What is it about?
Creativity is a key skill in problem-solving and innovation, and understanding how it can be assessed is vital. In this study, we used a method called textual forma mentis networks to analyze how humans and AI (specifically GPT-3.5) evaluate the creativity of short stories. These networks measure the relationships between words in stories and their emotional tones. Our findings revealed that while humans focus on both emotional depth and structural connections between ideas, GPT-3.5 ratings often differ. For instance, GPT-3.5 tended to favor its own stories, emphasizing emotional features like joy or fear, over structural creativity. This suggests that AI has a different "understanding" of creativity compared to humans.
Featured Image
Photo by Growtika on Unsplash
Why is it important?
As AI becomes more involved in education and creative fields, it’s important to understand how it evaluates creativity. This study shows that AI, like GPT-3.5, sees creativity differently from humans, often focusing more on emotions than how ideas connect. These differences highlight the need to be careful when relying on AI for judging creativity, ensuring it works more like humans before being widely used.
Perspectives
Creativity is such a deeply human skill, and studying how AI understands it was really fascinating. For me, this research highlighted how complex human creativity is and how hard it is for machines to truly grasp it.
Edith Haim
Universita degli Studi di Trento
Read the Original
This page is a summary of: Forma mentis networks predict creativity ratings of short texts via interpretable artificial intelligence in human and GPT-simulated raters, November 2024, Center for Open Science,
DOI: 10.31234/osf.io/6zpre.
You can read the full text:
Contributors
The following have contributed to this page
