We are excited to welcome Gennadii Iakovlev on the 25th of February as a guest lecturer for our V-Shift Lendület project, presenting his groundbreaking research titled: Affective Elite Polarization in European Parliamentary Speeches: A Novel Measurement Approach Using Large Language Models.
The abstract of his research project is as follows.
This project introduces a novel measure of affective elite polarization using Large Language Models. We present polarization data aggregated at the party-quarter level over the past two decades for three countries: Hungary, the UK, and Italy. To perform a sentiment analysis on these corpora of parliamentary speeches, we are utilizing pre-trained generative machine learning software. In so-doing, we pinpoint elite attitudes toward various in-groups and out-groups and combine these attitudes into an index of affective polarization. This approach paves the way for creating a time-series, EU-wide dataset of affective polarization spanning the last twenty years. Our findings reveal that affective elite polarization is consistent with, and falls somewhat between, affective mass and ideological elite polarization. The fine-grained party-quarter analysis shows that affective elite polarization slightly increases during elections and country- or party-level crises but more than that significantly decreases afterward. Also, the parties’ polarization is more responsive to party-level than to country-level events —- and so is their aggregated country-level score. Additionally, we find that governing parties tend to exhibit significantly lower levels of polarization compared to those in opposition.
The lecture will take place at the HUN-REN Centre for Social Sciences, on February 25th, 2025 at 1:00 PM.