Large Language Models Reinforce Positivity Bias Toward Personal Future in Humans

Poster Session X

Keyword: Other

Abstract: This study investigates emotional tendencies in popular human queries and the corresponding responses from three Large Language Models (LLMs). Regression analyses reveal that humans show a positivity bias toward personal future events (n=60), and LLMs align emotionally with prompts (n=300). These emotional tendencies may reinforce unrealistic optimism during human-LLM interactions.

• Ya-Jie Tai (Presenting Author)
• Yu Ching Chen (Author)
• Tsung-Ren Huang (Author)
• Suparna Rajaram (Author)