Slovak MPs' response to the 2022 Russian invasion of Ukraine in light of conspiracy theories and the polarization of political discourse

The paper examines how Slovak Members of Parliament (MPs) framed and contested the 2022 Russian invasion of Ukraine on Facebook during one week before and one week after 24 February 2022. Against a backdrop of broad European condemnation of the invasion, Slovakia stands out with a quarter of MPs not supporting a parliamentary condemnation resolution. Prior work shows that crises can depolarize politics in some contexts, but also that Russian propaganda narratives and conspiracy theories persisted across Europe during the invasion. Slovakia has high susceptibility to conspiratorial thinking, low trust in government, and exposure to Russian disinformation. The study applies Discourse Network Analysis (DNA) to explore: (RQ1) the content of MPs’ discourse pre- and post-invasion; (RQ2) the structure of the discourse network across the two weeks; and (RQ3) mechanisms driving changes in the discourse network at the invasion’s onset. Facebook is chosen due to its central role in Slovak political communication and immediacy during crises.

The paper situates Slovakia within East-Central Europe as a NATO/EU member with historically low satisfaction with democracy, high conspiratorial beliefs, and vulnerability to Russian disinformation and Pan-Slavist, anti-Western narratives. Prior studies found that in Italy and Finland the invasion reduced polarization around NATO as parties repositioned under external threat, while Russian propaganda narratives (e.g., blaming NATO, biolabs conspiracy) spread widely on social media across Europe. The concept of "conspiracy singularity" posits that actors who spread one conspiracy tend to adopt and connect multiple conspiracies into a unified narrative; emerging evidence from QAnon-related channels supports this. The paper adopts Discourse Network Analysis (DNA) to capture discursive struggles by linking actors and concepts, integrating qualitative coding with social network analysis to examine content, actor positions, dynamics, agenda setting, reinforcement, and polarization. DNA’s compatibility with inferential network models (e.g., SAOMs) enables testing micro-mechanisms of discourse change.

Design: Exploratory analysis using Discourse Network Analysis over two adjacent weekly windows: pre-invasion (17–23 Feb 2022) and post-invasion (24 Feb–2 Mar 2022). Sample and data collection: Of 150 Slovak MPs, 117 (78.0%) maintained active public Facebook pages. Researchers manually retrieved all public posts (text, photos, videos, combinations) from these pages at the end of the post-invasion week (2 Mar 2022). Total posts: 1,880 (859 pre-invasion; 1,021 post-invasion). Additional MP attributes collected: coalition/opposition and party affiliation, membership in 19 parliamentary committees, and number of Facebook followers (as of 2 Mar 2022). Coding: Two coders conducted inductive thematic analysis to generate initial codebooks (~100 codes each), then reconciled to a final codebook of 120 codes grouped into 11 categories: pro-Ukraine, pro-Russia, pro-NATO, anti-NATO, Slovak neutrality, anti-opposition, anti-coalition, pro-president, anti-president, pro-Chief Prosecutor, anti-Chief Prosecutor; plus 20 miscellaneous codes outside categories. Deductive coding applied one or more codes per post. Inter-rater reliability on 10% double-coded posts: Krippendorff’s Alpha = 0.88 (SD = 0.02). The final codebook (Supplementary Table S2) and adjacency matrices (Supplementary Table S3) are provided. Network construction: Two bipartite networks (MPs × discourse codes) for each week were built with undirected, weighted ties representing counts of code usage by each MP. For structural measures requiring unweighted ties, networks were dichotomized. Content analysis (RQ1): Popularity of codes and categories measured by weighted degree (frequency of usage). Structural analysis (RQ2): Visualization in Gephi using ForceAtlas2. Computed bipartite measures (migraph R package): Jaccard overlap between weeks, density, degree centralization (separately for MPs and codes), and equivalency (closed four-cycles over open three-paths) ranging 0–1. Mechanisms of change (RQ3): Applied stochastic actor-oriented models for bipartite networks (SAOM) in RSiena. MPs are the agency mode; simulations estimate objective function parameters (log-odds). Included endogenous effects: outdegree (density), basic rate, indegree popularity (sqrt) and outdegree activity (sqrt) for agenda reinforcement, outgoing isolate, and 4-cycles (bipartite closure). Attribute-based effects: MP activity (ego) by gender, party, number of committees, number of followers; MP homophily (alter in distance 2) by gender, party, followers; and code popularity (alter) by code categories. Estimation used two runs (second with increased simulations and multi-core) to achieve convergence; model fit assessed via outdegree/indegree distributions (goodness-of-fit in Supplementary Figure S1). Analytical code provided (Supplementary Code S1).

