The Brain Basis of Emotion: A Category Construction Problem

Northeastern University, Boston MA

Investigators

Abstract

Emotions play a central role in human life, yet their neural basis remains poorly understood. Within the science of emotion, it is currently debated whether there is a common brain pattern for a specific emotion category, such as anger, sadness, or fear, or whether there is meaningful variation in the patterns within each category and similarity across categories. An interdisciplinary research team will develop innovative modeling algorithms to learn, rather than stipulate, the number of categories justified by the brain data of each participant. This innovative approach will yield fundamental insights into the nature of emotion in the brain. More broadly, the modeling techniques will provide a means for the human neuroscience community to flexibly investigate psychological categories, without imposing theory-laden assumptions on their data as to the nature of the categories. This work will also help lay the groundwork for a more context-sensitive, personalized approach for relating neural activity with psychological categories. Standard classification approaches to analysis of neuroscience data assume that fixed, experimenter-assigned labels are representative of the ground truth. The present research will question these assumptions and investigate a novel hypothesis that understanding the brain basis of emotion is a category construction problem, not a classification problem. A classification problem assumes that the brain already represents the categories in question, whereas a category construction problem tests this assumption. This project will involve fMRI scanning as participants view emotional videos and subsequently rate their subjective experiences. The research will extensively sample multiple instances of the same emotion category within an individual. The investigators will model the data as a category construction problem by flexibly testing and discovering latent constructs using unsupervised machine learning algorithms combined with empirical model order selection analyses. This research will allow for future researchers to more flexibly and reliably analyze fMRI data, opening up new intellectual schema for understanding how such data reflects brain organization and mental experience in the science of emotion and beyond. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.