Blaming Across the Aisle: Political Contrasting and Blame Attribution in the Danish Parliament

Markus Lundsfryd Jensen; Rune Egeskov Trust; Kenneth Christian Enevoldsen; Sara Kolding

cs.CLSep 22, 2026

Blaming Across the Aisle: Political Contrasting and Blame Attribution in the Danish Parliament

Authors: Markus Lundsfryd Jensen, Rune Egeskov Trust, Kenneth Christian Enevoldsen, Sara Kolding

Organizations: School of Communication and Culture Aarhus University · Center for Humanities Computing

Abstract

Political discourse is widely perceived to be growing more hostile, yet robust evidence remains scarce. This study examines blame attribution in the Danish Parliament from 1997 to 2026, combining a purpose-built classifier, BlameBERT (F1: 0.80), with multilevel statistical modeling. The classifier is constructed using an annotation-efficient pipeline for blame attribution in low-to-mid resource languages. The results reveal a banana-shaped trajectory, with blame declining until around 2016 before entering a significant and sustained increase in recent years (2019-2026). Government status consistently influenced blame attribution - an effect we term political contrasting - with opposition parties blaming substantially more than governing parties. This effect was moderated by ideology: The blame-dampening effect of governing was less pronounced among right-wing parties, and ideological extremity amplified blame more strongly on the right. In recent years, the interaction between political wing and ideological extremity intensified, suggesting an ideological hardening of the blame rhetoric concentrated on the right of the political spectrum. Taken together, these patterns suggest that the perceived rise in harsh political language reflects not merely a general rhetorical drift, but an ideologically asymmetric hardening of political discourse. A sensitivity analysis showed that the conclusions were robust to varying classification thresholds.

Figures & tables

Figure 1: Flowchart of creation of datasets and models (more detailed version in Appendix A.0.1 ).

	Class 1 (Blame)			Average
	Precision	Recall	F1	Macro F1
Embedding	0.52	0.87	0.65	0.67
Generative	1.00	0.42	0.60	0.75
BlameBERT	0.72	0.79	0.75	0.80

Table 1: Classification performance for Qwen embedding and -generative models against BlameBERT.

Party		Wingness
Left wing		(-)
EL	Enhedslisten	1.74
Å	Alternativet	1.39
SF	Socialistisk Folkeparti	1.15
S	Socialdemokratiet	0.60
RV	Radikale Venstre	0.15

Table 2: Overview of parties and their wingness score, derived from the Chapel Hill Survey, grouped by wing.

Figure 2: Estimated effect of time on blame for both time periods with 95% CI. Left : Analysis H1, Time post October 1997, Right : Analysis H1.1, Time post January 2019

Figure 3: Estimated effect of Government status on the two political wings for both time periods with 95% CI. Left : Analysis H2, Right : Analysis H2.1

Figure 4: Estimated effect of wing affiliation and wingness for both time periods with 95% CI. Left : Analysis H2, Right : Analysis H2.1

Appendix figures & tables19 assets

Supplementary material from the paper’s appendix.

Appendix

Figure 5: Detailed flow chart of data processing steps.

Figure 6: Monthly counts of sentences toughout the entire period. Time of dataset merge is shown by red dashed line.

Figure 7: Mean monthly counts of sentences per active day in parliament. Time of dataset merge is shown by the dashed red line.

Figure 8: Density plot of number of monthly sentences in the dataset split by merge date

Figure 9: Boxplot of number of monthly sentences in the dataset split by merge date.

