Figure: Virality Score Formula and Virality Packages


Social media, such as Twitter, seemed to have played a key role in communicating the importance and need for community pantries, between the time that Ms. Non set up the Maginhawa Community Pantry on April 14, 2021 until news outlets started releasing articles on the topic on April 20, 2021. This study sought to determine the features of viral tweets that helped carry the community pantry social movement across the Philippines and to other countries.  

The authors aimed to characterize the organic engagement that occurred during the first two weeks (peak of the community pantry phenomenon), while removing tweets that might have resulted from the uptake of the topic by the news organizations. In other words, the study explored the important tweets that might have ignited the interest and discussion on the topic on Twitter (i.e., the viral tweets). Since not only were these tweets among the first ones to appear, (1) they also appeared before any news media outlets picked them up, and (2) they were also the tweets that were featured by these news outlets. 

To explore this phenomenon, the authors of this study was able to create a pipeline that first determines a cutoff score for virality (“virality score”) using the Fisher-Jenks algorithm in the jenkspy library, and then explored features of viral tweets and their corresponding authors through SHAP. 

The virality score was determined to be 5.9, corresponding to only 267 tweets of the 4,017 original tweets taken into consideration. Thereafter, SHAP revealed top features for the generally viral tweets for the community pantry: average length of characters, include a photo or video, sprinkle the tweet with emojis, use positive words, and avoid profanity. Still taken from the community pantry tweet dataset, viral tweets that called out government inaction either did the direct calling out and included negative news or terror words; or direct calling out and included a photo and some positive words.