Quantifying Latent Moral Foundations in Twitter Narratives: The Case of\n the Syrian White Helmets Misinformation

For years, many studies employed sentiment analysis to understand the\nreasoning behind people's choices and feelings, their communication styles, and\nthe communities which they belong to. We argue that gaining more in-depth\ninsight into moral dimensions coupled with sentiment analysis can potentially\nprovide superior results. Understanding moral foundations can yield powerful\nresults in terms of perceiving the intended meaning of the text data, as the\nconcept of morality provides additional information on the unobservable\ncharacteristics of information processing and non-conscious cognitive\nprocesses. Therefore, we studied latent moral loadings of Syrian White\nHelmets-related tweets of Twitter users from April 1st, 2018 to April 30th,\n2019. For the operationalization and quantification of moral rhetoric in\ntweets, we use Extended Moral Foundations Dictionary in which five\npsychological dimensions (Harm/Care, Fairness/Reciprocity, In-group/Loyalty,\nAuthority/Respect and Purity/Sanctity) are considered. We show that people tend\nto share more tweets involving the virtue moral rhetoric than the tweets\ninvolving the vice rhetoric. We observe that the pattern of the moral rhetoric\nof tweets among these five dimensions are very similar during different time\nperiods, while the strength of the five dimension is time-variant. Even though\nthere is no significant difference between the use of Fairness/Reciprocity,\nIn-group/Loyalty or Purity/Sanctity rhetoric, the less use of Harm/Care\nrhetoric is significant and remarkable. Besides, the strength of the moral\nrhetoric and the polarization in morality across people are mostly observed in\ntweets involving Harm/Care rhetoric despite the number of tweets involving the\nHarm/Care dimension is low.\n