Political Delegitimization Discourse Analysis

• Political Delegitimization Discourse (PDD) is a framework describing how symbolic rhetoric undermines legitimacy by targeting political actors, groups, or institutions.
• It employs systematic annotation and machine learning techniques, achieving high reliability (Cohen’s Kappa ≈0.82) and notable F1 scores in binary and multilabel detection.
• Longitudinal and cross-platform studies reveal fluctuating PDD trends tied to election cycles, gender differences, and political blocs, highlighting its impact on democratic discourse.

Political Delegitimization Discourse (PDD) is defined as symbolic attacks on the normative validity of political actors, groups, or institutions, aimed at undermining their legitimacy independently of specific policy debate. PDD operates through strategic, affective, and rhetorical mechanisms that exclude, ridicule, or frame targets as threats to the common good. Large-scale computational research has established annotation schemas and detection pipelines that identify key features of PDD, analyze trends over time and across platforms, and quantify actor-level and event-driven variation, demonstrating its rising prevalence and its differentiated usage patterns in modern democratic discourse (Rivlin-Angert et al., 21 Aug 2025).

1. Conceptual Foundations and Annotation Criteria

PDD is operationalized as language that questions or symbolically denies the legitimacy of a political actor, group, or institution. It is distinct from general incivility or hostility in that it is rhetorically and affectively focused on delegitimization rather than mere disagreement. Annotation frameworks specify PDD at the sentence level, identifying:

• Explicit delegitimization cues: ridicule, disgust, denial of political inclusion, vilification, or linkage to stigmatized outgroups.
• Target identification: specifying whether a person, group, or institution is the subject of delegitimization, with token-level marking to enable span-based analysis.
• Supplementary attributes: intensity (ordinal rating, e.g., 0=weak, 2=strong), incivility (mockery, insults, swearing), “outgroup” claims (framing adversaries as external), and appeals to the common good (portraying the target as a threat to society).

Annotation reliability in large-scale studies is high (correlation ≈ 0.91, Cohen’s Kappa = 0.82), suggesting stability and reproducibility across analysts (Rivlin-Angert et al., 21 Aug 2025).

2. Computational Methods for Detection and Characterization

PDD detection leverages both encoder and decoder architectures in a two-stage classification pipeline:

Stage 1 – Binary PDD Detection

• Fine-tuning transformer encoders (e.g., mBERT, AlephBERT, HeRO, DictaBERT) over up to 10 epochs and with varied loss functions.
• Incorporating decoder LLMs (DictaLM 2.0, Gemma-2B/9B, Qwen3-8B) to further classify discourse instances.
• Best model performance: DictaLM 2.0 achieves F$_1$ = 0.74 for binary PDD categorization.

Stage 2 – Characteristic Classification and Target Span Identification

• Multi-task classifiers assign binary attributes (incivility, outgroup, common good, etc.) and ordinal intensity (3-point scale).
• Decoder models use a separator token approach (e.g., marking mentions with “%%%”) for precise span identification of the delegitimization target; DictaLM 2.0 reaches macro-F$_1$ = 0.67 in multilabel and span detection.

All results are reported using standard LaTeX notation for F$_1$ and macro-F$_1$ scores, underlining methodological rigor (Rivlin-Angert et al., 21 Aug 2025).

3. Longitudinal and Cross-Platform Trends

Temporal analysis of Israeli political speech (Knesset debates from 1993–2023) and digital media (Facebook posts, major news outlets) reveals:

• Marked increase in PDD: sharp rise since 2008, with pronounced spikes during election campaigns and major institutional events.
• Cross-platform variation: PDD prevalence is similar in parliamentary and social media contexts (~7%), but social platforms show higher rates of all delegitimization features and greater average intensity (1.225 on Facebook vs. 1.000 in the Knesset).
• Gender disparity: male politicians use PDD more frequently than female politicians; mean scores are statistically significant (t = 2.613, p = 0.010).
• Political bloc variation: historical dominance by right-wing actors (mean PDD ≈ 0.074), but episodic surges among left-wing actors during specific campaigns.

