Empathy Alignment in Online Communities

Abstract

Inspired by the phenomenon that a growing number of people are talking about their own hard times on social media to seek for comfort and advice, the project focuses on how to be more empathetic in our responses to those posts. We refer to the appraisal theory to find the alignment in conversations where the response is closely related to the post. This is a sign of empathy, defined as a person feels the same way as another person who is experiencing some situation. We developed a codebook for labeling dataset and a website tool to visualize and compare annotations across multiple annotators. We also proposed a deep-learning model to detect the alignments for future analysis.