Neurophilosophy Forum

Title: Towards identifying kinds of human emotion with machine learning

Abstract: Debates concerning human emotion have centered around whether emotions are natural kinds or if they are collections of highly variable instances that are categorized in the moment. Although many studies have examined patterning of antecedent events, brain responses, and self-reported experience during emotional episodes, they mostly focus on whether it is possible to map such measures onto distinct categories. It is unclear whether categories identified in experimental studies are useful for scientific inference. In this talk, I describe how this challenge can be reframed as the machine learning problem of induction: learning general principles from a limited number of observations. I present research showing that the statistical regularities of emotional events enable precise inferences about human behavior and brain function that generalize across individuals and studies. I argue that examining the generalizability of inductive models can ultimately reveal the extent to which common sense categories like ‘fear’, ‘anger’, and ‘sadness’ are natural categories or psychological constructions. This machine learning framework allows for new, objective definitions of emotion categories and promises to resolve longstanding contention about human emotion.

Title: TBD

Abstract: TBD