When Do Language Models Endorse Limitations on Human Rights Principles?
This research is significant for policymakers and AI developers to understand and mitigate biases in LLMs regarding human rights principles, especially given their potential to shape public discourse.
This paper evaluated how 11 major LLMs navigate trade-offs involving the Universal Declaration of Human Rights (UDHR) across 1,152 scenarios in 24 rights articles and eight languages. The analysis revealed systematic biases, including higher acceptance of limiting Economic, Social, and Cultural rights, significant cross-linguistic variation with higher endorsement rates in Chinese and Hindi, and susceptibility to prompt steering.
As Large Language Models (LLMs) increasingly mediate global information access with the potential to shape public discourse, their alignment with universal human rights principles becomes important to ensure that these rights are abided by in high stakes AI-mediated interactions. In this paper, we evaluate how LLMs navigate trade-offs involving the Universal Declaration of Human Rights (UDHR), leveraging 1,152 synthetically generated scenarios across 24 rights articles and eight languages. Our analysis of eleven major LLMs reveals systematic biases where models: (1) accept limiting Economic, Social, and Cultural rights more often than Political and Civil rights, (2) demonstrate significant cross-linguistic variation with elevated endorsement rates of rights-limiting actions in Chinese and Hindi compared to English or Romanian, (3) show substantial susceptibility to prompt-based steering, and (4) exhibit noticeable differences between Likert and open-ended responses, highlighting critical challenges in LLM preference assessment.