A new audit reveals that large language models (LLMs) display a more favorable bias towards governments in countries with less media freedom. This effect is more pronounced in languages spoken in these regions compared to those from countries with higher media freedom. The research suggests a link between dataset control and the LLM's output, mirroring observations in historical prompt responses. AI
IMPACT Highlights potential biases in LLMs related to geopolitical and media freedom contexts, impacting global information access and fairness.
RANK_REASON The cluster describes a research paper presenting findings from a cross-national audit of LLMs. [lever_c_demoted from research: ic=1 ai=1.0]
Read on Mastodon — fosstodon.org →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
