Measuring the Centrality of Authors in a Reply Tree
In our project we are interested in how and why certain users try to moderate online discussions. One idea was to find these motivated users by analyzing the reply tree structures. The assumption being that the more central a user is to the discussion (frequency of posts and their distribution throughout the discussion) the more likely it would be to see moderating behaviour.
Although this approach did not yield immediate results, some interesting tools and research bits were produced which we are happy to present here:
First, we would like to to draw attention to the python library delab_trees which was created in order to analyze the reply tree structures. It can load, validate and analyze millions of reply trees in parallel and can even predict which author is most likely to write next in a given conversation. It is still incubating but we are continously adding new features and testing existing ones. Please feel free to fork the github project and make this open source project fly.
Second, we published a preprint which details some of the algorithms and their application. It is a rather technical paper but it shows significant differences between reply trees in Twitter and Reddit as to the involvement of their authors.
Third, we applied these measurements to the topic of climate change and analyzed how different strategic actor groups are positioned in social media discussions. The preprint is also available online. Please feel free to contact
Research is always a journey and the small stuff matters.