A Study on the Polarisation Effects of Biased Conversational Agents in Online Debates

Abstract

Artificial Intelligence is revolutionising our communication practices and the ways in which we interact with each other. This does not only impact how we communicate, but affects the nature of the partners with whom we communicate. Online platforms now allow humans to communicate with artificial agents in the form of socialbots or chatbots. Such agents have the potential to persuade humans and polarise opinions. We propose to investigate how a conversational agent influences the evolution of an online debate. We look at the persuasive consequences associated with an agent that uses persuasive cues to manipulate debaters. The agent will react to the arguments of the debaters by either supporting or attacking anyone that agrees or disagrees with the stance. The results suggest that when the debaters have prior a knowledge on the issue, their stances do not change under the effect of the agent. On the other hand, debaters with no knowledge of the issue did change their stances.