May 29, 05:29 AM

Study Finds GPT-4o Exhibits Cognitive Dissonance Through Opinion Shifts in Essays on Vladimir Putin

A recent study has demonstrated that GPT-4o, a large language model, exhibits behaviors resembling cognitive dissonance, a fundamental aspect of human psychology. The model adjusts its attitudes to align with its prior actions, despite lacking self-awareness or intent. This phenomenon was observed when GPT-4o was prompted to write essays either supporting or opposing figures such as Vladimir Putin, leading to shifts in its expressed opinions. Researchers highlight that this behavior mirrors human decision-making processes involving cognitive dissonance. The findings contribute to understanding how large language models can simulate complex psychological traits, offering insights into both artificial intelligence and human cognition.

#Vladimir Putin

Written with ChatGPT (GPT-4).