIs excluding zero. Variables shown meet this criterion in at least six seeds. baseline odds of winning. Posterior means for Colorado, Idaho, and Ohio are s… view at source ↗
Figure 4
Figure 4. Figure 4: Changes in frame usage before and after elections, [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
read the original abstract

Digital media have become central to how labor unions communicate, organize, and sustain collective action. Yet little is known about how unions' online discourse relates to concrete outcomes such as representation elections. This study addresses the gap by combining National Labor Relations Board (NLRB) election data with 158k Facebook posts published by U.S. labor unions between 2015 and 2024. We focused on five discourse frames widely recognized in labor and social movement communication research: diagnostic (identifying problems), prognostic (proposing solutions), motivational (mobilizing action), community (emphasizing solidarity), and engagement (promoting social media interaction). Using a fine-tuned RoBERTa classifier, we systematically annotated unions' posts and analyzed patterns of frame usage around election events. Our findings showed that diagnostic and community frames dominated union communication overall, but that frame usage varied substantially across organizations. Greater use of diagnostic, prognostic, and community frames prior to an election was associated with higher odds of a successful outcome. After elections, framing patterns diverged depending on results: after wins, the use of prognostic and motivational frames decreased, whereas after losses, the use of prognostic and engagement frames increased. By examining variation in message-level framing, the study highlights how communication strategies correlate with organizational success, contributing open tools and data, and complementing prior research in understanding digital communication of unions and social movements.

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 / 1 minor

Summary. The paper claims to analyze 158k Facebook posts by U.S. labor unions from 2015 to 2024, using a fine-tuned RoBERTa model to classify them into diagnostic, prognostic, motivational, community, and engagement frames. It reports that greater pre-election use of diagnostic, prognostic, and community frames is associated with higher odds of successful NLRB representation elections. Post-election, frame usage diverges: prognostic and motivational frames decrease after wins, while prognostic and engagement frames increase after losses. The study integrates NLRB data and contributes open tools and data to the field.

Significance. If the classifier is shown to be reliable and the associations hold after appropriate controls, this work would be significant for bridging social movement framing theory with empirical digital trace data in labor organizing. It provides a large-scale, longitudinal view of communication strategies tied to concrete outcomes, and the open data/tools aspect supports cumulative science in computational social science.

major comments (2)
  1. [Abstract] The abstract describes the use of a fine-tuned RoBERTa classifier for annotating posts with discourse frames but omits all details on model validation, including training set size, human annotation agreement (e.g., Cohen's kappa), held-out test performance (precision, recall, F1 per frame), or error analysis. This is load-bearing for the central claims, as the statistical associations with election outcomes are based on these frame counts; systematic misclassification could produce spurious correlations.
  2. [Analysis] The reported associations between frame usage and election outcomes do not mention controls for confounding factors such as union size, industry, or regional differences, nor any robustness checks. These omissions weaken the ability to attribute the differences in odds to framing strategies specifically.
minor comments (1)
  1. The abstract could more explicitly state the number of unions or elections covered to provide context for the 158k posts.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments, which identify key areas where greater transparency and robustness will strengthen the manuscript. We address each major comment below and have incorporated revisions to improve the presentation of our methods and analyses.

read point-by-point responses

  1. Referee: [Abstract] The abstract describes the use of a fine-tuned RoBERTa classifier for annotating posts with discourse frames but omits all details on model validation, including training set size, human annotation agreement (e.g., Cohen's kappa), held-out test performance (precision, recall, F1 per frame), or error analysis. This is load-bearing for the central claims, as the statistical associations with election outcomes are based on these frame counts; systematic misclassification could produce spurious correlations.

    Authors: We agree that the abstract should include key details on classifier validation to support the reliability of the frame counts used in our outcome analyses. The full manuscript reports these elements in the Methods section, including training set size, inter-annotator agreement, held-out test performance, and error analysis. To address the comment, we have revised the abstract to add a concise summary of the validation results (e.g., training data, agreement metric, and average F1). This change improves transparency without altering the core claims. revision: yes

  2. Referee: [Analysis] The reported associations between frame usage and election outcomes do not mention controls for confounding factors such as union size, industry, or regional differences, nor any robustness checks. These omissions weaken the ability to attribute the differences in odds to framing strategies specifically.