Text	Translation	Label	Confidence
Man indrømmede så her i salen, at det nok var for dårligt, at man havde skåret ned på det, men så løb man ind i en anden udemokratisk ting, nemlig at når man bag lukkede døre først har vedtaget sådan en aftale, så kan den kun laves om, hvis samtlige partier, der er med i aftalen, ønsker at lave den om.	It was then admitted in this Chamber that it was probably too bad to have cut it down, but then you ran into another undemocratic thing, namely that when you have first adopted such an agreement behind closed doors, it can only be changed if all the parties involved in the agreement want to change it.	Blame	0.60
Jeg må indrømme, at jeg finder, at en kønsopdelt fagbevægelse er med til at fastholde og fortsætte et kønsopdelt arbejdsmarked og dermed er en selvstændig årsag til kønsbetingede lønforskelle.	I must admit that I believe that a gender-separated trade union movement is helping to maintain and continue a gender-based labour market and is thus an independent cause of gender pay differences.	Blame	0.58
Når FN’s Sikkerhedsråd har besluttet, at Saddam Hussein skal afvæbnes, at Saddam Hussein skal destruere sine kemiske og biologiske våben, at han ikke må udvikle atomvåben, at han ikke må samarbejde med terrorister, at han skal skrotte sine langtrækkende missiler, og han alligevel ikke gør det, må der være nogle, der siger: På et eller andet tidspunkt skal der altså magt bag ordene, hvis der skal stå respekt om FN’s Sikkerhedsråd.	When the United Nations Security Council has decided that Saddam Hussein must be disarmed, that Saddam Hussein must destroy his chemical and biological weapons, that he must not develop nuclear weapons, that he must not cooperate with terrorists, that he must scrap his long-range missiles, and yet he does not, there must be some who say: at some point, then, if there is to be respect for the UN Security Council, there must be power behind the words.	Blame	0.59
I dag har man 20.000 arbejdspladser placeret rundtomkring i verden.	Today there are 20,000 jobs located around the world.	No blame	0.89
Vi er glade for, at et enigt Lagting bakker op om den her ændring, og Liberal Alliance støtter det selvfølgelig også her fra Folketinget med vores ja til lovforslaget.	We are pleased that a united Laying supports this change, and of course the Liberal Alliance also supports it here from the Folketing with our ’yes’ to the bill.	No blame	0.95
Og det er også fuldstændig rigtigt, at der i sandhed er en langsigtet strategi med regeringens boligpolitik.	And it is also absolutely true that there is indeed a long-term strategy with the government’s housing policy.	No blame	0.96

Appendix

Table 3: Randomly sampled examples of sentences classified by BlameBERT.

	Precision	Recall	F1	Macro F1
DIAL-1	0.62	0.88	0.73	0.77
DIAL-2	0.67	0.85	0.75	0.79
DIAL-3	0.68	0.69	0.68	0.76
DIAL-4	0.70	0.80	0.75	0.80
DIAL-5	0.72	0.79	0.75	0.80

Appendix

Table 4: Test set performance across datasets. All metrics for blame class.

	Precision	Recall	Macro F1
Embedding	0.70	0.72	0.67
Generative	0.91	0.71	0.75
BlameBERT	0.80	0.81	0.80

Appendix

Table 5: Qwen 3-Embeddings, Generative Qwen 3.5, and BlameBERT compared on macro averaged metrics.

	Precision	Recall	Macro F1
Abstain	0.91	0.71	0.75
Default no blame	0.86	0.64	0.64
Default blame	0.82	0.77	0.79

Appendix

Table 6: Generative Qwen 3.5 macro-averaged zero-shot sensitivity to null-handling strategy.

Condition		Precision	Recall	F1-score	Support
Default no blame	No blame	0.72	1.00	0.84	278
	Blame	1.00	0.28	0.44	148
	Accuracy			0.75	426
	Macro avg	0.86	0.64	0.64	426
	Weighted avg	0.82	0.75	0.70	426
Default blame	No blame	0.82	0.93	0.87	278

Appendix

Table 7: Qwen 3.5:9B zero-shot classification performance under different null-handling strategies.

Figure 10: Confusion matrix per party in gold-labeled test-set. (Only parties with more than 10 sentences included.)

M1.2, Time and Blame (1997–2026)
(Intercept)	$-2.380\,691\,216\,3$ ∗∗∗
	( $0.064\,458\,956\,1$ )
scale(month_index)	$-0.009\,802\,540\,0$
	( $0.005\,997\,880\,9$ )
scale(month_index) 2	$0.013\,519\,743\,7$ ∗
	( $0.006\,402\,173\,0$ )

Appendix

Table 8: Full summary with effect and SE (b (SE)) of all included paramteres of model M1.2 of Analysis 2.3 . Here time (scaled month index) is treated as the focal predictor, while government status is included as a controlling fixed-effect variable.

