AI Outperforms Humans in Emotional Intelligence Assessments
Researchers from the University of Geneva and the University of Bern conducted a groundbreaking study comparing the emotional intelligence (EI) of AI models with that of humans. The results were surprising: AI models not only matched but exceeded human performance in EI assessments.
The Experiment
The research team tested six AI models, including ChatGPT, using five well-established emotional intelligence assessments commonly used in research and business settings. These tests presented emotionally charged scenarios to evaluate how effectively the AI understood, regulated, and managed emotions. One such scenario involved a situation where a colleague stole an idea and received unfair congratulations. The AI was asked to choose the most appropriate reaction from four options: arguing with the colleague, talking to a superior, silently resenting the colleague, or retaliating by stealing an idea. The correct response, talking to a superior, demonstrated a professional and fair approach to resolving the situation.
AI’s Superior Performance
The results showed that AI models scored significantly higher than humans, achieving an average of 82% correct answers compared to 56% for humans. This suggests that AI can understand emotions and respond intelligently in complex social situations. To further validate their findings, the researchers had ChatGPT-4 generate new emotional intelligence tests, which were then administered to over 400 people. The AI-generated tests were found to be just as reliable, clear, and realistic as the original assessments that had taken years to develop.
Implications for Future Applications
These findings have significant implications for various fields, particularly education, coaching, and conflict resolution. The ability of Large Language Models (LLMs) like ChatGPT to not only solve emotional intelligence tests but also create new context-specific assessments demonstrates their potential in AI-assisted decision-making and education. The study suggests that LLMs possess emotional knowledge and can effectively reason about emotions, marking a promising step forward in AI development.
The study’s results, published in Communication Psychology, highlight the advancements in AI’s emotional intelligence. As AI continues to evolve, its applications in understanding and managing human emotions could lead to more sophisticated tools for personal development and professional training.
Journal Reference: Schlegel, K., Sommer, N.R. & Mortillaro, M. Large language models are proficient in solving and creating emotional intelligence tests. Commun Psychol 3, 80 (2025). DOI: 10.1038/s44271-025-00258-x
