AI struggles to tell stories across cultures
Summary and headline written by AI from the source article. How we work
Researchers evaluated how well artificial intelligence understands morals in stories from different cultures. They created a dataset of story morals written by people in 14 languages and compared those to morals generated by AI models like GPT-4o and Gemini.
The AI produced morals similar to human responses and generally pleased human reviewers. However, the AI tended to offer less variety in its morals across different languages and focused on values widely accepted around the world.
This suggests current AI can grasp common moral ideas, but it doesn't fully capture the range of interpretations people have when reading stories. The researchers propose using narrative interpretation as a way to assess cultural understanding in AI, moving beyond simple tests of knowledge. This approach could help developers build AI that better reflects diverse cultural values.