M2.5: Political Predictors (1997–2026)
(Intercept)	$-2.516\,227\,104\,3$ ∗∗∗
	( $0.102\,953\,298\,6$ )
in_gov1	$-0.478\,440\,756\,8$ ∗∗∗
	( $0.020\,497\,203\,3$ )
blocright_bloc	$-0.315\,517\,625\,7$
	( $0.168\,920\,505\,4$ )

Appendix

Table 9: Full summary with effect and SE (b (SE)) of all included parameters of model M2.5 of Analysis 2.3 . Here, government status (in_gov), wing affiliation (bloc), political extremity ( wingness ), and the interaction between these are treated as focal predictors. The quadratic relationship between blame and time for this period (see Table 8 ) was included as a controlling fixed-effect variable.

M3.1: Time and Blame (2019–2026)
(Intercept)	$-2.294\,317\,734\,2$ ∗∗∗
	( $0.068\,592\,382\,2$ )
in_gov1	$-0.533\,586\,659\,8$ ∗∗∗
	( $0.045\,888\,005\,2$ )
scale(month_index)	$0.092\,077\,966\,6$ ∗∗∗
	( $0.008\,613\,680\,2$ )

Appendix

Table 10: Full summary with effect and SE (b (SE)) of all included paramteres of model M3.1 of Analysis 2.3 . Here time (scaled month index) is treated as the focal predictor, while government status is included as a controlling fixed-effect variable.

M4.4, Political predictors (2019–2026)
(Intercept)	$-2.413\,112\,607\,7$ ∗∗∗
	( $0.101\,276\,300\,0$ )
in_gov1	$-0.515\,356\,191\,9$ ∗∗∗
	( $0.044\,216\,368\,6$ )
blocright_bloc	$-0.521\,797\,342\,9$ ∗∗∗
	( $0.151\,235\,974\,0$ )

Appendix

Table 11: Full summary with effect and SE (b (SE)) of all parameters included in model M4.4 of Analysis 2.3 . Here government status (in_gov), wing affiliation (bloc), political extremity ( wingness ) and the interaction between these are treated as focal predictors. The linear relationship between blame and time which was found for this period (see Table 10 ) was included as controlling fixed-effects variables.

Analysis H1: Temporal trend (1997–2026), M1.2
Threshold	$\tau=0.50$		$\tau=0.625$		$\tau=0.75$		$\tau=0.875$
Predictor	$b$	$(SE)$	$b$	$(SE)$	$b$	$(SE)$	$b$	$(SE)$
Month (linear)	$-0.00980$	$(0.00600)$	$-0.00995$	$(0.00720)$	$-0.0121$	$(0.00888)$	$-0.00113$	$(0.0134)$
Month 2 (quadratic)	$0.0135^{*}$	$(0.00640)$	$0.0187^{*}$	$(0.00767)$	$0.0237^{*}$	$(0.00945)$	$0.0404^{**}$	$(0.0143)$
Analysis H2: Political predictors (1997–2026), M2.5
Government Status	$-0.478^{***}$	$(0.0205)$	$-0.497^{***}$	$(0.0247)$	$-0.518^{***}$	$(0.0308)$	$-0.491^{***}$	$(0.0473)$

Appendix

Table 12: Sensitivity analysis: fixed-effect estimates across classification thresholds. Standard errors in parentheses. Significance:

{}^{.}p<.10

{}^{*}p<.05

{}^{**}p<.01

{}^{***}p<.001

Figure 11: Density distribution of probabilities of BlameBERT in the inference data.

Figure 12: Proportion of sentences labeled as blame, by political topic domain.

Figure 13: Topic domain distribution for parties in government and in opposition.

