Search Papers | Poster Sessions | All Posters
Poster B152 in Poster Session B - Thursday, August 8, 2024, 1:30 – 3:30 pm, Johnson Ice Rink
Towards a Latent Space Cartography of Individual Differences in Subjective Experience Using Large Language Models
Shawn Manuel1 (), Frédéric Gosselin, Jean Gagnon, Vincent Taschereau-Dumouchel; 1University of Montreal
Many theoretical perspectives posit that cognitive and affective functioning is greatly determined by how individuals subjectively experience the world. However, characterizing the breadth and depth of human experience remains a considerable challenge. One persistent problem is the lack of objective tools for quantifying and comparing narrative reports of subjective experiences. Here, we develop a new approach to map and compare reports of experience using modern large language models (LLMs). Using a series of 20 image prompts, we quantified how the verbal reports of experience provided by participants (n=210) deviate from one another and how these variations are linked to subjective experience and cognitive-affective profiles. We found that latent space embeddings of experience can accurately predict subjective valence and arousal judgments in a series of emotional pictures, as well as cognitive-affective profiles determined using computational factor modeling. As such, latent space cartography of experience could offer a promising avenue for objectively quantifying distortions of subjective experiences and ultimately linking them to patterns of neural activity.
Keywords: latent space cartography individual differences subjective experience large language models