Study finds LLMs exhibit 'silicon gain amplification' in affect ratings

A new study published on arXiv explores how large language models (LLMs) perceive and incorporate 'feelings' or affect into decision-making processes, particularly in scenarios like gambling or stock trading. Researchers found that LLMs exhibit a phenomenon termed 'silicon gain amplification,' where they rate gains more intensely than equivalent losses, contrasting with human loss aversion. However, this affective asymmetry diminished when models directly compared outcomes, and their self-rated feelings did not consistently predict their choices better than objective monetary utility. The findings raise important questions about how agentic LLMs should integrate anticipated emotions into their decision-making. AI

IMPACT Raises questions about how LLMs should incorporate emotions into decision-making, impacting future agentic AI development.

RANK_REASON Academic paper published on arXiv detailing novel findings about LLM behavior. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• arXiv
• CatalyzeX Code Finder for Papers
• Connected Papers
• CORE Recommender
• DagsHub
• Gotit.pub
• Hugging Face
• Influence Flower
• large language models
• Litmaps
• ScienceCast
• scite Smart Citations
• silicon gain amplification

• paper
• safety

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