AI Outshines Humans in Emotional Intelligence Tests: Study Reveals Startling Results

Researchers at the University of Geneva and University of Bern Find Generative AI Models Dominate Emotional Judgment Assessments

Published by Lee Bell on 31/05/2025

Can AI Outshine Humans in Emotional Intelligence? A Closer Look

By Lee Bell | 31/05/2025 | 10:00 AM | 3 min read

In a surprising turn of events, researchers have unveiled that advanced artificial intelligence models have the potential to outperform humans in emotional intelligence (EI) assessments. This finding, as unsettling as it may seem, has been brought to light by teams from the University of Geneva and the University of Bern, who rigorously tested six leading large language models (LLMs), including popular platforms like ChatGPT, Gemini, and Claude.

Outperforming Humans

The crux of the research involved evaluating the AI models using five standard emotional intelligence tests designed to simulate complex workplace dilemmas, assessing their ability to understand, regulate, and respond to emotional situations. Shockingly, the AI systems didn’t just meet expectations; they surpassed human performance significantly, scoring an impressive 82% on the questions compared to a mere 56% for the human participants.

Marcello Mortillaro, a senior scientist at Geneva’s Swiss Center for Affective Sciences, stated, “In the end, the LLMs achieved significantly higher scores. This suggests that these AIs not only understand emotions, but also grasp what it means to behave with emotional intelligence.” This revelation hints at an evolving perception of AI—not merely as tools for data processing but as competent entities in emotional reasoning.

A Shift in the Role of AI?

The second phase of the study leveraged ChatGPT-4 to generate entirely new scenarios for emotional intelligence assessments. The AI’s ability to create realistic test items proved remarkable, with over 400 individuals participating in the new evaluations. Researchers were astounded by the quality and reliability of the results from these AI-generated scenarios, which rivaled those developed through years of research.

Katja Schlegel, lead author and lecturer at the University of Bern, remarked, “They proved to be as reliable, clear and realistic as the original tests, which had taken years to develop.” This raises intriguing possibilities: not only can LLMs identify the best answers among existing options, but they can also generate context-sensitive scenarios, marking a significant leap in their utility.

As these findings suggest AI’s capacity to navigate human emotions in meaningful ways, discussions regarding their role are rapidly shifting. Beyond creative fields where they are presently utilized, applications could extend into education, coaching, and even conflict resolution.

Conclusion

The implications of this research are vast and multifaceted. While concerns may arise over the capabilities of AI in the emotional domain, it opens doors to innovative solutions in various sectors. As we continue to explore the realm of emotional intelligence within AI, the question remains: how will society adapt to these advanced technologies that may soon rival human emotional understanding? The answer could shape the future, blending human nuance with AI efficiency.

For more information, refer to the research published in Communications Psychology, May 2025.