Considering the Elaboration Likelihood Model for simulating hate and counter speech on Facebook

Abstract

Counter speech is perceived and has also been advocated by social networks as a measure for delimiting the effects of hate speech. To facilitate estimating the efficiency of counter speech in freely accessible blackboard communication as employed by Facebook, we extend an existing simulation model by integrating Elaboration Likelihood Model (ELM) mechanisms. We model four different user groups (core, clowns, followers and counter speakers), each with a specific set of properties, namely need for cognition and involvement as ELM personal characteristics, and opinion, volatility and activity as borrowed from opinion formation models. We also add argument strength as important message characteristic. Our simulation experiments show that the updated model provides similar but much more detailed results: potentially temporal opinion changes via peripheral processing get visible. Furthermore, we give more evidence that the opinions of counter speakers shall not be too extreme in all cases but sometimes rather moderate in order to achieve maximum impact.