Research Project: Horizon Europe Excellent Science ERC POC Grant şablonu
Loading...
Contributors
Funders
ID
EC.00168
Authors
Yörük, Erdem
Faculty Member
Publications
A computational analysis of ideological positions, emotional stance, and support for presidential candidates in Turkey
(Wiley, 2024) Atsızelti, Şükrü; Duruşan, Fırat; Gürerk, Oğuz; Hürriyetoğlu, Ali; Topçu, Işık Sulal; Yardı, Melih Can; Yörük, Erdem; Kina, M. Fuat; Mutlu, Osman; Etgu, Tolga; Koyuncu, Murat; CCSS (Center for Computational Social Sciences); Department of Sociology; Department of International Relations; Graduate School of Social Sciences and Humanities; Yes; College of Administrative Sciences and Economics; College of Social Sciences and Humanities; GRADUATE SCHOOL OF SOCIAL SCIENCES AND HUMANITIES; Research Center
Using artificial intelligence, this article explores the intricate dynamics between ideologies, emotions, and political preferences of the electorate in Turkey. Utilizing a dataset of one billion posts from X (formerly Twitter), the study maps out political opinions, focusing on support for presidential candidates, ideological stances, and collective emotions around the pivotal 2023 Turkish presidential elections. We discuss the limitations of conventional survey techniques and introduce an ERC-funded Politus project that processes digital trace data to offer timely insights into social and political trends. The study's findings, particularly around the "prayer rug (seccade) crisis," underscore the complexity of electoral politics and the potential of digital trace data in capturing the evolving sentiments and ideological orientations of voters. Through this computational approach, the research provides a granular depiction of Turkey's ideological map and electoral behavior, contributing significantly to the discourse on political analysis in the digital era.
Assessing the predictive power of social media data-fed large language models on vote behavior
(Association for Computing Machinery, 2024) Atsızelti, Şükrü; Barkhordar, Ehsan; Graduate School of Social Sciences and Humanities; Graduate School of Sciences and Engineering; No; GRADUATE SCHOOL OF SCIENCES AND ENGINEERING; GRADUATE SCHOOL OF SOCIAL SCIENCES AND HUMANITIES
This article investigates how large language models (LLMs) reflect human preferences and exhibit biases influenced by the diversity and nature of their input data. We used survey data related to Turkish presidential elections alongside tweets to assess the predictive performance and bias manifestations of LLMs under three different data inclusion strategies: (1) using only demographic information, (2) integrating demographic information with tweets, and (3) relying solely on tweets. Our findings reveal that prompts enriched with tweets typically achieve higher F1 Macro scores. However, this trend differs significantly when examining classes individually. While user-generated content significantly improves performance in predictions related to Recep Tayyip Erdogan, it does not show the same effect for Kemal Klllçdaroglu. This study shows that different models and prompting styles result in varied biases for each candidate, leading to mixed outcomes. These results underscore the importance of exploring how biases vary across different scenarios, models, and prompting strategies for each case.