These trends are indicative of an environment where symbolic exclusion and legitimization battles are central to electoral and public sphere competition (Rivlin-Angert et al., 21 Aug 2025).

4. Functional Attributes and Strategic Usage

PDD manifests in multiple functional forms:

• Incivility: linked with mockery, insult, and affective derision, but not all incivility qualifies as delegitimization—PDD specifically targets legitimacy rather than interpersonal relations.
• Intensity scaling: annotators distinguish weak, moderate, and strong delegitimization, revealing heterogeneous rhetorical strategies.
• Outgroup claims: delegitimizing actors often frame their targets as “external,” drawing boundaries of political community and leveraging societal fears.
• Common good framing: targets are depicted as threats to societal welfare; this framing escalates the perceived urgency and justifies exclusionary rhetoric.

The presence of these features is algorithmically tractable and can be tracked at scale, allowing for fine-grained analysis in comparative studies and real-time monitoring (Rivlin-Angert et al., 21 Aug 2025).

5. Democratic Implications and Policy Relevance

Findings from automated PDD analysis have implications for democratic health and institutional resilience:

• Polarization and erosion of norms: Spikes in PDD during election periods or institutional crises signal attempts by political actors to consolidate support by redefining boundaries of legitimate participation and discourse.
• Strategic reduction post-election: Coalitional actors reduce PDD after unity government formation, suggesting adaptive rhetorical moderation tied to institutional context.
• Early warning systems: Automated pipelines (F$_1$ ≥ 0.74) for monitoring delegitimization trends can inform policymakers about risks of escalating polarization and rhetorical exclusion.
• Research integration: PDD detection frameworks enable the paper of incivility, hate speech, and populist strategies, providing empirical tools to assess democratic dynamics.

Thus, PDD is both an indicator and a mediator of political change, offering leverage points for intervention and scholarly understanding (Rivlin-Angert et al., 21 Aug 2025).

6. Limitations and Directions for Research

Current studies are subject to several limitations:

• Linguistic specificity: Findings are drawn from Hebrew-language corpora focused on Israeli political speech and may not generalize across languages or polities without adaptation.
• Granularity: Sentence-level analysis may miss intersentential or dialogic delegitimization (e.g., irony, metaphor, cross-turn rhetorical moves).
• Descriptive orientation: Research to date tracks variance, not causality; links between PDD usage and audience attitudes or political outcomes remain to be explored.
• Expansion and deepening: Future research aims to incorporate more platforms, broader contextual analysis (thread-level, multi-modal), and causal inference, expanding the utility and generalizability of PDD frameworks.

This suggests a trajectory toward more nuanced, multilingual, and causal approaches to political delegitimization discourse in global democratic studies.

Summary Table: Key PDD Features in Large-Scale Computational Analysis

Attribute	Description	Quantification (DictaLM 2.0 F$_1$)
Binary PDD	Symbolic attack undermines legitimacy	0.74
Intensity	Ordinal scale: weak (0), strong (2)	0.67 (macro-F$_1$)
Incivility	Mockery, insults, swearing	0.67 (macro-F$_1$)
Outgroup Claim	Frames target as external enemy	0.67 (macro-F$_1$)
Common Good Appeal	Portrays target as threat to society	0.67 (macro-F$_1$)
Target Span	Token-level marking of delegitimized actor/group/institution	0.67 (macro-F$_1$)

These metrics illustrate the operationalization and performance of state-of-the-art systems for the automated paper of political delegitimization discourse.

Concluding Perspective

Political Delegitimization Discourse constitutes a distinct, measurable phenomenon in modern political communication. Through rigorous annotation, advanced modeling, and large-scale empirical paper, the field is positioned to quantify not only the prevalence but also the strategic dynamics and democratic implications of symbolic rhetorical exclusion. While initial findings are robust, broadening linguistic, contextual, and causal perspectives will be essential to fully understand the impact and evolution of PDD in global democratic societies (Rivlin-Angert et al., 21 Aug 2025).