Can LLMs improve the accuracy of behavior prediction for personnel in high-stakes scenarios by predicting emotions? — A fine-tuning study based on scarce homogeneous cultural data in high-pressure environments
Yibo Chen, Yang Ping, Shuhang Zhou, Caleb Jojo · 2026
Read Can LLMs improve the accuracy of behavior prediction for personnel in high-stakes scenarios by predicting emotions? — A fine-tuning study based on scarce homogeneous cultural data in high-pressure environments by Yibo Chen, Yang Ping, Shuhang Zhou, Caleb Jojo (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/plos-pone-0352988.txt.
Description
Yibo Chen and three colleagues ask, in this 2026 PLOS One study, whether large language models predict people's behavior better when they first predict emotions. The setting is deliberately narrow: personnel in high-pressure, high-stakes roles, drawn from a single cultural background so the data stay consistent. The team built a dataset linking scenarios, emotions, and behaviors, then fine-tuned a language model on it.
The emotion-first model beat both a plain baseline model and a model trained to predict behavior directly, across several accuracy measures. The authors connect this to decision theories that treat emotion as a step in choosing actions. They present the work as a bridge between artificial intelligence and behavioral science, with lessons for forecasting behavior where mistakes are costly.
Catalog record
| CREATOR | Yibo Chen, Yang Ping, Shuhang Zhou, Caleb Jojo |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 43369 |
| LICENSE | CC_BY |
| RIGHTS | CC BY |
| SOURCE | PLOS |