    Authors: We acknowledge that the primary reported models do not explicitly include or discuss controls for union size, industry, or regional differences, nor do they present formal robustness checks. This limits the strength of causal attribution to framing. We agree this is a substantive gap. In the revised manuscript we will add regression specifications that incorporate these controls (using NLRB-derived measures of union size, industry fixed effects, and regional indicators) and include a set of robustness checks (alternative specifications and subsample analyses). Results will be reported in the main text or a dedicated appendix table. revision: yes

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper combines external NLRB election outcome records with public Facebook posts from unions and applies a fine-tuned RoBERTa classifier to annotate five discourse frames. Frame counts are then used in statistical associations with the independently sourced win/loss outcomes. No equations, self-definitional constructs, fitted parameters renamed as predictions, or load-bearing self-citations appear in the methodology or abstract. The central claims rest on empirical patterns derived from verifiable external data and standard classification techniques rather than reducing to the inputs by construction, making the analysis self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the domain assumption that the chosen five frames are valid and distinguishable categories for union discourse and that the classifier faithfully recovers them from text; no free parameters or invented entities are described in the abstract.

axioms (1)
  • domain assumption The five discourse frames (diagnostic, prognostic, motivational, community, engagement) are valid and distinct categories for analyzing union communication on social media.
    The study selects and trains the classifier on these frames as widely recognized in labor and social movement research.

pith-pipeline@v0.9.0 · 5790 in / 1337 out tokens · 77079 ms · 2026-05-18T11:01:28.006721+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

12 extracted references · 12 canonical work pages · 3 internal anchors

  1. [1]

    Bennett, W

    Framing processes and social movements: An overview and assessment.Annual review of sociology, 26(2000): 611–639. Bennett, W. L.; and Segerberg, A

    work page 2000

  2. [2]

    Bronfenbrenner, K

    How Digitized Strategy Impacts Movement Outcomes: Social Media, Mobilizing, and Organizing in the 2018 Teachers’ Strikes.Politics & Society, 50(3): 485–518. Bronfenbrenner, K

    work page 2018

  3. [3]

    ILR Review, 62(1): 3–21

    The eyes of the needles: A sequential model of union organizing drives, 1999–2004. ILR Review, 62(1): 3–21. Ferguson, J.-P

    work page 1999

  4. [4]

    Administrative Science Quarterly, 63(2): 441–477

    Os- motic mobilization and union support during the long protest wave, 1960–1995. Administrative Science Quarterly, 63(2): 441–477. Fern´andez-Zubieta, A.; Guevara, J. A.; Caballero Roldan, R.; and Robles, J. M

    work page 1960

  5. [5]

    BERTopic: Neural topic modeling with a class-based TF-IDF procedure

    BERTopic: Neural topic model- ing with a class-based TF-IDF procedure. arXiv preprint arXiv:2203.05794. Herrigel, G

    work page internal anchor Pith review Pith/arXiv arXiv

  6. [6]

    Media and Communication, 4(4): 8–12

    Social movements and political agency in the digital age: A communication approach. Media and Communication, 4(4): 8–12. Kelly, J. 1998.Rethinking industrial relations: Mobilization, collectivism and long waves. Routledge. Kinder, D. R

    work page 1998

  7. [7]

    RoBERTa: A Robustly Optimized BERT Pretraining Approach

    Roberta: A robustly optimized bert pretraining ap- proach. arXiv preprint arXiv:1907.11692. Lovejoy, K.; and Saxton, G. D

    work page internal anchor Pith review Pith/arXiv arXiv 1907

  8. [8]

    https://doi.org/10.48680/meta.metacontentlibrary.5.0

    Meta Content Library version v5.0. https://doi.org/10.48680/meta.metacontentlibrary.5.0. Online; accessed 3 March

    work page doi:10.48680/meta.metacontentlibrary.5.0

  9. [9]

    arXiv preprint arXiv:2404.07340

    RIP Twitter API: A eulogy to its vast research contributions. arXiv preprint arXiv:2404.07340. National Labor Relations Board

    work page arXiv

  10. [10]

    New Technol- ogy, Work and Employment , 36(2): 201–218

    Digital audiences of union organising: A social media analysis. New Technol- ogy, Work and Employment , 36(2): 201–218. eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1111/ntwe.12184. Pasquier, V .; Daudigeos, T.; and Barros, M

    work page doi:10.1111/ntwe.12184

  11. [11]

    DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter

    Dis- tilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter. arXiv preprint arXiv:1910.01108. Schaller, Z

    work page internal anchor Pith review Pith/arXiv arXiv 1910

  12. [12]

    Ameri- can Sociological Review, 75(2): 227–251

    Rival unionism and membership growth in the United States, 1900 to 2005: A special case of inter-organizational competition. Ameri- can Sociological Review, 75(2): 227–251. Tarrow, S

    work page 1900