Content (RQ1): - Pre-invasion discourse was dominated by domestic antagonism (anti-coalition, anti-opposition). Post-invasion, invasion-related categories surged, especially pro-Ukraine, alongside notable increases in anti-NATO, Slovak neutrality, and pro-Russia codes. Structure (RQ2): - Network descriptors (pre → post): Jaccard coefficient = 0.13 (low overlap), density 0.05 → 0.06 (+0.02, +40.74%), MP degree centralization 0.42 → 0.26 (−0.16, −39.05%), code degree centralization 0.48 → 0.65 (+0.17, +36.21%), equivalency 0.25 → 0.34 (+0.09, +37.20%). Interpretation: more MPs discussing a smaller set of highly popular codes; greater closure/interconnectedness; substantial reconfiguration across weeks. - Visual structure showed a pronounced coalition–opposition split. Pre-invasion, coalition MPs emphasized pro-Ukraine, pro-NATO, anti–Chief Prosecutor, and anti-disinformation content; opposition MPs (notably SMER-SD, ĽSNS, Republika) emphasized pro-Russia, anti-NATO, anti-coalition narratives and opposed the US Defense Agreement. Post-invasion, coalition MPs concentrated on pro-Ukraine solidarity, while major opposition parties (excluding HLAS-SD) pushed Slovak neutrality, anti-NATO, and pro-Russia frames. HLAS-SD’s discourse moved closer to the coalition. - The same opposition MPs promoting Russian propaganda pre-invasion also spread COVID vaccine conspiracies and narratives of foreign control over Slovakia, then post-invasion amplified justifications for Russia, neutrality appeals, and whataboutism referencing NATO actions. Mechanisms of change (RQ3; SAOM): - Agenda reinforcement: Significant positive indegree popularity (0.46) and outdegree activity (0.24) indicated that popular topics became more popular and already-active MPs became more active; 4-cycles not significant after controls. - Agenda setters (MP activity, ego effects vs. independents): Significant increases for OĽaNO (0.32), SaS (0.52), SMER-SD (0.54), ĽSNS (1.73), and Republika (1.23). No significant activation for SME RODINA, ZA ĽUDÍ, HLAS-SD, ŽIVOT. Gender activity effect was marginally positive (0.29). - Polarization (homophily, alter in distance 2): Significant within-party homophily for OĽaNO (0.91), SaS (0.69), SMER-SD (0.93), HLAS-SD (0.96), ĽSNS (1.91), Republika (1.09), ŽIVOT (1.05), indicating rising party-based polarization. Gender homophily also significant (0.59). Followers-based homophily was non-significant (−0.34). - Agenda setting by code categories (alter effects): Invasion-related categories became significantly more popular: pro-Ukraine (1.08), Slovak neutrality (0.95), pro-NATO (0.78), anti-NATO (0.64), and pro-Russia (0.53). Anti-opposition also positive (0.39). Other categories showed null or weak effects. Overall: Coalition MPs largely adopted pro-Ukraine frames; opposition MPs, particularly SMER-SD, ĽSNS, and Republika, amplified pro-Russian, neutrality, and anti-NATO narratives. A subset of opposition MPs consistently propagated multiple conspiracies, supporting the "conspiracy singularity" pattern.

The findings demonstrate that the invasion rapidly reoriented Slovak parliamentary discourse from domestic coalition–opposition conflict toward invasion-related issues, yet without unified condemnation: the discourse bifurcated along coalition–opposition lines. Coalition MPs emphasized pro-Ukraine support, whereas major opposition parties promoted pro-Russia narratives, neutrality, and anti-NATO positions. Network measures and SAOM results jointly show that a small set of invasion-related topics dominated discussion (agenda reinforcement), and that party-based homophily increased (polarization). A stable cluster of opposition MPs (SMER-SD, ĽSNS, Republika) linked invasion narratives to pre-existing conspiracy and anti-Western frames, evidencing a tendency to merge multiple conspiracies into an overarching narrative (conspiracy singularity). These results contrast with depolarization observed in Italy and Finland, where parties repositioned under external threat; in Slovakia, only HLAS-SD shifted closer to the coalition, while prominent propagators of pro-Russian content intensified their stance. Given Slovak society’s susceptibility to elite cues, continued polarization and conspiratorial messaging among elites likely exacerbate societal fragmentation and vulnerability to foreign influence.

The study contributes by integrating qualitative coding with bipartite network analysis and SAOMs to capture rapid discourse shifts among Slovak MPs around the invasion of Ukraine. It shows: (1) a dramatic topic realignment toward invasion-related frames; (2) a persistent and strong coalition–opposition split; (3) agenda reinforcement concentrating attention on few highly salient topics; (4) party-based polarization intensifying; and (5) evidence that key opposition actors fused multiple conspiracies, supporting the conspiracy singularity thesis. Future research should: (a) examine longer time horizons to assess persistence or attenuation of polarization; (b) test whether politicians who adopt conspiratorial frames during one crisis consistently pivot to new conspiracies in subsequent crises; (c) investigate how extreme-end parties converge across issue domains; and (d) leverage automated text analysis to scale coding and assess cross-platform dynamics.

- Timeframe: Only two weeks (one pre-, one post-invasion) limits observation of longer-term mechanisms and comparisons to other crises (e.g., COVID, Kuciak’s murder). - Manual workload: Extensive manual coding constrained temporal scope; automated methods could enable broader coverage. - Single-time-point retrieval: Posts were collected at the end of the post-invasion week; deleted posts may be missing, potentially biasing discourse representation. - Platform scope: Focus on Facebook public pages; excludes private profiles and other platforms, which may host different dynamics.