The emotional trajectory of non-suicidal self-injury: sentiment analysis using social media data

Abstract Studies using self-report data have shown that emotion is related to non-suicidal self-injury (NSSI). However, the specific roles of emotion in the initiation and continuation of this type of behavior remain unmapped. This study conducted a sentiment analysis of social media posts to map the emotional trajectory of NSSI behavior in China. We collected data from 462,287 social media posts by 398 females who disclosed their NSSI behavior on Weibo, mainland China’s most popular social media platform. Using a lexicon-based sentiment analysis approach, we assigned sentiment scores to each post at the person-per-date level, then subjected these scores to latent growth modeling to map the emotional trajectory of NSSI behavior. During the four days preceding NSSI disclosure, the Weibo users showed significant increases in arousal ( β = .317; p = .014), positive emotions ( β = .175; p = .022), and negative emotions ( β = .805; p = .032). During the four days following NSSI disclosure, they experienced significant decreases in positive emotions compared with the preceding four days ( diff β = .318; p = .003), with no significant changes in negative emotions or arousal. Our findings indicated that the levels of arousal, positive emotions, and negative emotions all rose in the four days preceding NSSI disclosure. However, contrary to the common notion that NSSI may improve mood, our results showed that positive emotions decreased following the disclosure.