projects


Automatic Morality Classification for Social Media Posts

Large bodies of relatively innocuous texts can be useful for predicting a surprising amount about a person. Previous work has shown that assessment only of the text given in social media profiles can be used to predict a person's OCEAN personality type. Unfortunately, personality types are quite wobbly, and people retest to quite different values.

Moral Foundations Theory reduces human morality down to a series of differently-weighted values. It is one of the more rigorous and experimental approaches to understanding morality.

My proposal is simple: extend the type of work being done to predict personality types to predict MFT values. The major obstacle would be collection of ground truth data, which would take the form of the MFT questionnaire, and would have to be administered to the training/test dataset.