AI vs AI vs Humans (2025)

This research explores large language models (LLMs) as tools to promote pro-environmental behavior, comparing their impact across 3,200 participants: real humans, simulated humans based on actual participant data, and fully synthetic personas. Results reveal a 'synthetic persuasion paradox': synthetic agents significantly change their behavior post-intervention, while human responses barely shift. This disconnect underscores LLM's potential for pre-evaluating interventions but warns of limits in predicting real-world behavior.

Open external link