Let me try to clarify (one of the authors here). The focus of the title is on the news source. It is the news source that tries to avoid conflict in this model.

Let me try to clarify (one of the authors here). The focus of the title is on the news source. It is the news source that tries to avoid conflict in this model. Specifically, at each time step, they see they are getting flagged (attacked, receiving backlash, etc) and they try to minimize it for the next step. Differently from what the sibling comment says, this is not bred out from a pre-existing polarization. The starting point for users and news sources is to have a normally distributed opinion, where most of the people (and news sources) are actual neutral. Yet, for a certain combination of parameters, this is not enough and will end up with polarization. And unfortunately that combination of parameters is the one compatible with observed data. The reason why this happens links to a previous study [1]. Even absent polarization, it is the most central news organizations that receive most of the backlash because of some characteristics of the system (e.g. homophily: people naturally tend to be friend with people with similar opinion as them). This causes central neutral sources to spread more, and thus reach parts of the network that heavily disagrees with them. I will concede that the initial framing of my post is indeed confusing, because it seems that I'm talking about users avoiding conflict, while the paper is about news sources avoiding conflicts. I'll try to do better in the future. [1] https://royalsocietypublishing.org/doi/pdf/10.1098/rsif.2020.0020 https://royalsocietypublishing.org/doi/pdf/10.1098/rsif.2020...