LLMs Replicate Implicit and Explicit Attitudes, Study Finds

A master's dissertation from Utrecht University by Hedwig Oldenhof suggests that large language models (LLMs) can replicate both implicit and explicit attitudes. The research indicates that LLMs demonstrate biases similar to those found in human implicit association tests (IAT) and explicit questionnaires. AI

IMPACT Suggests LLMs may internalize and replicate human biases, impacting AI ethics and development.

RANK_REASON The cluster reports on findings from an unpublished master's dissertation, which constitutes academic research. [lever_c_demoted from research: ic=1 ai=1.0]

Read on Mastodon — mastodon.social →

• Hedwig Oldenhof
• large language models
• Utrecht University

• paper
• safety

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →