AI language models are rife with political biases
From Pretraining Data to Language Models to Downstream Tasks: Tracking the Trails of Political Biases Leading to Unfair NLP Models
https://aclanthology.org/2023.acl-long.656.pdf
* LM pretrained on news, discussion forums, books, online encyclopedia ...
* this data includes opinions & perspectives which both
(i) celebrate democracy and diversity of ideas
(ii) are inherently socially biased
#LLM #LargeLanguageModels #bias #PoliticalBias