GuiltRoBERTa: An AI-assisted tool to uncover the emotional dynamics of blame rhetoric in Hungarian Politics.

In Hungary's increasingly polarized political landscape, understanding the emotional mechanisms behind blame rhetoric is essential. This paper introduces GuiltRoBERTa, an AI-powered natural language processing tool designed to analyze emotions in political discourse, focusing on guilt-evoking language. Guilt-evoking language is a cornerstone of blame rhetoric, which assigns responsibility for negative outcomes. It goes beyond stating facts, aiming to elicit emotions like remorse and responsibility in the target (Turner & Rains, 2021). This form of rhetoric often highlights moral transgressions—violations of shared norms—emphasizing their wrongfulness and consequences (Tangney et al., 2006). Politicians use guilt strategically to influence behaviour, advance agendas, and foster group cohesion by blaming external entities, which reinforces in-group solidarity (Engels, 2010). Additionally, guilt-evoking language contributes to political polarization by idealizing the in-group while devaluing the out-group (Szabó, 2024). While advances in Large Language Models (LLMs) and natural language processing (NLP) have enhanced emotion analysis in political texts, these tools often prioritize English, overlooking low-resource languages (Sebők et al., 2024). To address this gap, we developed and validated a theory-driven XLM-RoBERTa model tailored to detect guilt-evoking language in Hungarian parliamentary speeches from 1998 to 2022. Guilt-evoking language operates through a combination of moral anger directed at a perceived transgression, attempts to induce feelings of guilt in the target, and intense blame emphasizing the negative consequences of the action, particularly the harm caused to others (on moral anger see Lindebaum & Geddes, 2015). While such rhetoric can sometimes motivate corrective behaviour, in the Hungarian context, it often serves to assert moral superiority rather than pursue pro-social goals for socio-political improvement. Our analysis drew from over 5 million sentences across political parties in the Hungarian National Assembly, revealing unique patterns of guilt-oriented rhetoric. The study examines how opposition and pro-government parties differ in their use of this rhetoric, factoring in variables like policy topics, issue ownership, and ideological positioning. By exploring the interplay of moral anger, blame, and guilt in political communication, this research offers a deeper understanding of emotional management in Hungarian politics and contributes to the broader literature on political emotions and polarization. References: Engels, J. (2010). The Politics of Resentment and the Tyranny of the Minority: Rethinking Victimage for Resentful Times. Rhetoric Society Quarterly, 40(4), 303–325. Lindebaum, D., & Geddes, D. (2015). The place and role of (moral) anger in organizational behavior studies. Journal of Organizational Behavior, 37(5), 738–757 Sebők, M., Máté, Á., Ring, O., Kovács, V., & Lehoczki, R. (2024). Leveraging Open Large Language Models for Multilingual Policy Topic Classification: The Babel Machine Approach. Social Science Computer Review, 0(0). Szabó, G. (2024). Managing moral emotions in divided politics: Lessons from Hungary’s 2022 general election campaigns. Springer Nature Tangney, J. P., Stuewig, J., & Mashek, D. J. (2007). Moral emotions and moral behavior. Annual Review of Psychology, 58(1), 345–372 Turner, M.M., & Rains, S.A. (2021). Guilt Appeals in Persuasive Communication: A Meta-Analytic Review. Communication Studies, 72, 684 - 700