Explore similar work

Aug 5, 2026cs.CL

German parties shifted towards intuition-based rhetoric after the far right's parliamentary breakthrough

The spread of misinformation is widely perceived as a threat to democratic deliberation, yet how political elites' rhetorical commitments to truth shift alongside the rise of populist actors remains poorly understood. Analysing 4.5 million tweets and 59,170 parliamentary speeches by German political elites between 2015 and 2025, we measure evidence-based and intuition-based rhetoric using a validated distributed dictionary representation. Across both arenas, intuition-based language has become more prominent, and right-leaning actors consistently exhibit the lowest Evidence Minus Intuition (EMI) scores. The parliamentary entry of the extreme-right Alternative for Germany (AfD) in 2017 coincides with sharp downward shifts in EMI across the broader chamber, while a more gradual decline is observed on Twitter. These findings document an association between far-right visibility and a changing approach to truth in elite discourse in a multiparty European democracy.

Peer Saleth, Segun T. Aroyehun, Fabio Carrella +3

Department of Politics and Public Administration, University of Konstanz, Universitätsstraße 10, Konstanz, 78457, Baden-Württemberg, Germany. · Artificial Intelligence Lab (Recod.ai), Institute of Computing, University of Campinas, R. Saturnino de Brito, 573, 13083-852,2026 Campinas, S˜ao Paulo, Brazil. · Department of Psychology, University of Potsdam, Karl-Liebknecht-Str. 24–25, 14476, Potsdam, Brandenburg, Germany. +1

Parliamentary Debates Democracy German Misinformation Rhetoric

Ask PDF

Apr 23, 2026cs.CL

Mapping the Political Discourse in the Brazilian Chamber of Deputies: A Multi-Faceted Computational Approach

Analyses of legislative behavior often rely on voting records, overlooking the rich semantic and rhetorical content of political speech. In this paper, we ask three complementary questions about parliamentary discourse: how things are said, what is being said, and who is speaking in discursively similar ways. To answer these questions, we introduce a scalable and generalizable computational framework that combines diachronic stylometric analysis, contextual topic modeling, and semantic clustering of deputies' speeches. We apply this framework to a large-scale case study of the Brazilian Chamber of Deputies, using a corpus of over 450,000 speeches from 2003 to 2025. Our results show a long-term stylistic shift toward shorter and more direct speeches, a legislative agenda that reorients sharply in response to national crises, and a granular map of discursive alignments in which regional and gender identities often prove more salient than formal party affiliation. More broadly, this work offers a robust methodology for analyzing parliamentary discourse as a multidimensional phenomenon that complements traditional vote-based approaches.

Flávio Soriano, Victoria F. Mello, Pedro B. Rigueira +4

Computer Science Department, Universidade Federal de Minas Gerais, Belo Horizonte, Minas Gerais, Brazil

Parliamentary Debates Political Discourse Discourse Stylometric Topic Modeling Portuguese+1

Ask PDF

Apr 21, 2026cs.CL

Epistemic orientation in parliamentary discourse is associated with deliberative democracy

The pursuit of truth is central to democratic deliberation and governance, yet political discourse reflects varying epistemic orientations, ranging from evidence-based reasoning grounded in verifiable information to intuition-based reasoning rooted in beliefs and subjective interpretation. We introduce a scalable approach to measure epistemic orientation using the Evidence--Minus--Intuition (EMI) score, derived from large language model (LLM) ratings and embedding-based semantic similarity. Applying this approach to 15 million parliamentary speech segments spanning 1946 to 2025 across seven countries, we examine temporal patterns in discourse and its association with deliberative democracy and governance. We find that EMI is positively associated with deliberative democracy within countries over time, with consistent relationships in both contemporaneous and lagged analyses. EMI is also positively associated with the transparency and predictable implementation of laws as a dimension of governance. These findings suggest that the epistemic nature of political discourse is crucial for both the quality of democracy and governance.

Segun Aroyehun, Stephan Lewandowsky, David Garcia

Department of Politics and Public Administration, University of Konstanz, Germany. · School of Psychological Science, University of Bristol, UK. · Department of Psychology, University of Potsdam, Germany. +1

Parliamentary Debates Epistemic States Democracy Political Discourse Deliberation Discourse+1

Ask PDF