| Resumen |
This work studies the differential capabilities of hu-man analysts and ChatGPT-4 in dream interpretation, employing lexical features and semantic categories defined by the LIWC tool. A dataset comprising interpretations from 38 analysts across 20 patients provided a basis for comparison, with distinct patterns emerging between human and AI-generated analyses. ChatGPT-4's interpretations showed a marked preference for semantic over grammatical categories, particularly words associated with vision and health. The study demonstrates the utility of Naive Bayes classification in distinguishing between human and AI analyses, achieving a high accuracy rate. Findings indicate that ChatGPT-4 does not present a consistent position relative to the centroid of human analysis, often diverging in its narrative style. The results contribute to the discourse on AI's role in dream analysis, suggesting that while ChatGPT-4 can generate coherent narratives, it exhibits a distinct analytical approach from humans, warranting further research into the integration of AI tools in psychological practices. © 2024 IEEE. |
