RESEARCH · RESEARCH · #2984
arXiv preprint: The psychophysics of affect in large language models
This arXiv preprint tests how large language models predict how gambling and stock-trading outcomes will feel and whether they use those predicted feelings in choices. Models' isolated outcome ratings showed diminishing sensitivity to magnitude and an unexpected gain-weighting (termed “silicon gain amplification”); ratings linearized when outcomes were presented together, and in choice tasks the gain–loss asymmetry largely disappeared with the models' self-reported feelings explaining choices no better than monetary utility.
KEY POINTS
- This arXiv preprint tests how large language models predict how gambling and stock-trading outcomes will feel and whether they use those predicted feelings in choices.
- Models' isolated outcome ratings showed diminishing sensitivity to magnitude and an unexpected gain-weighting (termed “silicon gain amplification”); ratings linearized when outcomes were presented together, and in choice tasks the gain–loss asymmetry largely disappeared with the models' self-reported feelings explaining choices no better than monetary utility.
- Findings matter because they show LLMs map outcomes to affect in context-dependent ways that could alter how agentic systems predict and use anticipated feelings in decisions.
WHY IT MATTERS
Findings matter because they show LLMs map outcomes to affect in context-dependent ways that could alter how agentic systems predict and use anticipated feelings in decisions.