Quantifying moral foundations from various topics on Twitter conversations

TL;DR

This paper uses natural language processing on Twitter data to quantify and analyze how different moral foundations are expressed in everyday conversations about moral topics, revealing the prominence of Purity and Care.

Contribution

It introduces a novel method to measure moral foundations from social media conversations, expanding moral psychology research into real-world, spontaneous discourse.

Findings

Purity is the most distinctive moral foundation in Twitter conversations.

Care is the most dominant foundation in discussions about immorality.

Moral foundations are mutually related but function differently in spontaneous speech.

Abstract

Moral foundations theory explains variations in moral behavior using innate moral foundations: Care, Fairness, Ingroup, Authority, and Purity, along with experimental supports. However, little is known about the roles of and relationships between those foundations in everyday moral situations. To address these, we quantify moral foundations from a large amount of online conversations (tweets) about moral topics on the social media site Twitter. We measure moral loadings using latent semantic analysis of tweets related to topics on abortion, homosexuality, immigration, religion, and immorality in general, showing how the five moral foundations function in spontaneous conversations about moral violating situations. The results indicate that although the five foundations are mutually related, Purity is the most distinctive foundation and Care is the most dominant foundation in everyday conversations on immorality. Our study shows a new possibility of natural language processing and social big data for moral psychology.