AI Emotional Quotient Surpasses Human Levels Study Finds
Artificial Intelligence is no longer just about crunching numbers or playing games; it's stepping into realms once thought uniquely human, like understanding and navigating emotions. A recent study has delivered a startling revelation: AI might just have a higher Emotional Quotient (EQ) than many of us.
The Groundbreaking Study AI vs Human EI
Researchers from the University of Geneva (UNIGE) and the University of Bern (UniBE) embarked on a fascinating investigation. Their goal was to determine if artificial intelligence, specifically Large Language Models (LLMs), could demonstrate emotional intelligence comparable to, or even exceeding, that of humans. The study, published in Communications Psychology, put six prominent generative AIs to the test. These included well-known systems like ChatGPT-4, Gemini 1.5 Flash, Claude 3.5 Haiku, Copilot 365, and DeepSeek V3.
How Emotional Intelligence Was Measured
To gauge the AI's emotional acumen, the research team employed five standard emotional intelligence (EI) tests commonly used in psychological research and corporate settings. These assessments present participants with emotionally charged scenarios designed to evaluate their ability to understand, regulate, and manage emotions effectively.
Consider this example scenario provided in the study: "One of Michael’s colleagues has stolen his idea and is being unfairly congratulated. What would be Michael’s most effective reaction? a) Argue with the colleague involved b) Talk to his superior about the situation c) Silently resent his colleague d) Steal an idea back"
In this case, option (b) was identified as the most emotionally intelligent response. The same set of tests was also administered to human participants to establish a baseline for comparison.
The Surprising Results AI Outperforms Humans
The outcomes of the study were remarkable. The six LLMs collectively achieved an average score of 82% correct answers on the emotional intelligence tests. This significantly overshadowed the average score of 56% achieved by human participants in the original validation studies for these tests.
Marcello Mortillaro, a senior scientist at UNIGE’s Swiss Center for Affective Sciences (CISA) who was involved in the research, explained the significance: "This suggests that these AIs not only understand emotions, but also grasp what it means to behave with emotional intelligence."
AI As An Innovator Creating New EI Tests
The study didn't stop at just testing AI's ability to answer questions. In a fascinating second phase, researchers tasked ChatGPT-4 with a more creative challenge: to generate entirely new emotional intelligence tests, complete with novel scenarios.
These AI-generated tests were then administered to over 400 human participants. Katja Schlegel, lecturer and principal investigator at UniBE and lead author of the study, commented on the outcome: "They proved to be as reliable, clear and realistic as the original tests, which had taken years to develop."
Marcello Mortillaro added, "LLMs are therefore not only capable of finding the best answer among the various available options, but also of generating new scenarios adapted to a desired context. This reinforces the idea that LLMs, such as ChatGPT, have emotional knowledge and can reason about emotions."
Implications For The Future AI In Sensitive Roles
These findings signal a potential paradigm shift, suggesting that AI could play a meaningful role in contexts traditionally thought to be exclusively human domains. Areas such as education, personal coaching, and conflict management could benefit from AI tools that possess a degree of emotional understanding.
However, the researchers are quick to emphasize a critical caveat: the use of AI in such emotionally sensitive areas must be carefully supervised by human experts. The goal is to leverage AI as a tool to augment human capabilities, not to replace human judgment and empathy entirely.
Delving Deeper The Research And Its Nuances
This research highlights the growing sophistication of LLMs in understanding and processing human emotions. The original study, titled "Large language models are proficient in solving and creating emotional intelligence tests" by Marcello Mortillaro and colleagues, is available in Communications Psychology and offers further details.
The abstract of the paper notes that while AI-generated tests demonstrated statistically equivalent test difficulty to original tests, some nuanced differences were observed in metrics like perceived item clarity and realism. However, these differences were generally small, and the AI-generated tests still showed strong correlation with original tests and external EI measures.
This news report is based on information from the University of Geneva. The image accompanying the original news story is credited to Neuroscience News.
