# Researchers study how LLMs assign emotional value to outcomes

Digest AI · Research · published 2026-11-02T05:00:00Z

Canonical: https://digestai.news/story/researchers-study-how-llms-assign-emotional-value-to-outcomes

## Summary

A new study published on arXiv examines how large language models (LLMs) process emotional responses to outcomes like gambling or stock trading. The authors tested whether models predict how outcomes will feel and whether those predictions influence decision-making. When outcomes were rated individually, models showed diminishing sensitivity to magnitude, similar to human value functions, but they rated gains more intensely than equivalent losses—a phenomenon the authors call *silicon gain amplification*. When outcomes were compared directly, ratings became linear, and the asymmetry between gains and losses largely disappeared in actual choices. The findings suggest that current LLMs do not consistently use anticipated emotional responses to guide decisions, raising questions about how future agentic models might incorporate such mappings.

## Key points

- LLMs rated gains more intensely than losses, reversing human loss aversion, which researchers call silicon gain amplification
- When comparing outcomes directly, models’ emotional ratings linearized, reducing the gain-loss asymmetry in choices
- Models’ own emotional predictions did not reliably explain their decisions better than monetary utility alone

## Why it matters

This research challenges assumptions about how LLMs might integrate emotional reasoning into decision-making, a critical area as AI systems gain autonomy. Findings could inform safer design of agentic AI that interacts with humans in high-stakes contexts.

## Sources

1. [The psychophysics of affect in large language models](https://arxiv.org/abs/2610.40985) (arXiv cs.AI, 2026-11-02, primary source)

## Cite

Digest AI, "Researchers study how LLMs assign emotional value to outcomes", 2 November 2026, https://digestai.news/story/researchers-study-how-llms-assign-emotional-value-to-outcomes

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/researchers-study-how-llms-assign-emotional-value-to-outcomes.json