Computing Beneficence: a Study of Pro-Social Attitudes in Comments of Online Social Media Users

Beneficence is a social phenomenon that has rarely been modeled computationally so far. In this paper, we propose to study the beneficence of online opinions and comments published on social media on essential topics for society. Our computational approach is based on measuring semantic similarity. We apply three measures to assess the beneficence of $\sim 41 K$ social media users: average Confidence, Normalized Google Distance, and Pointwise Mutual Information. As a use case, we analyze opinions on the topic of vaccinations on Facebook, where two distinct groups (Pro-Vax vs. Anti-Vax) are present. The results reveal a shared connection to beneficence among social media users, with both groups exhibiting similar levels of similarity and no significant clustering into echo chambers.