Modeling the Spread of Online Hate as a Diffusion Network
Source: https://www.sciencedaily.com/releases/2019/08/190821135258.htm
The article “First of its kind mapping model tracks how hate spreads and adapts online” discusses how researchers at George Washington University are seeking to track online hate as it spreads through social networks. The article explains how “hate clusters,” or groups of people who share a hateful belief, transcend language, social platform, and country, such that they persist even when, for example, they are banned from a particular social media platform. The team is seeking ways to reduce the spread of online hate by exploiting social network structure. For example, they propose reducing the power of large hate clusters by banning smaller clusters that feed into them. They also suggest banning particularly influential users, and adding new clusters that cause the hateful clusters to question their beliefs.
Although the article doesn’t explicitly discuss network diffusion, the spread of hate can clearly be viewed through the lens of diffusion. For example, if we assume that everyone in a network has initially good beliefs, and then introduce a few members with hateful beliefs, this hate will spread through the network according to the structure of the network and value of adopting that hateful belief. For example, let A denote that a node has a good belief and B denote that a node has a hateful belief. When researchers propose banning particularly influential users in order to control the spread of hate, this is like eliminating nodes with belief B who allow for significant spread of B (perhaps, for example, nodes on a local bridge). Adding new, non-hateful clusters to make the hateful groups question their beliefs is similar to inserting more nodes with belief A which are connected to B nodes. With so many more A nodes, the B nodes may convert to A, potentially causing a full cascade to A.