Two new research papers explore the political leanings and societal modeling capabilities of large language models (LLMs). The first paper demonstrates that LLMs can reconstruct aggregate voting behavior from sociodemographic profiles, accurately predicting election outcomes and political structures. The second paper argues that LLMs exhibit political biases, not by being moderate overall, but by presenting strongly partisan views on specific topics, and can exert significant persuasive influence on users' political attitudes, especially when steered. AI
IMPACT These studies highlight LLMs' potential for both understanding and influencing societal dynamics, raising questions about their role in shaping public opinion and political discourse.
RANK_REASON Two academic papers published on arXiv detailing research into LLM capabilities and biases.
- 2021 Czech parliamentary election
- alphaXiv
- arXiv
- CatalyzeX
- DagsHub
- Gotit.pub
- Hazem Ibrahim
- Hugging Face
- large-language models
- ScienceCast
- U.S.
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
