English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Zendy Plus

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Zendy Tools

Zendy Open

The rapid advancement of Large Language Models (LLMs) has sparked intensedebate regarding the prevalence of bias in these models and its mitigation.Yet, as exemplified by both results on debiasing methods in the literature andreports of alignment-related defects from the wider community, bias remains apoorly understood topic despite its practical relevance. To enhance theunderstanding of the internal causes of bias, we analyse LLM bias through thelens of causal fairness analysis, which enables us to both comprehend theorigins of bias and reason about its downstream consequences and mitigation. Tooperationalize this framework, we propose a prompt-based method for theextraction of confounding and mediating attributes which contribute to the LLMdecision process. By applying Activity Dependency Networks (ADNs), we thenanalyse how these attributes influence an LLM's decision process. We apply ourmethod to LLM ratings of argument quality in political debates. We find thatthe observed disparate treatment can at least in part be attributed toconfounding and mitigating attributes and model misalignment, and discuss theconsequences of our findings for human-AI alignment and bias mitigation. Ourcode and data are at https://github.com/david-jenny/LLM-Political-Study.

Exploring the Jungle of Bias: Political Bias Attribution in Language Models via Dependency Analysis