LLMs show consistent, but more extreme, social attraction judgments than humans

A new study published on arXiv investigates the ability of large language models (LLMs) to judge social attraction. Researchers constructed persona profiles based on psychological and relational constructs, categorizing them into socially attractive, mixed, and unattractive tiers. In two studies, 34 LLMs consistently rated these profiles, showing stability across runs and high agreement in relative ordering, though they tended to rate attractive profiles more positively and unattractive ones more negatively than human participants in a third study. The LLMs did not show significant effects related to gender presentation in their ratings. AI

IMPACT LLMs demonstrate a consistent, though potentially more polarized, ability to judge social attraction compared to humans, raising questions about their use in subjective evaluations.

RANK_REASON The cluster contains an academic paper detailing research findings on LLM capabilities. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• arXiv
• Humans
• Khawaja Abaid Ullah
• Large language models
• LLMs

• paper
• other

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