AI prompts offer a new tool for mapping how emotion concepts relate
How similar are "joy" and "surprise"? Conversely, how far apart are "joy" and "sadness"? We naturally organize words that express emotions—such as "happy," "afraid," and "angry"—according to similarities and differences in their meanings. However, investigating this entire structure using human participants alone is not easy because the number of comparisons rises rapidly as more emotion terms…
Researchers at the National Institute for Basic Biology have developed a method to investigate the semantic organization of emotion terms using the AI language model GPT-4. The study, published in Scientific Reports, compared the arrangement of 99 emotion terms produced by GPT-4 with those made by human participants, finding that both groups produced similar structures.
The researchers presented three emotion terms simultaneously on a grid, allowing them to express relationships between the terms in a single trial. By expanding the vocabulary to 99 terms, the researchers were able to detect a structure related to emotional intensity, or arousal, which had been difficult to discern with smaller sets of emotion terms.
This arousal-related structure emerged more clearly when using GPT-4's responses compared to conventional word embeddings. While GPT-4 does not experience emotions itself, the study suggests that it captures semantic relationships among emotion concepts embedded in human language. The research team next plans to expand the analysis to other languages and explore how cultural differences may influence the arrangement of emotion terms.
Written by urgent.news from Phys.org's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
