Internal narratives parameterise affective states
Researchers have developed a method to quantify depressive states by analyzing participants' internal narratives using large language models. Across two studies involving over 1200 participants, they found that verbal descriptions of symptoms contained granular information predictive of depression scores. The study also demonstrated that changes in these quantified narratives could lead to subsequent changes in self-reported affective states, suggesting a computational approach to understanding and potentially treating psychological conditions. AI
IMPACT This research offers a novel computational approach to understanding and potentially treating psychological conditions by leveraging LLMs to analyze internal narratives.