Relating conversational topics and toxic behavior effects in a MOBA game

Abstract Multiplayer Online Battle Arena (MOBA) games are very competitive and victory relies on effective teamwork. Toxic behavior is a common concern, as it corrodes team effort and harms game ambiance. Related work has focused on the automatic detection of toxic behavior, mostly based on features from the communication channels used by players. In this paper, we investigate the conversational patterns used by players of a popular MOBA game, League of Legends, and their effects on performance and contamination. We identify the main conversation topics exchanged by players using in-game chats; characterize the behavior of each type of player according to these topics; and develop an in-depth analysis of how they affect game performance and toxic contamination. Our findings show that some topics are directly related to toxic behavior. The allies of a toxic player are in general more affected by toxic behavior than the opponents, and in some cases, they act together as a source of toxic behavior. Opponents are more affected when the toxic behavior is directly targeted at them (e.g. racist slurs). Players with no significant contact with toxic players tend to be more positive, concentrating on game tactics and socialization.