The deepening hostility among political elites is a central concern for democratic stability, but treating this polarization as a generalized atmosphere obscures its underlying mechanics. In reality, affective and issue-based polarization are deeply intertwined, with political dyads deliberately escalating out-party animosity within the boundaries of specific policy debates. While existing comparative research captures broad legislative sentiment and elite ideological distance, it struggles to measure how this affective warfare is tactically distributed across the active policy agenda.
To address this gap, this study introduces a novel computational framework for mapping issue-specific affective polarization in the Australian House of Representatives (2010–2026). An agentic Large Language Model pipeline classifies parliamentary speeches into core policy domains, extracts directed out-party hostility from general legislative sentiment, and standardizes these evaluations across the majoritarian dyad. We then apply a non-parametric information-theoretic approach—capturing the Normalized Mutual Information (NMI) between policy topics and hostility—to aggregate these speech-level evaluations into a dynamic, longitudinal measure of how intensely parties weaponize the agenda over time.
Applying this framework reveals that what traditional metrics classify as generalized legislative toxicity is, in fact, highly concentrated issue weaponization. Preliminary results demonstrate that affective polarization does not rise uniformly across the parliament. Instead, partisan hostility clusters intensely around a narrow subset of policy domains—most notably border security and economic restructuring. By tracking this dyadic behavior over time, we uncover periods of asymmetrical escalation, illustrating how parties actively weaponize the legislative agenda to generate, rather than merely reflect, democratic polarization.
