With this project, we explored whether Large Language Models (LLMs) trained on different corpora of data generate politically biased headlines. In the context of an increasingly politically polarized US, the potential influence of social media companies (e.g. Facebook) over their users’ political behavior through consumption of their algorithmic products is essential to consider within the context of data ethics.
Jupyter notebooks containing analysis and research findings.
CNN Fine-Tuned Model FOX Fine-Tuned Model GPT Political Sentiment Analysis LLM Testing NLP Sentiment Analysis