Dissecting the Meme Magic: Understanding Indicators of Virality in Image Memes

Despite the increasingly important role played by image memes, we do not yet have a solid understanding of elements that might make meme go viral on social media. In this paper, investigate what visual distinguish memes are highly media from those get re-shared, across three dimensions: composition, subjects, and target audience. Drawing research in art theory, psychology, marketing, neuroscience, develop codebook to characterize use it annotate set 100 collected 4chan's Politically Incorrect Board (/pol/). On one hand, find more likely close-up scale, contain characters, include positive or negative emotions. other present clear subject viewer can focus attention on, long text be re-shared users. We train machine learning models between unlikely obtaining an AUC 0.866 our dataset. also show indicators virality identified model help most posted mainstream online networks too, as classifiers able predict 19 out 20 popular Twitter Reddit 2016 2018. Overall, analysis sheds light non-viral content online, basis for developing better techniques create moderate is catch viewer's attention.