Method for identifying trolls in online communities

Abstract

In the article the problem of recognizing users of social networks, chats and other virtual spaces that are provoked by other users, inciting conflicts between participants of various online communities is investigated. In this work the authors give a brief description of the trolling concept. The relevance of solving the problem of trolling in the social communities of the Internet is shown in connection with the widespread aggressive provocative behavior of individual users in the virtual space, as well as the influence of such behavior on the psyche and general condition of a person. The paper uses the theory of trust functions to process expert observations, estimates or measurements. The proposed approach involves the use of this theory in conjunction with the input of a value that determines the conflict between two combined trust functions. The goal of the work is to test a new approach for calculating potential trolls using a method based on the degree of conflict of trust functions between different messages of the discussion branch in question-answer communities. It is concluded that the possibility of assessing the user’s conflict with the help of a parameter called the measure of the conflict of his messages in relation to the messages of all other users. To simplify the calculation process, the k-means clustering algorithm is applied.